CoChat

The Rise of Agentic AI: A Review of Definitions, Frameworks, Architectures, Applications, Evaluation Metrics, and Challenges

Open access

The DOI matches its Crossref record; 136 of the 137 references it cites were found in a registry or an index, and 1 are not checked yet; 75 of the 457 sources its claims cite were checked for support.

Abstract

Agentic AI systems are a recently emerged and important approach that goes beyond traditional AI, generative AI, and autonomous systems by focusing on autonomy, adaptability, and goal-driven reasoning. This study provides a clear review of agentic AI systems by bringing together their definitions, frameworks, and architectures, and by comparing them with related areas like generative AI, autonomic computing, and multi-agent systems. To do this, we reviewed 143 primary studies on current LLM-based and non-LLM-driven agentic systems and examined how they support planning, memory, reflection, and goal pursuit. Furthermore, we classified architectural models, input–output mechanisms, and applications based on their task domains where agentic AI is applied, supported using tabular summaries that highlight real-world case studies. Evaluation metrics were classified as qualitative and quantitative measures, along with available testing methods of agentic AI systems to check the system’s performance and reliability. This study also highlights the main challenges and limitations of agentic AI, covering technical, architectural, coordination, ethical, and security issues. We organized the conceptual foundations, available tools, architectures, and evaluation metrics in this research, which defines a structured foundation for understanding and advancing agentic AI. These findings aim to help researchers and developers build better, clearer, and more adaptable systems that support responsible deployment in different domains.

Verified 2026-10-10 · 1 reference not checked yet

Overview

CoChat Summary

TL;DR

This review synthesizes 143 primary studies to define agentic AI, compare it with generative AI, autonomic computing, and multi-agent systems, and map its frameworks, architectures, applications, input-output formats, evaluation metrics, and challenges. It reports rapid growth in 2024–2025 and identifies recurring architectural and evaluation approaches, while highlighting unresolved technical, coordination, ethical, and security limitations.

Key findings

  • The review analyzed 143 primary studies; more than 90% were published in 2024 and 2025. Of these, 46% were peer-reviewed journals, 28% preprints, and 17% conference proceedings (methodology text).
  • LLM-based frameworks such as LangChain, AutoGPT, BabyAGI, AutoGen, OpenAgents, CAMEL, MetaGPT, SuperAGI, and TB-CSPN enable planning, memory, reflection, tool use, and goal pursuit; non-LLM systems use rules (CLIPS, Drools, JADE), RL (Q-learning, DQN, policy gradients), planning (STRIPS/PDDL), and simulation (NetLogo, MASON).
  • Five architectural models recur: ReAct single-agent, supervisor/hierarchical, hybrid reactive-deliberative, BDI, and layered neuro-symbolic; core components include perception, memory, planning, execution, reflection, orchestration, and interaction. Trade-offs include BDI explainability vs. brittleness, hierarchical scalability vs. supervisor bottlenecks, ReAct lightweight loops vs. limited parallelism, hybrid speed/planning vs. arbitration complexity, and neuro-symbolic uncertainty handling vs. integration overhead.
  • Evaluation combines testing methods and qualitative/quantitative metrics. Reported examples include ~94% accuracy for LangGraph/CrewAI, 96% for Claude 3.5/LLaMa-3.2, 12 ± 2 min task completion time (34.2% improvement over traditional AI), and ~94% rule fidelity; challenges include 78% of organizations reporting data safety problems, plus architectural, tool-integration, coordination, human-AI interaction, ethical/governance, and security issues.

Methods

  • Literature review/survey: searched Library Search, ACM Digital Library, ScienceDirect, Google Scholar, and Semantic Scholar with a query combining agentic AI/agent/autonomous AI/multiagent systems/autonomic computing/generative AI/AI planning/AI memory systems/AI feedback loops and survey/overview/review/summary/literature review terms; removed duplicates and screened abstracts/conclusions.
  • Sample: 143 primary studies published from 2005 to present, with >90% from 2024–2025; also examined 21 survey/review articles across six dimensions using L/M/H/NA ratings.
  • Analysis: extracted key information and synthesized definitions, frameworks, architectures, applications, input-output formats, evaluation metrics/testing methods, and challenges, using narrative synthesis and tabular summaries.

