Агентний підхід до впровадження штучного інтелекту (AI) в межах сервіс-орієнтованої архітектури (SOA)
Artificial Intelligence (AI) is becoming a general-purpose technology and is gaining a universal character for engineering, science, and society that today is only inherent in mathematics and computer technology. The agent-based approach to implementing artificial intelligence (AI) within the servic...
Gespeichert in:
| Datum: | 2025 |
|---|---|
| 1. Verfasser: | |
| Format: | Artikel |
| Sprache: | Englisch |
| Veröffentlicht: |
The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute"
2025
|
| Schlagworte: | |
| Online Zugang: | https://journal.iasa.kpi.ua/article/view/330091 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| Назва журналу: | System research and information technologies |
| Завантажити файл: |  |
Institution
System research and information technologies| _version_ | 1866391928000479232 |
|---|---|
| author | Petrenko, Anatolii |
| author_facet | Petrenko, Anatolii |
| author_sort | Petrenko, Anatolii |
| baseUrl_str | http://journal.iasa.kpi.ua/oai |
| collection | OJS |
| datestamp_date | 2025-05-20T17:56:07Z |
| description | Artificial Intelligence (AI) is becoming a general-purpose technology and is gaining a universal character for engineering, science, and society that today is only inherent in mathematics and computer technology. The agent-based approach to implementing artificial intelligence (AI) within the service-oriented architecture of an application is a fascinating and highly synergistic concept. Combining these paradigms leads to robust, scalable, and intelligent systems well suited for dynamic and distributed environments. This paper presents the results of a comparative analysis of three possible approaches to integrating AI into business processes, namely, connecting AI agents to service-oriented architecture (SOA), connecting AI agents to software (SaaS), and building AI as a service (AIaaS). The paper provides some insights into the potential benefits, challenges, examples, and considerations when adopting each of these approaches. |
| doi_str_mv | 10.20535/SRIT.2308-8893.2025.1.08 |
| first_indexed | 2025-07-17T10:28:45Z |
| format | Article |
| fulltext |
Publisher IASA at the Igor Sikorsky Kyiv Polytechnic Institute, 2025
104 ISSN 1681–6048 System Research & Information Technologies, 2025, № 1
TIДC
МЕТОДИ, МОДЕЛІ ТА ТЕХНОЛОГІЇ ШТУЧНОГО
ІНТЕЛЕКТУ В СИСТЕМНОМУ АНАЛІЗІ
ТА УПРАВЛІННІ
UDC 004.89:004
DOI: 10.20535/SRIT.2308-8893.2025.1.08
AGENT-BASED APPROACH TO IMPLEMENTING
ARTIFICIAL INTELLIGENCE (AI) IN SERVICE-ORIENTED
ARCHITECTURE (SOA)
A.I. PETRENKO
Abstract. Artificial Intelligence (AI) is becoming a general-purpose technology and
is gaining a universal character for engineering, science, and society that today is
only inherent in mathematics and computer technology. The agent-based approach
to implementing artificial intelligence (AI) within the service-oriented architecture
of an application is a fascinating and highly synergistic concept. Combining these
paradigms leads to robust, scalable, and intelligent systems well suited for dynamic
and distributed environments. This paper presents the results of a comparative
analysis of three possible approaches to integrating AI into business processes,
namely, connecting AI agents to service-oriented architecture (SOA), connecting AI
agents to software (SaaS), and building AI as a service (AIaaS). The paper provides
some insights into the potential benefits, challenges, examples, and considerations
when adopting each of these approaches.
Keywords: AI (Artificial intelligence), agentic AI, AI-agent, SOA (Service oriented
architecture), SaaS (Software-as-a-Service), RAG (Retrieval-Augmented Genera-
tion), large language models (LLM), single-agent and multi-agent systems, AI agent
development platforms, AI agent integration with SaaS, AI agents and SOA.
AGENTIC AI
As artificial intelligence becomes an increasingly integral part of how we live and
work, it’s important to understand the differences between agent-based AI and
generative AI [1; 2; 5; 10; 11; 12; 18].
Generative AI is a type of AI that focuses on creating new content, such as
text, images, music, or even video. It works by learning from large amounts of
data to understand patterns, styles, or structures, and then generating original con-
tent based on what it has learned. For example, generative AI such as ChatGPT
can generate unique text answers to questions, while image generation models
such as DALL-E can create images from text descriptions. In essence, generative AI is
like a digital artist or writer, creating creative works based on what it has learned.
Agentic AI, on the other hand, is a step forward. Unlike generative AI, it can
take initiative, set goals, and learn from its own experience (Fig. 1). It is proac-
tive, able to adjust its actions over time, and can handle more complex tasks that
Agent-based approach to implementing artificial intelligence (AI) in service-oriented architecture (SOA)
Системні дослідження та інформаційні технології, 2025, № 1 105
require constant problem solving and decision making. This transition from reac-
tive to proactive AI opens up new possibilities for technology in many fields, al-
lowing machines to operate with near-human understanding and creating a seis-
mic technological shift. Machines now understand us better than ever before.
They can learn, predict, intuit, and reason. They can take on uncertain tasks, man-
age complex processes, and make subtle decisions that only a year or two ago
could only be made by humans. Imagine a robot that operates without a human
controller, determining what to do next based on its environment, or a self-driving
car that strives to get you to your destination safely, with every action, from steer-
ing to braking, serving that goal.
In short, at its core, agent-based AI is a type of AI that prioritizes autonomy.
Agents have true autonomy, making decisions and taking actions independently
with minimal human supervision. The level of autonomy is determined by the
number of iterations an AI agent can go through to reach a conclusion, as well as
the number of tools at its disposal.
Fig. 1. Difference between Generative AI and Agentic AI [5]
An AI agent is an interactive computer program with pre-defined goals that
can perform a variety of tasks on behalf of a user or another program. AI agents
have the potential to understand and learn from their environment, make deci-
sions, act, and even continuously improve with minimal human intervention. This
is achieved by integrating several key components that allow it to interact with
data, interpret its environment, choose appropriate responses, and communicate
meaningfully with users. In addition, AI agents can benefit from human feedback,
which enhances their adaptability and performance.
In particular, AI agents are used to automate the process of developing web
applications without coding. They can analyze user needs, generate code, test ap-
plications, and even deploy them. This makes application development faster and
more cost-effective. However, the use of AI agents requires the availability of
LLMs (large language models such as GPT-4, Claude, Gemini, etc.), which can
be expensive because LLMs require a lot of resources:
Significant computational resources for training and operation.
Huge amounts of data for training.
A.I. Petrenko
ISSN 1681–6048 System Research & Information Technologies, 2025, № 1 106
Highly qualified machine learning and artificial intelligence specialists to
develop and maintain LLMs.
Significant energy for training and running LLMs.
But there are alternatives and ways to reduce costs:
Cloud providers such as Google Cloud, Amazon Web Services, and Mi-
crosoft Azure offer access to LLMs on a pay-as-you-go basis.
For some tasks, smaller SLMs can be used that require fewer computing
resources (Microsoft phi-4, DistilBERT, TinyBERT, Albert).
There are open-source LLMs that can be used for free (Llama 2 and OPT
from Meta, GPT-Neo, GPT-J, GPT-3 from EleutherAI). However, they may be
inferior in quality to commercial models.
The choice depends on the specific needs of your project. If you need maxi-
mum performance and functionality, commercial LLMs may be a better option. If
efficiency, speed and resource savings are important, smaller SLMs or open
source LLMs may be a better choice. In general, open source LLMs and small
SLMs play an important role in the development and dissemination of artificial
intelligence technologies, making them more accessible, efficient and versatile.
It should be noted that the boundaries between generative and agent-based
AI are not always clear. Many modern AI systems include elements of both, cre-
ating hybrid models that can generate content and make autonomous decisions.
