11/07/2025, 08:20 Answering the Top Trending AI Agent Questions, Part 4: The Future of AI Agents
Federal Policy Guidance AI Policy Tariffs & Trade
Artificial Intelligence / Shape AI Strategy / Answering the Top Trending AI...
Licensed for Distribution
This research note is restricted to the personal use of Iim Lucknow Up
(
[email protected]).
Answering the Top Trending AI Agent
Questions, Part 4: The Future of AI
Agents
3 June 2025 - ID G00835194 - 10 min read
By Leinar Ramos, Ben Yan, and 9 more
TOPICS
Shape AI Strategy Future Direction for GenAI
AI agents have emerged as one of today’s most widely discussed trends, taking the tech
and business world by storm. AI leaders can dive into this research, which answers the
key questions for navigating and shaping the future of AI agents, ensuring they’re ready
for what’s next.
Overview
Key Findings
An emerging set of trends to improve AI agent capabilities — including reasoning
models, multimodality, large action models, neurosymbolic AI and world models —
will shape the future of AI agents, increasing their autonomy and capability.
https://www.gartner.com/document-reader/document/6554602?ref=solrResearch&refval=479953854& 1/13
11/07/2025, 08:20 Answering the Top Trending AI Agent Questions, Part 4: The Future of AI Agents
Multiagent systems are expected to progress from single-platform implementations
to cross-platform systems. This evolution may lead to an “Internet of Agents,” where
distributed agents autonomously discover, interact and collaborate dynamically to
tackle complex challenges.
The future of AI agents is ultimately uncertain. However, AI leaders can take certain
actions now that will position them well for this potential future.
Recommendations
Prepare for the future of AI agents by laying the foundations of application
composability, governance and data that will be required.
Start your journey by experimenting with single AI agent systems, as opposed to
more complex multiagent systems. When working on multiagent systems, prioritize
approaches where AI agents are coordinated in a fixed or semistructured workflow,
and not “swarm” approaches where AI agents collaborate without central planning.
Invest in improving your integration and API, workflow mapping and redesign,
retrieval augmented generation (RAG), security and governance, monitoring, and
experimentation capabilities. These are investments that are sensible in their own
right for even nonagentic use cases, but that also position you well to capture future
opportunities in the AI agent space.
Introduction
There has been a rapid surge of interest in AI agents. Gartner saw a 750% increase in AI-
agent-related inquiries between the second and fourth quarter of 2024. This is driven by
technology progress and increased market hype.
AI leaders will face increased pressure to separate the hype from reality, understand AI
agents and their limitations and risks, and learn how to begin taking advantage of these
emerging technologies. A lack of clear understanding and focus on realistic business
opportunities and outcomes will lead organizations to make the wrong investments and
wind up with unnecessary failed projects.
https://www.gartner.com/document-reader/document/6554602?ref=solrResearch&refval=479953854& 2/13
11/07/2025, 08:20 Answering the Top Trending AI Agent Questions, Part 4: The Future of AI Agents
This research is the fourth of a four-part series on the top trending AI agent questions
(see Table 1).
Table 1: Top Trending AI Agent Questions Enlarge Table
Research Note Questions Covered Link to Research
Part 1: What are AI agents? Answering the Top Trending
Foundational AI Agent Questions, Part 1:
What is the difference
Knowledge between AI agents and Foundational Knowledge
agentic AI?
What is the difference
between AI agents and
AI assistants?
What are the core
building blocks of AI
agents?
Part 2: Current Do AI agents require an Answering the Top Trending
State of AI LLM? AI Agent Questions, Part 2:
Agents Current State of AI Agents
What are the current
challenges and
limitations of AI agents?
Source: Gartner
Top Questions on the Future of AI Agents
What are the trends shaping the future of AI agents?
How will AI agents be combined into multiagent systems?
How do I prepare for the future of AI agents?
Analysis
https://www.gartner.com/document-reader/document/6554602?ref=solrResearch&refval=479953854& 3/13
11/07/2025, 08:20 Answering the Top Trending AI Agent Questions, Part 4: The Future of AI Agents
What Are the Trends Shaping the Future of AI Agents?