Limitations

  • The paper does not state limitations of the review itself; it lists limitations/challenges of agentic AI systems: architectural/technical (dynamic task decomposition, hierarchical planning, adaptive reasoning, symbolic rigidity, long-term strategic planning, data volume/velocity/variety/veracity, opaque architectures), performance/tool integration (computational inefficiencies, lack of standard interfaces, limited design/test/deploy support), multi-agent coordination (communication, synchronization, scaling, correlated failures/collusion), human-AI interaction (limited transparency/interpretability, insecure interface design, misalignment with human intention), ethical/societal/governance (privacy/security risks; 78% of organizations reported data safety problems; accountability; communication failures; poor data quality), and security (adversarial attacks, data poisoning, misuse of tools/APIs, lack of oversight).
  • Evident from the text: it is a narrative review without a stated formal quality appraisal, risk-of-bias assessment, or meta-analysis; no review-level limitations section is provided.
  • Evident from the text: some reported publication counts appear inconsistent across sections (e.g., methodology text reports 32 arXiv, 27 conference proceedings, and 43 journal-based papers, while Table A2 reports 70 journals, 38 preprints, and 26 conference proceedings).

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/fi17090404

Cites

136 of 137 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

56 Valid · 5 Not valid · 196 Not checked · 42 Own results and interpretations

80 more claims · 13 with their cited sources checked

Create a free account to see all 299 claims and to verify the 113 references still open.

See all claims — free account
  • It may state two separate claims as one. 137
  • Probably not a substantial claim. 131