AI AGENTS WITH RETRIEVAL-AUGMENTED GENERATION (RAG)
To make an agent contextually relevant, it is connected to an external knowledge
base or data source that supports its responses with accurate, domain-specific in-
formation. A common approach to such integration is the Retrieval-Augmented
Generation (RAG) pattern, which combines external data retrieval with generative
capabilities. In addition to basic understanding, the agent is equipped with a tool-
kit - specialized skills and abilities that allow it to autonomously perform actions,
initiate workflows, or solve tasks according to set goals. An orchestrator coordi-
nates all these components and ties the agent’s functionality together. The orches-
trator processes user input, manages internal operations, and delivers consistent
results either directly to the user or to other agents in multi-agent interaction systems.
LLMs are trained on static data sets, which can lead to outdated information.
RAG allows agents to access up-to-date information from dynamic sources such
as web pages and news feeds (Fig. 2). RAG allows agents to tailor responses to
Fig. 2. Comparing standard LLM calls with RAG [9]
Agent-based approach to implementing artificial intelligence (AI) in service-oriented architecture (SOA)
Системні дослідження та інформаційні технології, 2025, № 1 107
the specific context and needs of the user by retrieving relevant information from
personalized sources such as search history or user profiles. RAG also enables
agents to provide explanations and sources of information, increasing transpar-
ency and trust in their responses. RAG works in AI agents as follows [7–9]:
Request: The user asks a question or makes a request to the AI agent.
Retrieval: The agent uses information from the request to search for rele-
vant documents or data in external sources.
Extraction: The agent extracts the most important information from the
sources found.
Generation: The agent uses an LLM to generate a response, using the ex-
tracted information as context.
Overall, RAG is an important tool for creating more intelligent, accurate and
useful AI agents that can effectively interact with the real world and meet user
needs. However, an alternative and more modern approach has recently
emerged — Table-Augmented Generation (TAG). While RAG has proven effec-
tive in integrating AI with external data retrieval systems, TAG offers a paradigm
shift by allowing large language models (LLMs) to interact directly with struc-
tured databases. Table-Augmented Generation (TAG) provides a more direct and
structured approach, allowing LLMs to query databases using SQL or other data-
base-specific query languages [7].
TYPES OF AI AGENTS AND TOOLS FOR CREATING THEM
The following four rules define the functionality of an AI agent: autonomy, per-
ception, decision making and adaptability [3; 6].
Autonomy: This rule means that the AI agent must function independently
to perform tasks without constant user intervention.
Perception: The AI agent can interpret data from the environment, ob-
tained through sensors, cameras or other sources.
Decision-making: This involves the AI agent’s ability to choose appropri-
ate actions to achieve its goals.
Adaptability: Adaptability is the ability of an AI agent to learn from new
information or experience and improve its responses over time.
These principles are the foundation for the design of AI agents and are wide-
ly accepted in the AI community to describe the core capabilities that enable intel-
ligent, agent-like behaviour.
Table 1 below provides examples of different types of AI agents. Each type
reflects the functionality, adaptability, and level of autonomy of the agent, as well
as the specific ways in which AI agents interact with the environment, make deci-
sions, and process information [13; 14; 15; 17].
Since the end of 2024, there has been a significant shift from single-agent AI
solutions to multi-agent systems [16; 18].
Single-agent systems: These are focused artificial intelligence models
geared towards specific tasks, such as smart chatbots. While effective in isolated
scenarios, they have limitations in managing complex, interconnected workflows.
Single-agent systems typically require human involvement to provide ongoing
feedback.
A.I. Petrenko
ISSN 1681–6048 System Research & Information Technologies, 2025, № 1 108
Multi-agent systems: These involve a network of AI agents that collabo-
rate to solve problems or achieve goals that require diverse knowledge. Imagine a
team collaborating, communicating internally, critiquing each other, and improv-
ing each other’s results to solve a given task, as opposed to a single agent receiv-
ing feedback only from the human interacting with it.
T a b l e 1 . Types of AI Agents
№ Agent Type Description Examples
1 Simple Reflex
Agent
These agents act solely based on the current
state of the environment, without consider-
ing the history or planning for the future.
They follow a set of predefined rules
(condition-action rules), e.g., “if sensor
detects X, then perform Y”
Thermostats, basic robots,
or systems like spam filters
that react to input
based on simple criteria
2 Model-Based
Reflex Agent
These agents maintain a model of the world
that helps them keep track of the state of the
environment. They use the model to make
better decisions
by considering both the current state and
the history of interactions
Robotic vacuum cleaners
(like Roombas) that map
their environment
to navigate efficiently
3 Goal-Based
Agent
These agents not only track the state of the
environment but also act to achieve specific
goals. They use planning and search
algorithms to determine the best course
of action to achieve their objectives
Autonomous vehicles
deciding the optimal
route to a destination
4 Utility-Based
Agent
These agents evaluate different states or
actions based on a utility function, which
measures how desirable a state is.
They aim to maximize the utility
(or “happiness”) by choosing actions that
lead to the best possible outcome
AI in recommendation
systems (e.g., Netflix or
Spotify), where the goal
is to maximize
user satisfaction
5 Learning
Agent
These agents improve their performance
over time by learning from data or
experiences. They can adapt
to new tasks or environments
without explicit programming
Chatbots (like ChatGPT),
recommendation systems,
autonomous vehicles, and
game-playing AIs
(like AlphaGo)
6 Collaborative
Agent Works with other agents or humans Coordinating the work of
multiple robots
7 Mobile Agent Moves between networks to perform tasks Network management
scenario
8 Multi-Agent
System
Involves multiple agents working together
or competing to solve complex problems
that cannot be handled by a single agent.
Agents in a MAS can communicate,
negotiate, and collaborate to achieve shared
or individual goals
Traffic management
systems, distributed
supply chain systems,
and swarm robotics
9
Belief-Desire-
Intention
(BDI) Agent
Balances between beliefs, desires,
and intentions
Autonomous bots for
customer service
10 Interface Agent Helps users by learning preferences Personalized email sorting
11 Reactive Agent Reacts quickly without internal models Real-time game characters
12 Hybrid
Agents
These agents combine multiple types of
agent architectures (e.g., reflex and goal-
based) to leverage the strengths of each
Autonomous drones that use
reflex actions for obstacle
avoidance and goal-based
planning for navigation
Agent-based approach to implementing artificial intelligence (AI) in service-oriented architecture (SOA)
Системні дослідження та інформаційні технології, 2025, № 1 109
AI Agent Development Platforms [16; 19; 20] are specialized tools that
streamline the process of building, training, and deploying AI agents. These plat-
forms aim to abstract away much of the complexity involved in AI development,
allowing developers and even non-developers to create AI agents more effi-
ciently. They simplify AI Agent Development by:
Rapid Prototyping: Developers can quickly build prototypes without deep
dives into complex algorithms or infrastructure setup.
Reduced Technical Debt: By using standardized platforms, there’s less
custom code to maintain, reducing long-term technical debt.
Access to Cutting-edge Technology: Platforms often incorporate the latest
AI advancements, making them available to users without the need for extensive
research or development.
Focus on Business Logic: Developers can focus more on defining the
business logic of the AI agent rather than the underlying technology.
Some such platforms are listed in Table 2 although there are many other
platforms (H2O.ai, DataRobot, DOMO, etc.).
T a b l e 2 . Examples of applications of Agent-Based AI in SOA
№ Platform name Description of features
1
Google Vertex
AI Agent
Builder
This platform integrates Google’s foundation models, search function-
alities, and conversational AI technologies into a unified development
environment. Utilising Vertex AI Agent Builder, developers are em-
powered to construct AI agents through a no-code interface
or by employing more sophisticated frameworks
2
Microsoft Azure
Autonomous
Systems Platform
A plethora of instruments and facilities are available for the
development of artificial intelligence agents, encompassing Azure
Bot Service, Azure Cognitive Services and Azure Machine Learning
3
Amazon
SageMaker
Agents
AWS offers a variety of services for developing AI agents, such as
Amazon Lex, Amazon Polly, and Amazon Rekognition
4 Hugging Face
Transformers
An open-source platform that provides free access to a large number
of pre-trained AI models and tools for developing AI agents
5 Microsoft
AutoGen
Microsoft’s multi-agent conversational platform is designed to facili-
tate the development of Large Language Model (LLM) workflows,
with the objective of enabling the utilisation of diverse applications
across multiple industry sectors. In addition, AutoGen provides
“AutoGen Studio,” a tool that facilitates the creation of multi-agent
systems without the requirement for extensive coding
6 LangChain
This is an open-source platform for AI agents that features a low-code
drag-and-drop interface. It is available for download at no cost on
GitHub, but a paid OpenAI API key is required for operation
7 AgentGPT
A web platform for deploying AI agents directly from your web
browser using GPT 3.5. Supports tasks such as AI image creation,
Google search, and code writing
8
Salesforce Ein-
stein Agent
Builder
A platform for deploying AI agents from Salesforce, designed
to automate complex tasks. Includes Agent Builder for creating AI
agents with natural language instructions
9 Godmode
Web solution for AI agents powered by OpenAI and Microsoft.
