AI agent and Agentic AI applications: Emerging fields from latest patents

The potential of AI agents and Agentic AI is being actively debated across a wide range of industries. This article surveys the state of patent filings worldwide and what those patents actually contain.
We hope it serves as a useful reference for anyone weighing how to put AI to work or planning new lines of business.

How the Patents Were Selected

The patents covered here were selected using the following procedure.

1. Database

   • World Intellectual Property Organization (WIPO) PATENTSCOPE

2. Search keywords

   • ("AI Agent" OR "Agentic AI")

3. Fields searched

   • Title and Abstract

4. Initial search results

   • 610 patents identified

5. Screening

   • Time period

     • Patents published before 2016 were excluded. This cutoff was chosen to capture the technical progress that followed the introduction of the Transformer architecture, which underpins today's LLM-based agents (Vaswani et al., 2017)

   • Deduplication

     • Duplicates within patent families were consolidated so that only one patent per invention was included in the analysis.

6. Result

   • The final set analyzed here consists of 500 patents.

Filing Landscape

As a baseline, we first look at some basic statistics: where the patents were filed, when, and under which patent classifications.

Region

By region, the United States (US) leads in the number of filings, followed by China (CN) and South Korea (KR).

Timing

Following the release of the Transformer in 2017 and of BERT and GPT-1 in 2018, AI agent–related patent filings rose only modestly. Since the launch of GPT-3.5 in November 2022 and GPT-4 in March 2023, however, the number of patents has grown dramatically.

Patent classification

By classification, the largest share of filings falls under information processing (Section G), followed by mechanical engineering (F), electricity and telecommunications (H), and transport and traffic (B).

Abbreviations

• G06F Electric digital data processing

• G06N Computing arrangements based on specific computational models

• G06Q Information and communication technology specially adapted for administrative, commercial, financial, managerial or supervisory purposes

• H04L Transmission of digital information

• G10L Speech analysis or speech synthesis

• G16H Healthcare informatics

• G06V Image or video recognition or understanding

• G06T Image data processing or generation, in general

• H04W Wireless communication networks

• B60W Conjoint control of vehicle sub-units of different type or different function

• G05B Control or regulating systems in general

• G08G Traffic control systems

• H04M Telephonic communication

• A61B Diagnosis; surgery; identification

• H04N Pictorial communication

• A63F Card games, board games, roulette games

• B25J Manipulators

Technical Approaches

From here, we turn to the technical approaches described in the filed patents.

Models and training methods

Rather than simply plugging in a large language model (LLM), the proposed methods combine multiple training techniques and architectures to give agents a high degree of autonomy, domain expertise, and adaptability. The field is shifting from "static foundation models" toward "dynamic agent systems that keep learning."

• An LLM controls the agent's behavior (Mccarson, 2024b; O’Malia & Denmon, 2022)

  • Using a "very large language model" as the controller for an AI agent

  • Combining generative AI with Agentic AI in healthcare (an AI nurse platform), cybersecurity, and industrial equipment control

• Optimizing decision-making through reinforcement learning (Kuusela et al., 2024)

  • Agents based on deep reinforcement learning are employed in semiconductor design, thermal ablation planning for liver tumors, and control of treatment delivery

  • Using reinforcement learning to give in-game NPCs and conversational agents human-like behavior and natural dialogue flow

• An ensemble approach that, instead of relying on one giant model to solve everything, combines many task-specific "Narrow AI" models (Odinaev, 2024; Raichelgauz & Harel, 2025; Sohum et al., 2025b)

  • Coordinating a suite of small models, each optimized for a specific function such as manufacturing, vehicle control, or emergency braking

  • Systems that apply AI models composed of multiple sub-models, and cooperative AI agent systems that share resources

• Mechanisms for continuous adaptation and improvement based on feedback from the environment and users (Self-Evolution) (Shah, 2025)

  • Mechanisms by which an agent autonomously evaluates and corrects its own performance in order to improve alignment with humans

  • Implementations that pair large language models with knowledge graphs, and design-support systems using RAG (Retrieval-Augmented Generation), which incorporate external knowledge and improve inference accuracy without retraining the model

Architecture

Rather than running a single AI model (such as an LLM) on its own, the patents propose "composite systems" that combine memory, planning, tool use, and safety mechanisms. Put another way, these are evolving into organizational systems: a "team of specialists" (multi-agent) is assembled, a "manager" (control layer) supervises it, and the team is equipped with "external memory" (memory modules) and a "manual" (planning module) to carry out its work.