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

Research works printed but not linked: 28

  1. Jin, H., Huang, L., Cai, H., Yan, J., Li, B., and Chen, H. (2024). From llms to llm-based agents for software engineering: A survey of current, challenges and future. arXiv.
  2. Sami, A.M., Rasheed, Z., Kemell, K.K., Waseem, M., Kilamo, T., Saari, M., Duc, A.N., Systä, K., and Abrahamsson, P. (2024). System for systematic literature review using multiple ai agents: Concept and an empirical evaluation. arXiv.
  3. Schneider, J. (2025). Generative to agentic ai: Survey, conceptualization, and challenges. arXiv.
  4. Pavani, S., and Shwetha, H. (2025, August 14). Agentic AI: Redefining Autonomy for Complex Goal-Driven Systems. Available online: https://scholar.google.com/scholar?hl=en&as_sdt=0%2C26&q=Agentic+AI%3A+Redefining+Autonomy+for+Complex+Goal-Driven+Systems.&btnG=.
  5. Botti, V. (2025). Agentic AI and Multiagentic: Are We Reinventing the Wheel?. arXiv.
  6. Huang, L., Koutra, D., Kulkarni, A., Prioleau, T., Wu, Q., Yan, Y., Yang, Y., Zou, J., and Zhou, D. (May, January 28). Towards Agentic AI for Science: Hypothesis Generation, Comprehension, Quantification, and Validation. Proceedings of the Companion Proceedings of the ACM on Web Conference, Sydney, Australia.
  7. Wissuchek, C., and Zschech, P. (2025). Exploring Agentic Artificial Intelligence Systems: Towards a Typological Framework. arXiv.
  8. Raidas, M.A., and Bhandari, R. (2025). Agentic AI in Education: Redefining Learning for the Digital Era. Artificial Intelligence in EducationTransforming Learning for the Future, Scholars’ Press.
  9. (2025, January 19–23). Multimodal Agentic Model Predictive Control. Proceedings of the 24th International Conference on Autonomous Agents and Multiagent Systems, Detroit, MI, USA.
  10. Paleti, 2024, Agentic AI in Financial Decision-Making: Enhancing Customer Risk Profiling, Predictive Loan Approvals, and Automated Treasury Management in Modern Banking, Multidiscip. Sci. Work. Manag. J., 34, 832
  11. Trirat, P., Jeong, W., and Hwang, S.J. (2025). Agentic Predictor: Performance Prediction for Agentic Workflows via Multi-View Encoding. arXiv.
  12. Hong, S., Zhuge, M., Chen, J., Zheng, X., Cheng, Y., Zhang, C., Wang, J., Wang, Z., Yau, S.K.S., and Lin, Z. (2024, January 7–11). MetaGPT: Meta programming for a multi-agent collaborative framework. Proceedings of the International Conference on Learning Representations, ICLR, Vienna, Austria.
  13. Liu, Z., Yao, W., Zhang, J., Yang, L., Liu, Z., Tan, J., Choubey, P.K., Lan, T., Wu, J., and Wang, H. (2024). AgentLite: A Lightweight Library for Building and Advancing Task-Oriented LLM Agent System. arXiv, 15538.
  14. Thirumalainambi, R. (2007, January 9–12). Pitfalls of JESS for Dynamic Systems. Proceedings of the Artificial Intelligence and Pattern Recognition, Citeseer, Orlando, FL, USA.
  15. Silver, T., Hariprasad, V., Shuttleworth, R.S., Kumar, N., Lozano-Pérez, T., and Kaelbling, L.P. (2025, August 14). PDDL planning with pretrained large language models. In Proceedings of the NeurIPS 2022 Foundation Models for Decision Making Workshop. Available online: https://drive.google.com/file/d/1PqCYzzfdUFitxG7NxFs0D6dn-TzI97q_/view.
  16. (2025, January 16–19). Nurturing Code Quality: Leveraging Static Analysis and Large Language Models for Software Quality in Education. In Proceedings of the Adjunct 33rd ACM Conference on User Modeling, Adaptation and Personalization. New York, NY, USA.
  17. Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K., and Cao, Y. (2023, January 1–5). ReAct: Synergizing Reasoning and Acting in Language Models. Proceedings of the International Conference on Learning Representations (ICLR), Kigali, Rwanda.
  18. Kim, J., Wahi-Anwa, M., Park, S., Shin, S., Hoffman, J.M., and Brown, M.S. (2025). Autonomous Computer Vision Development with Agentic AI. arXiv.
  19. Jeyakumar, S.K., Ahmad, A.A., and Gabriel, A.G. (2024, January 16). Advancing Agentic Systems: Dynamic Task Decomposition, Tool Integration and Evaluation using Novel Metrics and Dataset. Proceedings of the NeurIPS 2024 Workshop on Open-World Agents, Vancouver, BC, Canada.
  20. Casper, S., Bailey, L., Hunter, R., Ezell, C., Cabalé, E., Gerovitch, M., Slocum, S., Wei, K., Jurkovic, N., and Khan, A. (2025). The AI Agent Index. arXiv.
  21. Sheriff, A., Huang, K., Nemeth, Z., and Nakhjiri, M. (2025). ADA: Automated Moving Target Defense for AI Workloads via Ephemeral Infrastructure-Native Rotation in Kubernetes. arXiv.
  22. Bennett, J., and Sterritt, R. (2024, January 10–14). Autonomic Computing in Total Achievement of Quality. Proceedings of the The Twentieth International Conference on Autonomic and Autonomous Systems, Athens, Greece.
  23. Inala, 2025, Building Trustworthy Agentic Ai Systems FOR Personalized Banking Experiences, Metall. Mater. Eng., 31, 1336
  24. Huang, C., Huang, H., Yu, T., Xie, K., Wu, J., Zhang, S., Mcauley, J., Jannach, D., and Yao, L. (2025). A Survey of Foundation Model-Powered Recommender Systems: From Feature-Based, Generative to Agentic Paradigms. arXiv.
  25. Zhang, Q., Wornow, M., and Olukotun, K. (2025). Cost-Efficient Serving of LLM Agents via Test-Time Plan Caching. arXiv.
  26. (2025, August 14). AI Agents: Potential implications for IS Research?. Available online: https://www.sciencedirect.com/science/article/pii/S0963868725000216.
  27. Zhou, R., Sikand, V., and Rao, S. (2025, January 27). AI Agents for Deep Scientific Research. Proceedings of the UIUC Spring 2025 CS598 LLM Agent Workshop, Urbana, Illinois, USA. Available online: https://openreview.net/forum?id=wODNrFtTT2.
  28. Mukherjee, A., and Chang, H.H. (2025). Agentic AI: Expanding the Algorithmic Frontier of Creative Problem Solving. arXiv.

Cited by

See all 207

Metadata

Authors
Bandi, A., Kongari, B., Naguru, R., Pasnoor, S., & Vilipala, S. V.
Affiliations
  • Northwest Missouri State University (US)
  • Independent Researcher, Omaha, NE 68022, USA
Corresponding author
Bandi, Ajay
Venue
Future Internet
Publisher
MDPI AG
Published
2025-09-04
Volume
17
Issue
9
Pages
404
ISSN
1999-5903 (ISSN-L)
OpenAlex
W4414023998
Cited by
207
References
100
Citations by year
2026: 191 · 2025: 16
FWCI
82.37
Type
journal article
Language
en
Open access
Locations
  • Future Internet (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › AI-based Problem Solving and Planning
  • Physical Sciences › Computer Science › Artificial Intelligence › Artificial Intelligence Applications
  • Physical Sciences › Computer Science › Artificial Intelligence › Psychiatry, Mental Health, Neuroscience
Keywords
agentic AI, autonomous agents, large language models, agent architecture, planning, memory, reflection, multi-agent systems, autonomic computing, generative AI, evaluation metrics, agent coordination

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.