Supports the creation of multiple AI agents simultaneously
using GPT-3.5 and GPT-4 based on user-defined goals
10 OpenAI Agents
With GPT-4o, o1, and subsequent versions, developers can create
increasingly complex agents and deploy them in their applications
or on the ChatGPT platform
A.I. Petrenko
ISSN 1681–6048 System Research & Information Technologies, 2025, № 1 110
Continued Table 2
11 Fetch.ai
This provides cryptographic and blockchain capabilities to AI agents,
thus allowing various types of agents (i.e. reflex agents, goal agents
and utility agents) to access blockchain features such
as crypto wallets and on-chain interaction
12 LlamaIndex A popular framework for developing AI agents, which provides
tools for working with large language models, such as GPT-3
13 CrewAI
CrewAI is an open-source framework in Python designed to support
the development and management of multi-agent artificial intelligence
systems. The enhancement of these AI systems is achieved by the alloca-
tion of specific roles, the enablement of autonomous decision-making, and
the facilitation of communication between agents. This collaborative ap-
proach enables the effective resolution of complex problems, surpassing
the capabilities of individual agents operating in isolation
14 PhiData
PhiData provides a comprehensive framework for the creation of com-
plex agents that possess enhanced memory and knowledge manage-
ment capabilities, utilising GPT-4 technology. This platform is particu-
larly well-suited for applications that necessitate profound contextual
understanding and long-term learning capabilities. Potential use cases
include long-term projects, knowledge-intensive tasks and personalised
user interaction
15 Atomic Agents
Atomic Agent is a versatile, open-source framework created by Brain-
Blend AI designed for developing multi-agent systems and AI applica-
tions. It emphasizes modularity and atomicity, allowing developers to
construct complex AI solutions by combining simple, interchangeable
components. By breaking down AI systems into smaller, self-
contained, reusable components, Atomic Agents promises a future
where AI development is both modular and predictable
Key Benefits of AI Agent Development Platforms are:
Abstraction of Complexity:
No-Code/Low-Code Interfaces: Platforms like Microsoft Azure AI, and
Google Cloud AI offer drag-and-drop interfaces or visual programming tools, re-
ducing the need for extensive coding.
Pre-built Components: Many platforms provide pre-built AI models,
templates, or modules for common tasks like NLP, image recognition, etc., which
can be easily integrated.
Integration and Deployment:
Seamless Integration: These platforms often come with built-in tools for
integrating AI agents with other systems or services, using APIs or direct connectors.
Deployment Automation: They handle the deployment process, including
scaling, which can be particularly useful for cloud-based solutions.
Data Management:
Data Pipelines: Platforms like DataRobot or H2O.ai offer tools to manage data
flows, from ingestion to preprocessing, making it easier to feed data into AI models.
Model Training: Automated or semi-automated model training features,
which can optimize hyperparameters and select the best model architecture.
User-Friendly Interfaces:
Dashboarding: Visual dashboards for monitoring model performance, da-
ta quality, and agent interactions.
Collaboration: Features for team collaboration, version control, and shar-
ing of AI assets.
Agent-based approach to implementing artificial intelligence (AI) in service-oriented architecture (SOA)
Системні дослідження та інформаційні технології, 2025, № 1 111
Scalability and Performance:
Cloud Resources: Leveraging cloud infrastructure for scalability, which is
crucial for AI applications that might require significant computational resources.
Security and Compliance:
Built-in Security Measures: Many platforms include security protocols to
protect data and models.
Compliance: Some platforms help in adhering to industry standards and
regulations, which is vital for deploying AI in regulated sectors.
In conclusion, AI agent development platforms significantly lower the entry
barrier for building AI applications. They democratize AI technology by making
it accessible to a broader audience, including those without deep technical exper-
tise in AI. However, for highly specialized or performance-critical applications, a
more custom approach might still be required. Many platforms operate on a sub-
scription or usage-based model, which can be costly for large-scale or long-term
projects. There’s also a risk of becoming dependent on the platform’s ecosystem,
which might complicate migration or integration with other systems.
CONNECTING AI AGENTS TO SOA
The agent-based approach to implementing artificial intelligence (AI) within a
service-oriented architecture (SOA) is a fascinating and highly synergistic con-
cept. Remind, that SOA is an architectural style where software components are
designed as independent, reusable services that communicate with each other over
a network. The combination of agent-based AI and SOA brings together the best
of both worlds—autonomous decision-making from agents and the modular, dis-
tributed nature of SOA [20; 21].
Potential Benefits of Connecting AI Agents to SOA:
Enhanced Intelligence in Services: AI agents can bring intelligence and
autonomy to SOA by enabling services to dynamically adapt, learn, and make
intelligent decisions based on real-time data.
Improved Automation: AI agents can automate complex tasks and
workflows within an SOA, leading to increased efficiency and reduced human
intervention.
Personalized Experiences: AI agents can analyze user data and prefer-
ences to personalize the delivery of services within an SOA, creating more tai-
lored and engaging experiences.
Dynamic Optimization: AI agents can continuously monitor and opti-
mize the performance of services within an SOA, ensuring optimal resource utili-
zation and responsiveness.
Dynamic and Adaptive Systems: Agents can adapt to changes in the en-
vironment or user demands in real-time. When embedded in SOA, this adaptabil-
ity allows services to be reconfigured dynamically based on the context, improv-
ing system responsiveness and robustness.
Decentralization: SOA is inherently decentralized, and the agent-based
approach aligns well with this philosophy. Each agent can act independently
while still interacting with other services, reducing bottlenecks and single points
of failure.
A.I. Petrenko
ISSN 1681–6048 System Research & Information Technologies, 2025, № 1 112
Scalability and Modularity: SOA is designed for modularity, and add-
ing intelligent agents to individual services allows for a scalable way to introduce
AI capabilities. New agents can be introduced or updated without disrupting the
entire system.
Interoperability: Agents can act as intermediaries or orchestrators be-
tween services in SOA, enabling better integration of heterogeneous systems or
legacy services.
Enhanced Decision-Making: Agents bring reasoning and decision-
making capabilities to SOA, enabling services to not just respond to requests but
also predict, optimize, and proactively act to improve outcomes.
Support for Complex, Multi-Agent Systems: In cases where multiple
agents are deployed (e.g., for supply chain management or IoT systems), SOA
provides a framework for communication and collaboration among agents, ensur-
ing interoperability and coordination.
Connecting AI agents to a Service-Oriented Architecture (SOA) involves
several technologies and methodologies to ensure seamless integration, interop-
erability, and effective communication between AI components and other ser-
vices. Here’s a breakdown of the key technologies and approaches:
API Integration
RESTful APIs: AI agents can expose their functionalities through REST-
ful services, allowing them to be consumed by other services within the SOA.
This method is stateless, making it scalable and easy to integrate.
GraphQL: For more flexible data fetching, GraphQL can be used where
clients can request exactly what data they need from the AI agent, reducing over-
fetching and under-fetching.
Messaging Systems
Message Brokers (e.g., RabbitMQ, Apache Kafka): These facilitate asyn-
chronous communication. AI agents can publish results or receive tasks through
messages, which is particularly useful for handling high volumes of data or when
real-time processing isn’t necessary.
Event-Driven Architecture (EDA): AI agents can react to events triggered
by other services, allowing for dynamic, responsive systems where AI capabilities
are invoked based on specific business events.