• Distributed, cooperative architectures that coordinate multiple function-specific agents instead of relying on a single general-purpose agent (Raichelgauz & Harel, 2025; Sohum et al., 2025a)

  • Combining a group of small, task-specific agents to increase the robustness and expertise of the system as a whole

  • Sharing resources among multiple AI agents and making recommendations and decisions through a Coordinator (Distributed Agentic AI)

• Lifecycle management and self-evolution (邓天悦 et al., 2025)

  • Platforms for operations (AIOps, the AI counterpart of DevOps) and self-improvement

  • Zero-code systems for developing and deploying agents, and development methods built on low-code platforms

Reliability and governance

• Cryptographic and blockchain technologies are integrated to guarantee "who" an agent operating in a distributed environment is, and "who" is operating it (T. Chen, 2025; Moore, 2015)

  • A meta-control layer built into the architecture monitors and governs an agent's reasoning and actions to prevent autonomous agents from running out of control or behaving inappropriately (for example, hallucinating)

  • Frameworks for assessing an agent's complexity and training status, and blockchain-based systems for auditing and attesting to an agent's actions

  • Voiceprint authentication, and authentication that combines proof of ownership with biometrics

  • Systems for detecting adversarial AI attacks

• Agents that make ethical judgments and act on calculated risk (Crabtree et al., 2025)

  • Decision-making platforms equipped with deontic reasoning, enabling judgments grounded in the obligations and rules that must be upheld

  • Functions for monitoring and correcting regulatory compliance in real time, and service-delivery devices grounded in ethics

Application Areas

This section surveys the application areas covered by the filed patents.

Healthcare

AI's role is expanding from "consultant" (chatbot) to "practitioner" (treatment control, surgical planning, drug discovery).

• Autonomous patient care and management (Kariguddaiah, n.d.; Vakili & Nayak, 2025)

  • Virtual agents that monitor patients around the clock and deliver personalized care. Rather than merely answering questions, they continuously track the patient's condition and intervene proactively.

  • An "AI nurse" platform that combines LLMs (large language models) with deep learning to improve personalized patient care and facility efficiency

  • A voice-based system for managing type 2 diabetes, LLM-powered mental health support tools, and psychological health screening systems for university students and adolescents

  • Personalized AI agents that act as a "case manager" for each patient, and systems for managing medication

• Control of treatment devices, surgical planning, and treatment control via reinforcement learning (Inamoto, 2000; Kuusela et al., 2024)

  • An AI proxy that automatically plans and guides thermal ablation of liver tumors, and a 3D surgical planning device for joint replacement surgery

  • Agents that interface directly with medical devices, such as a learning-capable oxygen flow control and monitoring system and an AI platform that works with RFID-enabled medical injectors

• Drug discovery, genomic analysis, and research support (Mccarson, 2025)

  • An AI system for identifying compounds to use in combination therapies, a system for processing molecular data on drugs, and a system for searching compound information on traditional Chinese medicines

  • Generative AI / agentic AI systems for medical and genetic analysis, diagnosis, and treatment recommendations

Manufacturing and infrastructure

• Smart manufacturing and process control(Mccarson, 2024a; Odinaev, 2024)

  • Generative AI and agentic AI systems for the autonomous control of manufacturing processes (analyzing, controlling, and optimizing industrial equipment and manufacturing systems)

  • Optimization of specific processes, such as the automated optimization of converter smelting and intelligent decision-making in production management for machinery manufacturing

  • Managing an entire factory by combining multiple small agents, each specialized in a particular task, as in an Ensemble of Narrow AI Agents for Manufacturing

  • In the semiconductor field, SoC (System on Chip) circuit design using reinforcement learning

• Autonomous operation of power and energy infrastructure(Duan et al., 2021)

  • In power grid management, agents autonomously balance supply and demand and maintain equipment

  • Advanced infrastructure management approaches that combine blockchain and AI

  • Autonomous voltage control in power systems, and AI-agent-based intelligent early warning and maintenance management for power equipment