Back to top questions
AI agents represent a rapidly evolving technology. There are a few trends that are set to
shape its future:
Planning and reasoning: Lack of planning capability limits AI agent adoption, but
new reasoning models like OpenAI’s o3 and DeepSeek’s R1 are starting to be used in
AI agents to improve their planning and reasoning capabilities. These new models
are currently expensive and have high latency, but as this changes, AI agents that
use them could become more performant and reliable.
Large action models: (LAMs) are foundation models trained and optimized to
generate a sequence of actions to meet a goal. They could represent a better model
for AI agents than LLMs, which are not directly trained for actions.
Multimodality: Multimodal AI could allow AI agents to process and understand
multiple forms of data simultaneously, including text, audio, images and video.
Together with advanced sensors, this capability will enhance the perception
capability of AI agents, enabling systems to interact with the physical world
effectively (see Innovation Insight: Multimodal AI Explained).
Neurosymbolic AI: Neurosymbolic AI fuses machine learning (ML) methods and
symbolic systems (for example, knowledge graphs) to create more robust and
trustworthy AI models. This approach provides a reasoning infrastructure for AI
agents to solve a wider range of tasks more effectively and reliably.
World models: World models are learned abstract representations of an
environment. They enable AI systems to make predictions via simulating potential
future states and helping to understand the consequences of the actions taken,
which is a crucial missing capability for AI agents.
Multiagent systems: A near-term future trend is to go beyond single agents and
create collections of AI agents that interact with each other. Despite some early
experiments, multiagent systems are not widely prevalent because of their increased
complexity. The future of multiagent systems extends beyond the creation of AI
https://www.gartner.com/document-reader/document/6554602?ref=solrResearch&refval=479953854& 4/13
11/07/2025, 08:20 Answering the Top Trending AI Agent Questions, Part 4: The Future of AI Agents
agents in a single environment and into cross-platform AI agents and even into a
potential future Internet of AI Agents.
“Loopy” AI agents: Most current AI agents act in a “one-shot” way, taking actions
based on inputs, but not observing the impact of their actions in the environment.
Future AI agents are more likely to be in an interactive feedback loop, as seen in
Figure 1. Early examples include OpenAI’s computer-using agent, which receives a
computer screenshot as input, takes actions (like scrolling and clicking) in a virtual
machine environment, and then receives a new screenshot after those actions are
executed. The loop repeats until the agent navigates the environment to achieve its
1
goal.
Figure 1: Feedback Loop Between AI Agents and Their Environment
This is not an exhaustive list of trends, but some that are positioned to drive the future
of the technology.
How Will AI Agents Be Combined Into Multiagent
Systems?
Back to top questions
The market is rapidly looking beyond single AI agents and into multiagent systems —
collections of AI agents that interact with each other to achieve individual or shared
https://www.gartner.com/document-reader/document/6554602?ref=solrResearch&refval=479953854& 5/13
11/07/2025, 08:20 Answering the Top Trending AI Agent Questions, Part 4: The Future of AI Agents
goals. This is an important part of the future of AI agents, but it’s mostly an
experimental technology at this point.
The multiagent trend is driven in great part by the limitations of individual AI agents,
which are typically not reliable enough to perform well in a broad range of tasks. Given
this limitation, a potential solution is to break down a process into different tasks and
have AI agents that are more narrow and specialized for each task. This is challenging to
do in practice, and very few organizations have successfully deployed multiagent
systems in production.
Organizations need to get a single AI agent
working before attempting to build a
complex multiagent system.
The future of multiagent systems is likely to evolve in three phases, as seen in Figure 2:
Phase 1 — Single-platform MAS (Current Phase): In the vast majority of current
implementations, the multiple agents in a MAS are created on a single platform. This
approach allows for streamlined integration, where agents can easily share
information and collaborate. Additionally, operating on a single platform simplifies
maintenance and updates, as changes can be implemented across the system
without disrupting individual agent functions (see Case Study: Multiple AI Agents’
Recommendations Improve Efficiency).