Microservices Architecture
Containerization (Docker, Kubernetes): AI services can be containerized,
making them portable and scalable. This approach fits well with microservices
where each AI function might be its own microservice.
Service Mesh (e.g., Istio): Enhances how services communicate, manage,
and secure inter-service communication, which is crucial when integrating AI
agents that might need specific network policies or security measures.
Data Handling and Integration
Data APIs: For AI agents that require or produce data, data APIs can be
used to integrate with data services or databases within the SOA.
Data Streaming Technologies: Tools like Apache Kafka or AWS Kinesis
can be used for real-time data streaming to and from AI agents, ensuring they
have the latest data for processing.
Orchestration and Workflow Management
Workflow Engines (e.g., Camunda, Apache Airflow): These can orches-
trate complex workflows where AI agents are just one part of a larger process,
ensuring that AI tasks are executed in the right sequence and context.
Agent-based approach to implementing artificial intelligence (AI) in service-oriented architecture (SOA)
Системні дослідження та інформаційні технології, 2025, № 1 113
Serverless Computing: Using platforms like AWS Lambda or Azure
Functions, AI tasks can be executed in response to events without managing the
underlying infrastructure.
Security and Governance
OAuth, JWT (JSON Web Tokens): For secure communication and authen-
tication between services.
API Gateways: To manage access to AI services, ensuring that only au-
thorized services can interact with AI agents.
Monitoring and Management
Service Monitoring Tools (e.g., Prometheus, Grafana): To monitor the
health, performance, and usage of AI services within the SOA.
Centralized Logging: Tools like ELK stack (Elasticsearch, Logstash, Ki-
bana) for logging and analyzing interactions and performance of AI agents.
AI-Specific Technologies
Model Serving Platforms (e.g., TensorFlow Serving, Seldon Core): These
platforms allow for the deployment of machine learning models as services, mak-
ing it easier to integrate AI models into SOA.
Feature Stores: Centralized repositories for managing and serving features
used by machine learning models, ensuring consistency and reducing data duplication.
Integration Patterns
Proxy Pattern: AI agents can act as proxies or facades, simplifying the in-
terface for other services.
Adapter Pattern: Used to convert the interface of an AI agent into another
interface clients expect, improving reusability.
The technology for connecting AI agents to SOA involves a blend of modern
software architecture practices, cloud technologies, and AI-specific tools. The
goal is to create a flexible, scalable, and maintainable system where AI capabili-
ties are seamlessly integrated into business processes, enhancing overall system
intelligence without disrupting existing services. This integration requires careful
planning, especially around data flows, security, and performance, to ensure that
the AI components work harmoniously within the broader service ecosystem.
While the agent-based AI approach within SOA is powerful, there are some
challenges and factors you need to consider:
Complexity: Introducing agents into SOA can increase system complexity,
especially when managing interactions between autonomous agents and services.
Communication Overhead: Agents and services need to communicate
frequently, which can introduce latency or bottlenecks in distributed systems if
not carefully designed.
Security: Autonomous agents might make unauthorized decisions or in-
teract with malicious services. Ensuring secure communication and decision-
making is critical.
Standardization: SOA relies on standard protocols (e.g., SOAP, REST),
while agents may require additional protocols for negotiation, collaboration, or
reasoning. Aligning these standards can be challenging.
Scalability of Decision-Making: As the number of agents and services
grows, ensuring that agents can make decisions in a timely manner without over-
whelming the system is essential.
A.I. Petrenko
ISSN 1681–6048 System Research & Information Technologies, 2025, № 1 114
Interoperability of AI Models: Agents might use different AI models,
which could lead to compatibility issues. A common framework or ontology may
be needed for agents to collaborate effectively.
Monitoring and Debugging: Debugging agent behaviors in a distributed
SOA environment can be complex, especially when agents are making autono-
mous decisions based on incomplete or uncertain information.
It should be emphasized that this approach differs significantly from previ-
ous developments, where a software agent-based service-oriented integration ar-
chitecture for collaborative intelligent systems was proposed. A unique feature of
mentioned approach was that the order planning process was organized online
through negotiations between agent-based web services. Of course, the software
agents used were not present AI agents [22].
Agent-based AI within a Service-Oriented Architecture (SOA) is a powerful
combination. Examples of applications of Agent-Based AI in SOA are presented
in Table 3.
T a b l e 3 . Examples of applications of Agent-Based AI in SOA
№ Sectors Use cases Agents’ role
Smart Grids
AI agents represent energy producers (solar panels, wind
turbines), consumers (homes, businesses), and storage
units. They interact and negotiate in real-time to balance
energy supply and demand, optimize grid stability, and
reduce costs 1
Resource
Management
and
Optimization Supply
Chain
Logistics:
Agents model suppliers, manufacturers, distributors, and
customers. They autonomously manage inventory, pre-
dict disruptions, and optimize delivery routes to improve
efficiency and responsiveness
E-commerce
AI agents act as virtual shopping assistants, learning cus-
tomer preferences and providing personalized product
recommendations, deals, and support, enhancing the
shopping experience 2
Personalized
Customer
Service
Financial
Services
Agents offer personalized financial advice, analyze mar-
ket trends, and manage investment portfolios based on
individual client goals and risk tolerance
Healthcare
Agents simulate patients, doctors, hospitals, and other
healthcare providers to model disease spread, evaluate
treatment strategies, and optimize healthcare resource
allocation 3
Complex
System
Modeling
and
Simulation Traffic
Management
Agents represent vehicles, pedestrians, and traffic sig-
nals. They interact to optimize traffic flow, reduce con-
gestion, and improve road safety
Robotics
Agents control robots in manufacturing, warehouse au-
tomation, and exploration. They can adapt to changing
environments, collaborate with other robots, and learn
from experience 4 Autonomous
Systems
Self-Driving
Cars
Agents perceive the environment, make driving deci-
sions, and coordinate with other vehicles to ensure safe
and efficient navigation
Examples of Real-World Applications:
IBM Watson: Used in healthcare to provide personalized cancer treat-
ment recommendations.
Amazon Alexa: Employs AI agents for natural language understanding
and task automation.
Agent-based approach to implementing artificial intelligence (AI) in service-oriented architecture (SOA)
Системні дослідження та інформаційні технології, 2025, № 1 115
Tesla Autopilot: Utilizes agent-based AI for autonomous driving features.
Agent-based AI in SOA is transforming how we design and build intelligent
systems. By combining the strengths of both approaches, we can create more flex-
ible, scalable, and responsive solutions to complex real-world problems.
CONNECTING AI AGENTS TO SAAS
Applying an AI agent to a Software as a Service (SaaS) platform can indeed be
highly beneficial, offering numerous advantages that can enhance the value prop-
osition of the SaaS product. It is important to recall the definition of SaaS, which
is a software delivery service in which a provider hosts a software service and
makes it available to customers over the Internet. Customers can access the soft-
ware through a web browser, eliminating the need to purchase, install and main-
tain software on their own servers. The SaaS provider assumes responsibility for a
wide range of tasks, including maintaining servers and databases, providing up-
dates and implementing security measures. There are a number of platforms for
creating sophisticated web applications without writing code [23]:
Customer Relationship Management (CRM): Salesforce, HubSpot, Zoho Creator.
Enterprise Resource Planning (ERP): SAP, Oracle NetSuite.
Project Management: Trello, Asana, Webflow.
Document Collaboration: Google Workspace, Microsoft 365.
E-commerce Platforms: Shopify, Magento, Airtable.
Marketing Automation: Mailchimp, Marketo, Bubble.
Human Resources: BambooHR, Workday.
Key Characteristics of SaaS Platforms:
Cloud-Based Delivery: SaaS applications are hosted on servers in the
cloud, accessible via web browsers or lightweight client applications. This elimi-
nates the need for users to install software on their local machines.
Subscription-Based Pricing: Users typically pay for SaaS on a subscrip-
tion basis, which can be monthly, annually, or based on usage. This model con-
trasts with traditional software where you buy a license outright.