• Self-optimizing telecommunications networks (AIOps)(McCord et al., 2025)

  • Training AI/ML models and leveraging contextual information in wireless networks, and improving the power and spectral efficiency of NR (New Radio) networks

  • Autonomous operations (AIOps), including context-aware AI assistants that troubleshoot network problems, automated root cause analysis, and the automation of network management activities

Business

• Advanced automation of business processes (RPA × AI agents)

  • Using AI agents to automatically generate process execution code and commands for RPA

  • Optimization systems for digital employees, and applications to creating and checking evaluation accounts

• IT operations and software development (AIOps & DevOps)(H. Chen et al., 2025; Fedoruk et al., 2025)

  • Automated code generation using multiple AI agents, systems that generate DevOps (development and operations) pipelines based on large language models, and agent development on low-code development platforms

  • Agents that resolve IT issues, agents that pre-build and configure cloud services, and systems that perform root cause analysis

• Specialization in specific business functions (marketing, HR, finance)(Umeda, 2023)

  • Generating marketing materials, generating advertising landing pages, shop growth management systems, everything from content generation to sales promotion, and commercial intelligent analytics

  • Building HR data warehouses, executing corporate online education and training plans, and recruitment support systems

• Customer engagement and service delivery

  • Monitoring conversation sessions and proactively offering ordering or reservation services based on context

  • By delivering seamless chat services across multiple messengers, regardless of the messaging platform (app), businesses can respond to customers consistently no matter which app they use

Mobility

Rather than serving merely as navigation aids, agents are proposed as "autonomous decision-makers" that directly control vehicles, ensure safety in emergencies, and optimize traffic flow across entire cities.

• Vehicle control and enhanced safety through AI ensembles(Raichelgauz & Harel, 2025)

  • Breaking driving tasks down into components, with each agent responsible for assessing a specific situation

  • In safety-critical operations where lives are at stake, such as emergency braking, specialized agents guarantee a high level of reliability

  • Technologies in which agents autonomously handle everything from environmental perception to future prediction and action decisions, along with technologies for situational response and learning in autonomous driving control systems

• Infrastructure management and traffic optimization

  • AI agents for monitoring and managing not only individual vehicles but entire road and rail networks

  • Methods for inferring traffic conditions across highway networks, and the dynamic generation of personalized traffic reports based on AI agents

Conclusion

The commercialization of AI agents and agentic AI is expected to advance significantly in the years ahead. Understanding the current patent landscape should serve as a valuable reference for research and development.

References

• Chen, H., Liu, K., Tong, J., Wang, J., Xu, Z., Ben Romdhane, S., & Yella, V. R. (2025). Multiple agent automatic root cause analysis (USPTO Patent No. 20250274336:A1). In US Patent (No. 20250274336:A1). https://patents.google.com/patent/US20250274336A1/en

• Chen, T. (2025). Blockchain-based artificial intelligence agent life cycle management and authentication systems and methods (USPTO Patent No. 12483411). In US Patent (No. 12483411). https://patents.google.com/patent/US12483411B1/en

• Crabtree, J., Kelley, R., Hopper, J., & Park, D. (2025). Ai agent decision platform with deontic reasoning (USPTO Patent No. 20250259041:A1). In US Patent (No. 20250259041:A1). https://patents.google.com/patent/US20250259041A1/en

• Duan, J., Wang, S., Shi, D., Diao, R., Zhang, B., & Lu, X. (2021). SYSTEMS AND METHODS OF AUTONOMOUS VOLTAGE CONTROL IN ELECTRIC POWER SYSTEMS. In Patent. Google Patents. https://patentimages.storage.googleapis.com/e9/5a/95/39c4bddb5e6843/US20210143639A1.pdf

• Fedoruk, R., Manton, J., Reagan, S., Roberts, G., & Stuntebeck, E. (2025). Dynamic execution of artificial intelligence agents through device management (USPTO Patent No. 12493540). In US Patent (No. 12493540). https://patents.google.com/patent/US12493540B1/en

• Inamoto, Y. (2000). RFID-ENABLED MEDICAL INJECTORS AND ARTIFICIAL INTELLIGENCE (AI) PLATFORM FOR PRE-INJECTION INTERROGATION. US Patent 6,049,891. https://patentimages.storage.googleapis.com/b6/6f/c6/96d2592a3eb0a8/US6049891.pdf