Phase 2 — Cross-platform MAS: This utilizes agent communication protocols to
enable interoperability and collaboration among agents distributed across different
environments. These protocols facilitate standardized communication, allowing
agents to exchange information and coordinate actions (see Innovation Insight:
Model Context Protocol). Despite the increased interest and adoption of emerging
standards, cross-platform MASs are not yet a reality for the vast majority of
organizations.
https://www.gartner.com/document-reader/document/6554602?ref=solrResearch&refval=479953854& 6/13
11/07/2025, 08:20 Answering the Top Trending AI Agent Questions, Part 4: The Future of AI Agents
Phase 3 — The Internet of Agents (agentic web): This envisions a global network of
interconnected agents functioning as a MAS. In this expansive ecosystem, AI agents
can automatically discover and interact with one another, forming dynamic
collaborations to address complex challenges. This interconnected network would
leverage advanced discovery mechanisms, enabling agents to identify compatible
partners and establish communication channels autonomously. These AI agents will
become increasingly sophisticated, capable of negotiating roles, sharing resources
and coordinating efforts. This remains a hypothetical future, but there is increased
2
vendor interest and investment.
Figure 2: The Evolution of Multiagent Systems
Regardless of the evolution of MASs above, a fundamental question will remain on how
to coordinate different AI agents. As seen in Figure 3, there is a continuum between
fixed and dynamic coordination, involving a key trade-off — the more dynamic the
agent coordination, the higher the flexibility, but the lower the reliability of the
combined system.
Figure 3: Style of Multiagent Coordination
https://www.gartner.com/document-reader/document/6554602?ref=solrResearch&refval=479953854& 7/13
11/07/2025, 08:20 Answering the Top Trending AI Agent Questions, Part 4: The Future of AI Agents
There are three main ways to coordinate multiple AI agents:
Orchestrated agents: The agents are always invoked in the same order to perform
the same tasks in a fixed, predefined workflow.
Semistructured orchestration: The agent connections are specified, but the
workflow can change depending on the goal. This could include a planning agent
that divides the work between different agents.
Collaborating agents: Agents collaborate dynamically based on their knowledge of
each other’s capabilities. There is no predefined workflow or centralized planning. A
dynamic plan is formed from these interactions.
Most successful multiagent systems will have orchestrated agents or semistructured
coordination. Collaborating agent approaches, like swarms of AI agents that don’t have
any centralized planning, will be too unreliable in practice for the great majority of
enterprise use cases.
How Do I Prepare for the Future of AI Agents?
Back to top questions
The future of AI agents is ultimately uncertain and difficult to predict. However, AI
leaders can take certain actions now that will position them well for this potential future.
https://www.gartner.com/document-reader/document/6554602?ref=solrResearch&refval=479953854& 8/13
11/07/2025, 08:20 Answering the Top Trending AI Agent Questions, Part 4: The Future of AI Agents
Success with AI agents will come down to
positioning. The best investments will be in
activities that are sensible to do regardless,
but that position you to capture the upside
of opportunities as they emerge. You can’t
predict, but you can prepare.
Below is a list of a few recommended actions:
Integration and APIs: Assess your key applications’ interfaces and integration
capabilities, including how well the functionality and data are exposed via APIs,
events and standards. This will help establish an agent interface management
capability to support secure, policy-controlled, observable agent integration.
Experiment with emerging standards like MCP, which are important for AI agents, but
also helpful for data and application integration even in nonagentic LLM-based
applications.
Workflow mapping and redesign: AI agents will be part of broader business
workflows, which will need to be carefully redesigned in order to drive real business
value. A first step AI leaders can take is to map important business workflows and
customer journeys to identify opportunities for different types of AI use cases,
including agentic ones.
Retrieval augmented generation (RAG): RAG is an important capability to improve,
given that AI agents will require context and memory. This foundational capability is
valuable on its own, but it also positions organizations well for future AI agent
developments (see How to Improve and Optimize Retrieval-Augmented Generation
Systems).
Security and governance: AI agents will bring about new risks given their increased
autonomy. Without security and governance controls, organizations will not be able
to experiment and scale AI agents (see Quick Answer: Mitigate New Risks and
Security Threats From AI Agents).
https://www.gartner.com/document-reader/document/6554602?ref=solrResearch&refval=479953854& 9/13
11/07/2025, 08:20 Answering the Top Trending AI Agent Questions, Part 4: The Future of AI Agents
Monitoring tools: Invest in observability capabilities to gain visibility into agent
behavior. These capabilities are already useful for regular LLM-based systems, but
will be particularly useful for AI agents to have visibility on agent activities. This
includes monitoring resource consumption, anomaly detection, deviations from
expected behavior and tool usage.