Multi-Tenancy: The software is designed to serve multiple customers
(tenants) from a single instance of the application. Each customer’s data is iso-
lated and secure, but the codebase and infrastructure are shared.
Scalability: SaaS platforms are built to scale easily, allowing them to ac-
commodate growth in users, data, or functionality without significant additional
setup or cost.
Automatic Updates: Updates, including new features, security patches,
and bug fixes, are automatically rolled out to all users, ensuring everyone has the
latest version without manual updates.
Data Management: The provider manages data storage, backup, and recov-
ery, which includes handling data security and compliance with relevant regulations.
Accessibility: Users can access the software from any device with an in-
ternet connection, often requiring only a web browser, which enhances mobility
and remote work capabilities.
Benefits of SaaS Platforms:
Cost Efficiency: Reduces the need for capital expenditure on hardware
and software licenses.
A.I. Petrenko
ISSN 1681–6048 System Research & Information Technologies, 2025, № 1 116
Ease of Use: Generally easier to deploy and use compared to on-premises
software.
Flexibility: Allows for flexible scaling of resources according to business
needs.
Innovation: SaaS providers often update their services with new features,
keeping the software current with market trends.
Challenges:
Dependency on Internet: Requires a reliable internet connection for access.
Data Security: Concerns about data privacy and security since data is
stored on external servers.
Customization: Some SaaS solutions might not offer the level of customi-
zation that on-premises software can provide.
Vendor Lock-in: Potential for dependency on the SaaS provider, making it
difficult to switch.
Key Benefits and Potential of AI-SaaS Integration:
Enhanced Automation: AI agents automate routine business processes
in SaaS applications, minimizing human intervention and can automate routine
tasks, customer inquiries, or even complex processes like data analysis, freeing up
human resources for more strategic tasks.
Improved Customer Experience: AI agents use SaaS platforms to de-
liver fast and personalized services to customers and to tailor the user interface,
content, and recommendations based on user behavior and preferences, making
the service more intuitive and engaging.
Data Analytics and Insights: AI agents analyze data within SaaS appli-
cations to deliver actionable insights and to automate routine tasks, customer in-
quiries, or even complex processes like data analysis, freeing up human resources
for more strategic tasks.
Collaboration Efficiency: AI optimizes collaboration processes within
SaaS applications and optimize the use of computational resources, ensuring the
SaaS platform scales efficiently with demand.
Cost and Time Savings: AI-SaaS integration reduces operational costs
by minimizing the need for manual labour and saves time.
Security and Compliance: AI agents enhance data security and compli-
ance within SaaS applications by identifying unusual patterns that might indicate
a security breach or unauthorized access.
Innovation and Competitive Edge: AI agents can be used to de-
velop new features or enhance existing ones, providing a competitive edge
by offering capabilities that are difficult for competitors to replicate quickly.
Market Differentiation: A SaaS with integrated AI can differentiate itself in
a crowded market by offering smart, proactive services.
Customer Engagement: Chatbots and Virtual Assistants can provide 24/7
customer support, handle FAQs, and guide users through the platform, improving
customer satisfaction and engagement.
AI agents integrated with SaaS (Software as a Service) applications offer a
wide range of opportunities for businesses, including process optimization, cus-
tomer satisfaction, cost savings, and strategic decision support. When integrating
AI into a SaaS platform, the AI functionalities are often provided as part of the
service, enhancing the platform’s capabilities (Table 4).
Agent-based approach to implementing artificial intelligence (AI) in service-oriented architecture (SOA)
Системні дослідження та інформаційні технології, 2025, № 1 117
T a b l e 4 . Examples of applications of Agent-Based AI in SaaS
№ Sectors Use cases Agents’ role
Learning
Platforms
AI agents act as virtual tutors, adapting to individual learning
styles, recommending relevant content, and providing person-
alized feedback to optimize learning outcomes
1
Personalized
User
Experience Content
Streaming
Service
Agents analyze user preferences and viewing history to
suggest personalized recommendations, discover new con-
tent, and create custom playlists
CRM
Systems
AI agents automate repetitive tasks like data entry, lead
qualification, and customer segmentation, freeing up hu-
man agents to focus on more strategic activities
2 Intelligent
Automation Project
Management
Software
Agents can monitor project progress, identify potential
risks, and automatically assign tasks to team members
based on their skills and availability
Help Desk
Software
AI agents provide instant answers to common questions,
troubleshoot issues, and escalate complex problems to human
agents, improving response times and customer satisfaction
3
Proactive
Customer
Support
Chatbots
Agents engage in natural language conversations with cus-
tomers, providing support, answering questions, and guid-
ing them through processes
Marketing
Automation
Platforms
Agents analyze customer data to identify patterns, predict
behavior, and personalize marketing campaigns for better
engagement and conversion rates
4
Data
Analysis
and Insights Business
Intelligence
Tools
Agents can sift through large datasets, identify trends, and
generate reports to provide valuable insights
for decision-making
Cybersecurity
Platforms
AI agents monitor network traffic, detect anomalies, and
respond to security threats in real-time, protecting sensitive
data and systems
5 Enhanced
Security Identity
and Access
Management
Agents can analyze user behavior, identify suspicious ac-
tivities, and prevent unauthorized access to critical re-
sources
6
Automated
Task
Manage-
ment
AgentForce
AgentForce is Salesforce’s innovative AI agent designed to
deliver speed, efficiency, and personalized solutions in
areas like customer service, sales, and marketing. Its core
strength lies in placing artificial intelligence at the heart of
business processes, helping companies become smarter and
more competitive
Examples in Action:
Salesforce Einstein: Uses AI agents to provide sales predictions, auto-
mate tasks, and personalize customer interactions.
HubSpot: Leverages AI for content optimization, lead scoring, and chat-
bot interactions.
Grammarly: Employs AI agents to provide grammar and writing suggestions.
By incorporating agent-based AI into SaaS applications, businesses can un-
lock new levels of efficiency, personalization, and intelligence, ultimately deliver-
ing greater value to their customers. In summary, a SaaS platform delivers soft-
ware applications over the internet, managed by third-party providers, offering
users a convenient, scalable, and often cost-effective way to access software. The
integration of AI into these platforms leverages cloud computing’s scalability and
data processing capabilities to provide advanced, data-driven features that can
significantly enhance the functionality and user experience of the software.
A.I. Petrenko
ISSN 1681–6048 System Research & Information Technologies, 2025, № 1 118
BUILDING AI AS A SERVICE (AIAAS)
With AIaaS, businesses of all sizes can access natural language processing (NLP),
machine learning (ML) algorithms, predictive analytics, and more to automate
tasks, analyze data, or improve business strategies and customer experience. They
can use and benefit from these AI tools, even without a large team of developers
or a huge budget, making it a lower-risk way to integrate AI into their business.
As a cloud computing service, AIaaS is flexible and can easily scale as its needs
grow without updating your hardware or infrastructure [24; 25].
Advantages:
Cost-effectiveness: Businesses can access AI capabilities without the need
for significant upfront investment in hardware and software.
Scalability and Flexibility: AIaaS solutions can be easily scaled up or
down to meet changing business needs.
Faster Deployment: Pre-built AI agents can be quickly integrated into ex-
isting systems and workflows.
Access to Expertise: AIaaS providers offer expertise and support to help
businesses effectively leverage AI technology.
Business Model Innovation: AIaaS can open new revenue streams by of-
fering AI capabilities as a subscription or pay-per-use model, potentially attract-
ing a broader customer base.
Scalability: Easier to scale AI services as they are managed centrally by
the service provider, who can optimize resources across multiple clients.
Focus on Core Business: Allows companies to focus on their core compe-
tencies while outsourcing AI development and maintenance.
Challenges:
Dependency: Clients become dependent on the service provider for AI
capabilities, which could pose risks if the provider faces issues or if there are
changes in service terms.
Customization: Generic AI services might not fully meet the specific
needs of every business, potentially requiring additional customization which
could negate some cost benefits.
Data Privacy and Security: Outsourcing AI means sensitive data might be
processed outside the organization, raising concerns about data privacy and com-
pliance with regulations like GDPR or CCPA.