• Kariguddaiah, A. (n.d.). AI NURSE PLATFORM: AN ARTIFICIAL INTELLIGENCE, CONVERSATIONAL (AI AGENTIC) AI NLP, GENERATIVE-AI & DEEP LEARNING LARGE LANGUAGE MODEL/S (LLM) DRIVEN VIRTUAL ASSISTANT/S FOR ENHANCED PATIENT PERSONALIZED CARE OUTCOMES & FACILITY EFFICIENCIES. Retrieved February 11, 2026, from https://patentscope.wipo.int/search/en/detail.jsf?docId=US471591817

• Kuusela, E. H., Basiri, S., Czeizler, E., Hakala, M. O., & Halko, L. J. (2024). Methods and apparatus for controlling treatment delivery using reinforcement learning (USPTO Patent No. 12068065). In US Patent (No. 12068065). https://patentimages.storage.googleapis.com/da/aa/96/bbb079e179ba37/US12068065.pdf

• Mccarson, B. (2024a). Generative AI and agentic AI systems and methods for industrial equipment and manufacturing systems analytics, control, and optimization (USPTO Patent No. 12140915). In US Patent (No. 12140915). https://patentimages.storage.googleapis.com/f9/12/eb/3e8365c194bdeb/US12140915.pdf

• Mccarson, B. (2024b). Generative AI and agentic AI systems and methods for prevention, detection, mitigation and remediation of cybersecurity threats (USPTO Patent No. 12088599). In US Patent (No. 12088599). https://patentimages.storage.googleapis.com/18/79/a7/6a66bf399c410d/US12088599.pdf

• Mccarson, B. (2025). Agentic AI, contextual AI, and generative AI systems and methods for medical and genetic analytics, diagnostics, and treatment recommendations (USPTO Patent No. 12450284). In US Patent (No. 12450284). https://patents.google.com/patent/US12450284B2/en

• McCord, A. W., Marin, G. M. D., & Sheppard, R. J. (2025). Adaptive communication management system using model-based and model-free reinforcement learning (USPTO Patent No. 12261893). In US Patent (No. 12261893). https://patentimages.storage.googleapis.com/5d/57/65/51f04bf3f4897f/US12261893.pdf

• Moore, D. C. (2015). METHOD, INTERACTIVE AI AGENT SYSTEM AND COMPUTER-READABLE RECODING MEDIUM FOR PROVIDING SEMANTIC-FREE USER VOICEPRINT AUTHENTICATION HAVING ENHANCED SECURITY. US Patent 8,930,815. https://patentimages.storage.googleapis.com/22/d2/98/5288a22e4236bf/US8930815.pdf

• Odinaev, K. (2024). Ensemble of Narrow AI agents for Manufacturing (USPTO Patent No. 20240242329:A1). In US Patent (No. 20240242329:A1). https://patentimages.storage.googleapis.com/e0/25/52/fe112db3009f58/US20240242329A1.pdf

• O’Malia, J. A., & Denmon, Z. (2022). Ultra large language models as ai agent controllers for improved ai agent performance in an environment (USPTO Patent No. 20220036153:A1). In US Patent (No. 20220036153:A1). https://patentimages.storage.googleapis.com/38/36/4a/a225c445cfb5ef/US20220036153A1.pdf

• Raichelgauz, I., & Harel, A. (2025). Ensemble of narrow AI agents for autonomous emergency breaking (USPTO Patent No. 12479423). In US Patent (No. 12479423). https://patents.google.com/patent/US12479423B2/en

• Shah, S. P. (2025). Rubric based self learning methods and systems in an artificial intelligence environment (USPTO Patent No. 20250299054:A1). In US Patent (No. 20250299054:A1). https://patents.google.com/patent/US20250299054A1/en

• Sohum, A. K., Foong, C. Y. J., & Ramakrishna, M. (2025a). Collaborative artificial intelligent (ai) agent systems with coordinators/recommendations for shared applications and methods thereof (USPTO Patent No. 20250356316:A1). In US Patent (No. 20250356316:A1). https://patents.google.com/patent/US20250356316A1/en