Experimentation: Start experimenting with agentic use cases to build familiarity and
proficiency with the emerging set of design patterns in the space. The skills
developed in these experiments are useful regardless of whether the most optimistic
future scenarios like an Internet of Agents comes to pass (see Emerging Patterns for
Building LLM-Based AI Agents).
Evidence
Related Topics
Coordinate AI Delivery and Execution Design and Engineer AI Systems
Develop Resources and Assets for AI
MORE ON THIS TOPIC
Key AI Predicts 2025: AI and the
Automation Future of Work
Trends Shaping 25 March 2025
the Future of
AI will have a transformational
SecOps impact on the future of work —
altering indelibly how we perform
work in the future. CIOs should act
on these bold predictions...
https://www.gartner.com/document-reader/document/6554602?ref=solrResearch&refval=479953854& 10/13
11/07/2025, 08:20 Answering the Top Trending AI Agent Questions, Part 4: The Future of AI Agents
Enterprises Are Your AI Needs a
Implementing AI, Boss; Appoint an
But Not AI Leader
Everywhere (Yet) 26 June 2025
18 June 2025
View More
Recommended by Authors
Innovation Insight: Innovation Insight AI Agents: Are You
AI Agents for the AI Agent Ready to Set Your
RESEARCH 3 April 2024 Platform AI Free?
· 15 min read Landscape RESEARCH 20 February
RESEARCH 26 March 2025
2025 · 2 min read
· 14 min read AI agents enable
levels of business
adaptability,
flexibility and agility
that traditional AI
systems can’t
achieve.
Autonomous
behaviors, despite
significant...
Your Peers Also Viewed
Answering the Top Answering the Top Answering the Top
Trending AI Agent Trending AI Agent Trending AI Agent
Questions, Part 2: Questions, Part 1: Questions, Part 3:
Current State Getting Started
https://www.gartner.com/document-reader/document/6554602?ref=solrResearch&refval=479953854& 11/13
11/07/2025, 08:20 Answering the Top Trending AI Agent Questions, Part 4: The Future of AI Agents
RESEARCH 3 June 2025 RESEARCH 3 June 2025
Foundational
· 12 min read · 11 min read
Knowledge
RESEARCH 3 June 2025
· 11 min read
The Future of AI: Innovation Insight
Less Talking, More for the AI Agent
Doing Platform
RESEARCH 21 October Landscape
2024 · 2 min read RESEARCH 26 March 2025
AI is moving fast.
· 14 min read
More than ever, AI
techniques should
bring adaptability to
an uncertain world.
Despite its
extraordinary power
and early promises,
we need...
Recommended Multimedia
How to Calculate
Business Value Generative AI:
When Not to Use
and Cost for From Emergence
Generative AI
Generative AI Use to Hype to Doubt
VIDEO 19 March 2024
Cases VIDEO 22 August 2023
VIDEO 25 March 2025
https://www.gartner.com/document-reader/document/6554602?ref=solrResearch&refval=479953854& 12/13
11/07/2025, 08:20 Answering the Top Trending AI Agent Questions, Part 4: The Future of AI Agents
01:11 00:55 06:26
View More
© 2025 Gartner, Inc. and/or its affiliates. All rights reserved. Gartner is a registered
trademark of Gartner, Inc. and its affiliates. This publication may not be reproduced or
distributed in any form without Gartner's prior written permission. It consists of the opinions
of Gartner's research organization, which should not be construed as statements of fact.
While the information contained in this publication has been obtained from sources
believed to be reliable, Gartner disclaims all warranties as to the accuracy, completeness or
adequacy of such information. Although Gartner research may address legal and financial
issues, Gartner does not provide legal or investment advice and its research should not be
construed or used as such. Your access and use of this publication are governed
by Gartner's Usage Policy. Gartner prides itself on its reputation for independence and
objectivity. Its research is produced independently by its research organization without
input or influence from any third party. For further information, see "Guiding Principles on
Independence and Objectivity." Gartner research may not be used as input into or for the
training or development of generative artificial intelligence, machine learning, algorithms,
software, or related technologies.
POLICIES PRIVACY POLICY TERMS OF USE OMBUDS Get The App
CONTACT US
© 2025 Gartner, Inc. and/or its affiliates. All rights reserved.
https://www.gartner.com/document-reader/document/6554602?ref=solrResearch&refval=479953854& 13/13