Market Saturation: The AIaaS market could become saturated, leading to
price wars and reduced profitability.
The “AI Agent as a Service” model involves selecting of an AI agent type
and tools for AI agent Development [25; 27]. The type of AI agent which is build-
ing depends on the complexity of the task and the environment in which it will
operate. From simple reflex agents to learning agents and multi-agent systems,
each type has strengths and applications. Similarly, the tools and frameworks
which are choosing will depend on project requirements, such as scalability, ease
of use, and the specific domain (e.g., robotics, gaming, or conversational AI). By
combining the right type of agent with suitable tools, it is possible to create pow-
erful AI systems tailored to your needs. Some existing use cases where this model
has been implemented: are summarizing in Table 5.
Agent-based approach to implementing artificial intelligence (AI) in service-oriented architecture (SOA)
Системні дослідження та інформаційні технології, 2025, № 1 119
T a b l e 5 . Examples of use cases for which AIaaS was developed
№ Sectors AIaaS Agents’ role
Customer
Service
Agents
Companies like Ada and Intercom offer AI-powered
chatbots that can be integrated into websites
and apps to handle customer inquiries, provide support,
and even process transactions
Sales and
Marketing
Agents
Tools like Drift and Conversica provide AI agents that
can qualify leads, schedule appointments, and nurture
prospects through personalized email campaigns
1
Specialized
AI
Assistants
HR and
Recruiting
Agents
Services like Ideal and Pymetrics use AI agents to screen
resumes, conduct initial interviews, and match candidates
with the best-fit jobs
UiPath and
Automation
Anywhere
These platforms offer AI agents that can automate repeti-
tive tasks, such as data entry, invoice processing, and
report generation, across various business applications
2
AI-Powered
Automation
Platforms Zapier and
IFTTT
These services use AI agents to connect different apps
and automate workflows, such as sending notifications,
creating tasks, and updating spreadsheets
3
Healthcare
Data Analy-
sis
Google Cloud
Healthcare
API with AI
capabilities
This AI agent allows healthcare providers to analyze
patient data for better treatment outcomes, operational
efficiency, and compliance with healthcare regulations
without needing to develop AI models in-house
4
Content
Generation
and Editing
Grammarly
for Business
This AI agent not only checks for grammar and spelling
but also provides suggestions for clarity, tone,
and engagement, ensuring high-quality content
production at scale
5
Personalized
Learning
Knewton’s
Alta
for adaptive
learning
This AI agent personalizes the learning experience for
each student, adapting content and difficulty based on
individual performance, thereby improving learning
outcomes
6
Energy
Sector
Optimization
Siemens’
MindSphere
This AI agent helps in scheduling maintenance
efficiently, reducing energy costs, and extending
the lifespan of infrastructure
7
Real-time
Traffic
Management
INRIX
Traffic AI
This AI agent helps in reducing traffic jams, improving
emergency response times, and enhancing overall city
mobility
These examples demonstrate how “AI Agent as a Service” can be applied
across various industries to provide specialized AI functionalities without the
need for clients to invest heavily in AI infrastructure or expertise. This model al-
lows businesses to leverage cutting-edge AI technologies for specific tasks, en-
hancing their operations, customer service, and decision-making processes.
COMPARISON OF DIFFERENT WAYS OF AI AGENTS’ USAGE
There are three basic approaches: connecting AI agents to SOA, connecting AI agents
to SaaS and building a new business model “AI as a Service”. Which is better?
Key Differences:
SOA: Emphasizes modularity, reusability, and interoperability across dif-
ferent systems and applications within an organization. It’s ideal for complex en-
terprise-level solutions where integration and flexibility are crucial.
SaaS: Focuses on enhancing specific applications with AI capabilities to
improve user experience, automate tasks, and provide intelligent insights. It’s of-
ten delivered as part of a cloud-based software subscription.
A.I. Petrenko
ISSN 1681–6048 System Research & Information Technologies, 2025, № 1 120
AIaaS: Provides access to pre-built AI agents or tools for specific tasks, such
as natural language processing, image recognition, or data analysis. It allows busi-
nesses to leverage AI without the need for extensive development or infrastructure.
The choice depends on the organization’s strategic goals, existing techno-
logical landscape, risk tolerance, and market positioning [26] The detailed analy-
sis of important factors, including features of integration, scalability, customiza-
tion, maintenance, deployment and cost, are summarized in Table 6.
T a b l e 6 . Comparison of different ways of AI agents’ usage
Feature Agent-Based AI in
SOA
Agent-Based AI in
SOA AI Agents as a Service
Architecture Decentralized,
service-oriented
Centralized,
application-specific
Typically centralized,
API-driven
Integration
Method
AI agents are integrated
as services within an
existing or new SOA
framework
AI agents are inte-
grated into or along-
side SaaS applications,
often through APIs
or as plugins
AI capabilities are offered as
standalone services, accessi-
ble via APIs, which can
be integrated into any
platform or application
Deployment On-premise or cloud Cloud-based Cloud-based
Focus
Complex system inte-
gration, enterprise-level
solutions
Specific application
functionality, user
experience
Specialized AI tasks,
pre-built agents
Customization High Moderate
Limited, but increasing
with API options
Scalability
Highly scalable
through add-
ing/removing agents
Agent-Based AI in
SaaS
Highly scalable, managed
by the provider
Data
Handling
Full control over data
flow and processing
Less control over data,
what simplifies data
management
Users have control over
what data they send to the
AI service, but the data
processing happens on the
provider’s infrastructure
Maintenance
Can be complex due to
the need to maintain
both AI services and
the underlying SOA
infrastructure
Lower maintenance
burden on the user
side as SaaS providers
handle much of the
backend maintenance
Low for users, as the service
provider manages the AI
infrastructure and updates
Cost Higher upfront
investment
Subscription-based,
predictable costs
Pay-as-you-go,
variable costs
Examples
Smart grids, supply
chain optimization,
healthcare systems
Personalized learning
platforms, CRM sys-
tems, help desk soft
Chatbots, virtual assistants,
automation tools
Choosing the Right Approach:
The best approach depends on your specific needs and goals:
SOA: Best for large organizations with complex systems and integration
requirements.
SaaS: Suitable for businesses seeking AI-powered features within spe-
cific applications.
AIaaS: Ideal for those wanting to quickly integrate AI capabilities into
their existing systems or build custom AI solutions without heavy investment.
Comparison Summary:
Integration Complexity: SOA > SaaS > AIaaS (AIaaS being the simplest
for external integration).
Agent-based approach to implementing artificial intelligence (AI) in service-oriented architecture (SOA)
Системні дослідження та інформаційні технології, 2025, № 1 121
Control Over Data and Infrastructure: SOA > SaaS > AIaaS (SOA of-
fers the most control).
Scalability and Flexibility: AIaaS > SaaS > SOA (AIaaS and SaaS excel
in cloud-native environments).
Ultimately, these approaches are not mutually exclusive. They can be com-
bined and integrated to create comprehensive AI solutions that address a wide
range of business challenges.
CONCLUSIONS
Businesses are increasingly leveraging AI to enhance their operations and cus-
tomer experiences. There are three primary ways to integrate AI:
AI with SOA: This involves incorporating AI agents into existing Service-
Oriented Architectures (SOA). SOAs are built on interconnected services, and
adding AI can automate tasks, analyze data within these services, and improve
overall system efficiency. This approach is useful for businesses with established
SOAs looking to enhance functionality.
AI with SaaS: This integrates AI agents into Software as a Service applica-
tion. This can personalize user experiences, automate tasks within the application,
and provide valuable insights from user data. This approach is beneficial for
businesses utilizing SaaS solutions and wanting to improve their capabilities.
AI as a Service: This is a business model where AI capabilities are offered as
a standalone service. Companies can access sophisticated AI tools without
investing heavily in infrastructure or expertise. This is ideal for businesses want-
ing to experiment with AI or needing specific AI functions without building them
from scratch.
Each approach has its own merits and the best choice depends on a busi-
ness’s specific needs, existing infrastructure, and AI goals. However, all three
approaches have the potential to help businesses improve their operations and
gain a competitive advantage. Careful attention to design, communication, secu-
rity, and scalability is required to fully realize the benefits of these approaches.