• Sohum, A. K., Foong, C. Y. J., & Ramakrishna, M. (2025b). Systems and methods for augmenting responses/recommendations and generating enhanced decisions through resource sharing between artificial intelligent (ai) agents (USPTO Patent No. 20250355915:A1). In US Patent (No. 20250355915:A1). https://patents.google.com/patent/US20250355915A1/en

• Umeda, T. (2023). Recruitment support system, recruitment support method, and program. In Patent. http://jglobal.jst.go.jp/public/202303019907126895

• Vakili, S., & Nayak, A. K. (2025). Voice-based method and system for management of type 2 diabetes (USPTO Patent No. 12251242). In US Patent (No. 12251242). https://patentimages.storage.googleapis.com/f4/aa/33/035fc239e4494e/US12251242.pdf

• Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30. https://proceedings.neurips.cc/paper/7181-attention-is-all

• 邓天悦, 李亚楠, 黄英杰, 袁婷, 马云洁, 陈明, & 韩华瑞. (2025). Method, medium and device based on AI agent and low-code generation model (Patent No. CN:119088378:B). In Patent (No. CN:119088378:B). https://patentimages.storage.googleapis.com/1e/2a/15/a7c878cd8178b8/CN119088378B.pdf

Appendix: List of patents covered (Application IDs)

US465822756
US469947015
US458062494
US438361528
US468508362
WO2025240652
US436106679
US472300303
US471833949
US471591817
AU470572375
US468861356
AU470572010
US471590829
US470603097
US471591850
CN417774840
CN432992756
WO2021084510
US461039877
US468461747
CN444137684
CN454338066
WO2025218356
US468462219
CN454487144
CN432885271
CN453000210
US468105106
US428680229
US468105518
CN444141545
US468462125
CN456694529
CN430578390
CN444453465
CN465775680
CN447744862
CN453081433
US472300490
US468105964
CN439186456
CN447744855
CN447250669
CN453087344
CN441985626
US468150301
CN437270409
US463552442
US465147979
CN444137683
US471309826
KR340252306
CN436538715
US469223574
CN432067482
US363143033
CN437332720
CN450096010
WO2025165454
US320341565
CN454506135
CN455789528
WO2025239713
CN428729880
WO2026028983
US461039796
CN437937266
US453094112
WO2023043695
WO2025221497
EP471600480
CN468784878
US468945591
WO2025254715
KR250898438
CN436469982
US465733753
CN451633888
EP467052499
US471309825
WO2023106446
CN471745226
US466435018
CN458538980
CN397769936
US468108101
US440284564
WO2026004526
CN457483947
JP443803171
CN436039493
CN449318624
CN458146134
CN453704396
CN454807033
WO2019156537
WO2023110478
KR251664886
US462218150
WO2025191572
CN469899062
US471833694
KR235301818
CN436540463
CN457494039
US455863280
WO2019132135
CN454508080
US468462447
IN334771181
MY446541712
US366100411
CN437343326
CN470523217
US472301538
KR250898439
CN429428779
CN436448717
CN440615225
CN447733691
US462838556
WO2025240965
CN427207925
CN467563462
US464101387
US251447284
CN438262359
US468105594
CN437337872
CN454313760
US468861337
KR251664862
US282302954
WO2025224432
CN468803343
CN448142885
US462837975
US468105113
US468507536
CN444876729
KR438349588
US470469487
US419732965
US464434280
CN471451036
CN441332374
CN459239933
US471591706
US396244490
JP444567521
US449217490
CN462331955
CN453150220
WO2018081833
WO2023085165
KR251664863
US464101419
CN460836721
WO2019156536
US469223082
CN468186867
US435545749
CN468571259
US468464175
CN465203502
CN471553880
KR365495002
US468105276
GB461052116
CN468798050
US323753444
US398110382
CN427212140
CN445653522
US459197968
US289043788
CN454526597
US391538821
CN471544876
CN438677885
US445576217
US456554215
WO2024064030
CN454958559
US471684621
KR356031543
US459618297
US458220044
WO2019156535
WO2022096942
WO2025235329
WO2025263923
US470605276
US375621248
JP435089957
JP453616146
US459873173
US470605281
US447188664
CN446890450
US468105114
WO2021090299
KR323635485
US349430640
JP441005757
JP450891794
CN467567358
CN467114583
CN468183922
KR243441267
CN442930340
US464434158
CN451626200
CN459671511
US461615949
US468105443
WO2026006913
CN456618652
CN451616205
CN451635234
US468106700
US468860825
US472300128
KR250898432
CN436979631
US462837324
WO2024064022
US471591135
CN464225158
CN464234032
CN471712261
KR236560074
JP432844376
CN438263465
CN444854171
US446174165
US460152088
US462220115
CN463214160
US471029110
CN470660364
US472300056
US398808168
CN431084135
CN451648754
US461040623
CN463499405
WO2019143141
US253706009
CN362200722
CN426570855
CN443422839
US458062717
WO2019124647
EP469233098
WO2021191111
WO2025074187
KR276060327
US424798713
CN437332726
KR467128603
CN465778987
CN458909657
CN467577966
US466435445
WO2025080402
CN455161387
US468461652
WO2025238628
CN467171673
CN466493323
CN346621539
US458061333
US456554840
US456557307
CN461828337
WO2022234930
US469223425
CN470507463
CN424935942
US431349475
CN432782673
CN436028395
CN467557261
US455634904
WO2019098638
WO2019157079
US451306352
US470469378
KR250898431
US361300888
KR414660073
KR432948220
US468108117
WO2025186744
CN463856576
CN471452876
US319906206
EP411272686
JP453624219
CN457483829
US465733337
CN461949910
US468105213
CN463243081
WO2025128158
US468861326
CN469127137
US468945666
WO2025250514
US471307270
KR251667878
KR260732870
CN420675918
CN463499913
JP450204408
WO2019168235
WO2025243204
CN469289508
WO2026007836
KR231643209
US373269119
CN347706760
CN346621514
CN453041105
US461615792
US323754858
CN346621515
US444402491
CN459685956
US461118204
CN460824318
CN462338277
US463436795
WO2019066132
WO2025106992
US435545875
US471028439
US357569112
US417726827
CN449311522
US458052755
US451438407
JP454280168
CN450099826
US456997171
WO2022195432
US468861738
CN464173473
WO2026015679
US471591858
CN308316758
KR391086587
US460791251
AU461827546
US468105511
CN468796289
US468464197
CN465209651
CN463841400
KR444078141
EP391888626
CN444452160
US455635233
JP453619759
AU464529348
US470602844
WO2025200581
KR250160682
US295247818
KR391086589
CN397478723
CN437954491
CN448138771
US455617327
CN451868507
CN453711600
CN467581635
JP453624221
US471590889
CN444139599
US449219488
US462218520
CN460823632
US471030780
KR442915027
US471309763
US471309878
US221772214
US281657371
KR323635483
US328944877
CN413703137
US446175105
CN445622057
JP451849915
AU447232948
CN440757145
US459197965
WO2018075658
WO2024063897
WO2024064197
US469223504
US468461867
US471591159
US470124917
US471833708
CN467578374
US427163859
JP454279726
CN450079613
WO2025042388
US469780268
US338560888
US381471253
EP426737269
JP437927317
US461040189
US456557366
WO2017177128
WO2021026385
WO2025193303
CN455167526
US468861620
US471028580
US471835557
IN294774678
CN445788879
US468105939
US464434524
CN454514858
US469223714
US471307698
KR340256968
AU400342297
CN443831002
US448392711
US463439084
US461040234
WO2025168947
CN455790204
CN469135451
WO2025244708
WO2025250549
CN464175233
US369783907
CN448140312
US430560678
US451827244
US455625483
WO2018184193
WO2024064021
WO2021220406
US468861707
US471030774
EP461050197
WO2023126063
CN467579698
US393004729
JP453615257
JP453620059
JP458182130
CN457149127
CN444137294
US467443020
WO2018128653
WO2020235693
US468864244
US468461878
US468461888
CN465503292
US471309729
US471833945
WO2025226216
KR240192500
US376427952
US428014959
AU458079570
US447188671
US465983088
WO2019103569
WO2022177103
WO2025250335
KR375104658
US429028758
CN439132078
US452615198
CN457949923
CN458539106
AU467472460
CN459670897
CN469129090
WO2025199119
US282302775
US459200096
WO2020012861
WO2024064025
WO2025068280
US471591158