AI agents powered by advanced generative AI (GenAI) technologies will be
the most disruptive force in technology in 2025 [29]. These autonomous systems,
capable of performing complex tasks with minimal human intervention, are
poised to revolutionize industries, rethink workflows, and increase productivity.
Using the results of recent research, including Deloitte’s 2025 Forecast Report
[30], we can predict the emergence, application, and future of AI agents in most
industries. Retrieval-augmented generation (RAG) extends the ability of AI
agents to work with unstructured data. Such agents can quickly find the informa-
tion they need and generate accurate answers, making them invaluable for knowl-
edge management and improving the efficiency of managing interdependent
workflows. AI agent architectures are becoming more modular, allowing for greater
customization and scalability. This evolution ensures that organizations can deploy
solutions tailored to their specific needs without significant AI agent overhauls.
The surge in popularity of AI agents in late 2024 mirrors how ChatGPT and
other LLMs transformed the AI market in 2022. Now, vendors and developers are
massively shifting from creating cutting-edge LLMs and AI chatbots to develop-
ing AI agents and exploring ways to implement them. Recently, a new approach
to language modelling (called Large Concept Models or LCMs) has been an-
nounced that operates at a higher semantic level, dealing with concepts that often
correspond to a sentence in text or an equivalent speech utterance [31].
A.I. Petrenko
ISSN 1681–6048 System Research & Information Technologies, 2025, № 1 122
It is predicted that starting in 2025, organizations that embrace this transfor-
mation will thrive, while those that cling to traditional SOA and SaaS paradigms
will struggle to remain relevant. Ethical considerations are at the forefront of AI
agent development. In 2025, the focus will be on explainability so that users can
understand and trust the decisions made by AI agents.
REFERENCES
1. SOAIS. Agentic AI: The Future of Autonomous, Action-Driven Automation. Available:
https://soais.com/agentic-ai-the-future-of-autonomous-action-driven-automation/
2. Lorena Nessi, What are AI Agents: How To Create a Based AI Agent. Available:
https://www.ccn.com/education/crypto/ai-agents-how-to-create-based-ai-agent/
3. Olesia, AI Agents: Types, Functions, Advantages & Challenges - Exploring the
World of Autonomous Intelligence. Available: https://lablab.ai/blog/the-best-ai-
agents-in-2023
4. Atomicwork. 12 AI agent frameworks for businesses to consider in 2025. Available:
https://www.atomicwork.com/itsm/best-ai-agent-frameworks
5. Haricharaun Jayakumar, Agent AI and Generative AI: understanding the differ-
ence. Available: https://www.solix.com/uk/blog/agentic-ai-and-generative-ai-
understanding-the-difference
6. Ilias Ism, 10 Real-World AI Agent Examples in 2024. Available: https://www.chat-
base.co/blog/ai-agent-examples
7. G. Sreedevi, From RAG to TAG: Leveraging the Power of Table-Augmented Gen-
eration (TAG): A Leap Beyond Retrieval-Augmented Generation (RAG). Available:
https://www.linkedin.com/pulse/from-rag-tag-leveraging-power-table-augmented-
leap-beyond-gogusetty-nh8hf/
8. Hiren Dhaduk, 6 Types of AI Agents: Exploring the Future of Intelligent Machines.
Available: https://www.simform.com/blog/types-of-ai-agents/
9. Avishek Biswas, The Ultimate Guide to RAGs — Each Component Dissected. Avail-
able: https://towardsdatascience.com/the-ultimate-guide-to-rags-each-component-
dissected-3cd51c4c0212
10. Emrecan Dogan, What are AI agents, and how do they work? Available:
https://www.glean.com/blog/ai-agents-how-they-work
11. Kanerika Inc. AI Agent Examples: The Future of Business and Technology.
Available: https://www.linkedin.com/pulse/ai-agent-examples-future-business-
technology-kanerika-tdbcc/
12. Agents in Artificial Intelligence. Available: https://www.javatpoint.com/agents-in-ai
13. Fuzen. AI Agents Examples and Types. Available: https://fuzen.io/ai-agents-
examples-and-types/
14. Samanyou Garg, 7 Types of AI Agents to Streamline Your Workflow in 2025. Avail-
able: https://writesonic.com/blog/types-of-ai-agents
15. Intelligent Agent Examples: Transforming Industries with AI. Available:
https://salescloser.ai/intelligent-agent-examples-transforming-industries-with-ai/
16. Sarfraz Nawaz, Top 24 Use Cases of AI Agents in Business. Available:
https://www.ampcome.com/post/24-use-cases-of-ai-agents-in-business
17. Safa Bouguezzi, Artificial Intelligence Agents Explained. Available:
https://www.baeldung. com/cs/artificial-intelligence-agents
18. The 4 biggest AI stories of 2024. Available: https://www.oksim.ua/2024/12/26/4-
najbilshi-istori%D1%97-pro-shtuchnij-intelekt-za-2024-rik/
19. OpenAI will launch an AI agent. Available: https://babel.ua/news/112657-openai-
zapustit-shi-agenta-vin-vikoristovuvatime-komp-yuter-vid-imeni-lyudini
20. Madhu Kumar, Paradigms and Transformations in Software Development.
Available: https://dev.to/aws-heroes/the-future-of-ai-agent-driven-paradigms-and-
transformations-in-software-development-55l
21. Jonghong Jeon, SOA, Agentic AI, AI as a Service, and Future Autonomous AI. Avail-
able: https://medium.com/@favorable_eminence_oyster_546/soa-agentic-ai-ai-as-a-
service-and-future-autonomous-ai-132fdd1765fc
Agent-based approach to implementing artificial intelligence (AI) in service-oriented architecture (SOA)
Системні дослідження та інформаційні технології, 2025, № 1 123
22. M. Alessandrini, W.M. Lippe, and W. Nuesser, “Intelligent Service System: An
Agent-Based Approach for integrating Artifical Intelligence Components in SOA
Landscapes,” 2008 IEEE/WIC/ACM International Conference on Web Intelligence
and Intelligent Agent Technology, Sydney, NSW, Australia, 2008, pp. 496–499. doi:
10.1109/WIIAT.2008.139
23. KOFANA. Digital. How Do AI Agents Integrate with SaaS Applications? Available:
https://www.kofana.com/en/trends/reports-and-blog/how-do-ai-agents-integrate-
with-saas-applications
24. DOMO. Guide to AI as a Service (AIaaS): Benefits, Types, Companies. Available:
https://www.domo.com/glossary/ai-as-a-service
25. How To Build LangChain Custom Agents (Tools). YouTube. 2024. Retrieved Au-
gust 4, 2023.
26. Oleksandra, A Brief Guide on AI Agents: Benefits and Use Cases. Available:
https://www.codica.com/blog/brief-guide-on-ai-agents/
27. G. Sreedevi, Building Multi-Agent LLM Systems with PydanticAI Framework: A
Step-by-Step Guide To Create AI Agents With Ease. Available: https://www.
linkedin.com/pulse/building-multi-agent-llm-systems-pydanticai-framework-
gogusetty-w4b5f/
28. Ori Ziv, How AI Agents Will Disrupt SaaS in 2025. Available: https://medium.
com/@oriziv4/how-ai-agents-will-disrupt-saas-in-2025-7567d793ca68
29. Deloitte Global’s 2025 Predictions Report: Generative AI: Paving the Way for a
transformative future in Technology, Media, and Telecommunication. Available:
https://www. deloitte.com/global/en/about/press-room/deloitte-globals-2025-
predictions-report.html
30. Edwin Lisowski, A List of AI Agents Set to Dominate in 2025. Available: https://medium.
com/@elisowski/a-list-of-ai-agents-set-to-dominate-in-2025-028f975c5b99
31. Mehul Gupta, Meta Large Concept Models (LCM): End of LLMs? Available:
https://medium.com/data-science-in-your-pocket/meta-large-concept-models-lcm-
end-of-llms-68cb0c5cd5cf
Received 16.01.2025
INFORMATION ON THE ARTICLE
Anatolii I. Petrenko, ORCID: 0000-0001-6712-7792, Educational and Research Institute
for Applied System Analysis of the National Technical University of Ukraine “Igor
Sikorsky Kyiv Polytechnic Institute”, Ukraine, e-mail: tolja.petrenko@gmail.com
АГЕНТНИЙ ПІДХІД ДО ВПРОВАДЖЕННЯ ШТУЧНОГО ІНТЕЛЕКТУ (AI)
В МЕЖАХ СЕРВІС-ОРІЄНТОВАНОЇ АРХІТЕКТУРИ (SOA) / А.І. Петренко
Анотація. Штучний інтелект (ШІ) стає технологією загального призначення і
набуває універсального характеру для техніки, науки і суспільства, який сьо-
годні притаманний лише математиці та комп’ютерним технологіям. Агентний
підхід до реалізації ШІ в межах сервіс-орієнтованої архітектури додатків є за-
хопливою і дуже синергетичною концепцією. Поєднання цих парадигм при-
зводить до створення надійних, масштабованих та інтелектуальних систем, які
добре підходять для динамічних і розподілених середовищ. Подано результати
порівняльного аналізу трьох можливих підходів до інтеграції ШІ в бізнес-
процеси, а саме: підключення агентів ШІ до сервіс-орієнтованої архітектури
(SOA), підключення агентів ШІ до програмного забезпечення (SaaS) та побу-
дова ШІ як послуги (AIaaS). Розглянуто потенційні переваги, виклики, при-
клади та міркування із застосуванням кожного з цих підходів.
Ключові слова: AI (Artificial Intelligence), агентний AI, AI-agent, SOA (Service
oriented architecture), SaaS (Software-as-a-Service), RAG (Retrieval-Augmented
Generation), великі мовні моделі (LLM), одноагентні і багатоагентні системи,
платформи розроблення АІ агентів, інтеграція агентів AI з SaaS, агенти АІ і SOA.
|
| id | journaliasakpiua-article-330091 |
| institution | System research and information technologies |
| keywords_txt_mv | keywords |
| language | English |
| last_indexed | 2025-09-17T09:26:02Z |
| publishDate | 2025 |
| publisher | The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" |
| record_format | ojs |
| resource_txt_mv | journaliasakpiua/4d/b754e4af611f304c88b6767d1402464d.pdf |
| spelling | journaliasakpiua-article-3300912025-05-20T17:56:07Z Agent-based approach to implementing artificial intelligence (AI) in service-oriented architecture (SOA) Агентний підхід до впровадження штучного інтелекту (AI) в межах сервіс-орієнтованої архітектури (SOA) Petrenko, Anatolii AI (Artificial Intelligence) агентний AI AI-agent SOA (Service oriented architecture) SaaS (Software-as-a-Service) RAG (Retrieval-Augmented Generation) великі мовні моделі (LLM) одноагентні і багатоагентні системи платформи розроблення АІ агентів інтеграція агентів AI з SaaS агенти АІ і SOA AI (Artificial intelligence) agentic AI AI-agent SOA (Service oriented architecture) SaaS (Software-as-a-Service) RAG (Retrieval-Augmented Generation) large language models (LLM) single-agent and multi-agent systems AI agent development platforms AI agent integration with SaaS AI agents and SOA Artificial Intelligence (AI) is becoming a general-purpose technology and is gaining a universal character for engineering, science, and society that today is only inherent in mathematics and computer technology. The agent-based approach to implementing artificial intelligence (AI) within the service-oriented architecture of an application is a fascinating and highly synergistic concept. Combining these paradigms leads to robust, scalable, and intelligent systems well suited for dynamic and distributed environments. This paper presents the results of a comparative analysis of three possible approaches to integrating AI into business processes, namely, connecting AI agents to service-oriented architecture (SOA), connecting AI agents to software (SaaS), and building AI as a service (AIaaS). The paper provides some insights into the potential benefits, challenges, examples, and considerations when adopting each of these approaches. Штучний інтелект (ШІ) стає технологією загального призначення і набуває універсального характеру для техніки, науки і суспільства, який сьогодні притаманний лише математиці та комп’ютерним технологіям. Агентний підхід до реалізації ШІ в межах сервіс-орієнтованої архітектури додатків є захопливою і дуже синергетичною концепцією. Поєднання цих парадигм призводить до створення надійних, масштабованих та інтелектуальних систем, які добре підходять для динамічних і розподілених середовищ. Подано результати порівняльного аналізу трьох можливих підходів до інтеграції ШІ в бізнес-процеси, а саме: підключення агентів ШІ до сервіс-орієнтованої архітектури (SOA), підключення агентів ШІ до програмного забезпечення (SaaS) та побудова ШІ як послуги (AIaaS). Розглянуто потенційні переваги, виклики, приклади та міркування із застосуванням кожного з цих підходів. The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" 2025-03-28 Article Article Peer-reviewed Article application/pdf https://journal.iasa.kpi.ua/article/view/330091 10.20535/SRIT.2308-8893.2025.1.08 System research and information technologies; No. 1 (2025); 104-123 Системные исследования и информационные технологии; № 1 (2025); 104-123 Системні дослідження та інформаційні технології; № 1 (2025); 104-123 2308-8893 1681-6048 en https://journal.iasa.kpi.ua/article/view/330091/319565 |
| spellingShingle | AI (Artificial Intelligence) агентний AI AI-agent SOA (Service oriented architecture) SaaS (Software-as-a-Service) RAG (Retrieval-Augmented Generation) великі мовні моделі (LLM) одноагентні і багатоагентні системи платформи розроблення АІ агентів інтеграція агентів AI з SaaS агенти АІ і SOA Petrenko, Anatolii Агентний підхід до впровадження штучного інтелекту (AI) в межах сервіс-орієнтованої архітектури (SOA) |
| title | Агентний підхід до впровадження штучного інтелекту (AI) в межах сервіс-орієнтованої архітектури (SOA) |
| title_alt | Agent-based approach to implementing artificial intelligence (AI) in service-oriented architecture (SOA) |
| title_full | Агентний підхід до впровадження штучного інтелекту (AI) в межах сервіс-орієнтованої архітектури (SOA) |
| title_fullStr | Агентний підхід до впровадження штучного інтелекту (AI) в межах сервіс-орієнтованої архітектури (SOA) |
| title_full_unstemmed | Агентний підхід до впровадження штучного інтелекту (AI) в межах сервіс-орієнтованої архітектури (SOA) |
| title_short | Агентний підхід до впровадження штучного інтелекту (AI) в межах сервіс-орієнтованої архітектури (SOA) |
| title_sort | агентний підхід до впровадження штучного інтелекту (ai) в межах сервіс-орієнтованої архітектури (soa) |
| topic | AI (Artificial Intelligence) агентний AI AI-agent SOA (Service oriented architecture) SaaS (Software-as-a-Service) RAG (Retrieval-Augmented Generation) великі мовні моделі (LLM) одноагентні і багатоагентні системи платформи розроблення АІ агентів інтеграція агентів AI з SaaS агенти АІ і SOA |
| topic_facet | AI (Artificial Intelligence) агентний AI AI-agent SOA (Service oriented architecture) SaaS (Software-as-a-Service) RAG (Retrieval-Augmented Generation) великі мовні моделі (LLM) одноагентні і багатоагентні системи платформи розроблення АІ агентів інтеграція агентів AI з SaaS агенти АІ і SOA AI (Artificial intelligence) agentic AI AI-agent SOA (Service oriented architecture) SaaS (Software-as-a-Service) RAG (Retrieval-Augmented Generation) large language models (LLM) single-agent and multi-agent systems AI agent development platforms AI agent integration with SaaS AI agents and SOA |
| url | https://journal.iasa.kpi.ua/article/view/330091 |
| work_keys_str_mv | AT petrenkoanatolii agentbasedapproachtoimplementingartificialintelligenceaiinserviceorientedarchitecturesoa AT petrenkoanatolii agentnijpídhíddovprovadžennâštučnogoíntelektuaivmežahservísoríêntovanoíarhítekturisoa |