Agentic AI · since 2022
Agentic AI
Systems that plan, call tools, and act over multiple steps rather than answering in a single turn — the shift from LLM-as-oracle to LLM-as-actor. The current frontier of deployment: the model stops being something you query and becomes something that does work.
48
events traced
51
source records
30 Jun 2026
first signal
9 Jul 2026
last activity
Who drove it
Key movements
2022+↑
tool-use → multi-step
2024+↑
reasoning models make agents reliable
2025–26↑
enterprise adoption
The story, event by event
Every point below is traced to a real source — nothing on this page is invented.
6 Oct 2022impact 72
ReAct — reasoning + acting interleaved
value: ReAct; metric: pattern
9 Feb 2023impact 66
Toolformer — self-supervised tool use
value: self-taught tools; metric: capability
30 Mar 2023impact 64
AutoGPT — autonomous agents hit the mainstream
value: mainstream; metric: moment
13 Jun 2023impact 74
Tool use / function calling goes native
value: tool-use; metric: capability
22 Oct 2024impact 70
Computer-use / operator agents
value: GUI; metric: action-space
7 Oct 2025impact 60
Introduction of Gemini 2.5 Computer Use Model
use cases: UI testing, workflow automation; performance: outperforms leading alternatives on multiple web and mobile control benchmarks; safety features: includes safety features to mitigate risks associated with AI agents controlling computers
10 Nov 2025impact 60
AI Integration in Education
AI tools significantly enhance teaching efficiency in Northern Ireland.
The pilot captured more than 600 unique use cases for Gemini, ranging from streamlining administrative work to brainstorming more engaging content.
Following the success of the pilot, C2k plans to roll out Gemini training to more teachers across Northern Ireland.
It's critical that our products empower educators, not replace them.
This initiative highlights the potential of AI to transform educational practices, emphasizing collaboration between technology and educators.
1 Jan 2026impact 65
MUFG adopts ChatGPT Enterprise for large-scale agentic AI deployment
MUFG is rolling out ChatGPT Enterprise to 35,000 employees and developing AI-powered customer experiences, marking a significant enterprise adoption of agentic AI.
Mitsubishi UFJ Bank began a phased rollout of ChatGPT Enterprise in 2026 for 35,000 employees, aiming to integrate AI into daily workflows and extend human thinking.
Employees created over 1,800 custom GPTs in four months and reported a 20-30% workload reduction in research tasks, with a focus on reinvesting saved time into customer interactions.
MUFG plans to integrate an AI concierge and Money Advisory Platform (MAP) in its digital bank, and the Moneytree wealth management app with Apps in ChatGPT for natural language queries, to enhance customer experiences.
This demonstrates the growing maturity and enterprise readiness of agentic AI, moving beyond theoretical capabilities to large-scale, practical deployment in critical sectors like finance.
27 Jan 2026impact 60
Unlocking Agentic RL Capabilities in GPT-OSS
key argument: The article argues that meticulous engineering is essential to unlock agentic RL capabilities in the GPT-OSS model.; evidence basis: The article presents experimental results and engineering solutions that address specific issues in the GPT-OSS training process, demonstrating improvements in stability and performance.
18 Feb 2026impact 60
Understanding Failure Modes in Agentic Systems
failure modes: {'Kimi-K2': {'description': 'Issues with task completion recognition'}, 'GPT-OSS-120B': {'description': 'Cascading failures', 'failure_modes_per_trace': 5.3}, 'Gemini-3-Flash': {'description': 'Isolated failures', 'failure_modes_per_trace': 2.6}}; recommendations: Externalize Verification for models like Gemini
19 Feb 2026impact 60
Release of Gemini 3.1 Pro
score: 77.1; benchmark: ARC-AGI-2
27 Mar 2026impact 40
Revival of OpenClaw Agents
method 1: Hugging Face Inference Providers; method 2: Local Model Running
10 Apr 2026impact 60
ChatGPT Enhancements in Finance
examples: Structuring inputs, Drafting outputs, Standardizing workflows; description: ChatGPT's functionalities streamline finance processes.
13 Apr 2026impact 60
Integration of OpenAI models into Cloudflare's Agent Cloud
codex agents: Enterprises can now deploy agents built on Codex harness to Cloudflare.; business usage: More than 1 million business customers around the world are directly using OpenAI, 3 million weekly active users use Codex.; enterprises access: Millions of enterprises can now access OpenAI frontier models directly within Cloudflare Agent Cloud.
14 Apr 2026impact 60
Expansion of TAC Program
focus: cybersecurity; model: GPT-5.4-Cyber; program: Trusted Access for Cyber
16 Apr 2026impact 60
Launch of Trusted Access for Cyber initiative
funding: $10 million; initiative: Trusted Access for Cyber
20 Apr 2026impact 60
Hyatt's Deployment of ChatGPT Enterprise
departments: operations, finance, marketing; description: Hyatt has deployed ChatGPT Enterprise across multiple departments to enhance guest experiences and improve operational efficiency.; expected outcomes: boost productivity, streamline operations, enhance collaboration
21 Apr 2026impact 60
Advocating Openness in AI Cybersecurity
summary: The article discusses the significance of openness in AI cybersecurity, particularly through the example of the Mythos AI model, which effectively identifies and addresses software vulnerabilities.; key argument: Openness in AI cybersecurity is essential for effective defense against vulnerabilities.
22 Apr 2026impact 60
DeepMind's AI Adoption Initiative
adoption gap: 25%; economic impact: $15.7 trillion by 2030
27 Apr 2026impact 60
Choco's AI Integration in Food Distribution
order processing: {'annual_orders': 8800000, 'productivity_gain': '2x', 'manual_workflow_reduction': 'millions', 'manual_order_entry_reduction': '50%'}
30 Apr 2026impact 60
Advancement of AI Co-Clinician in Telemedicine
goal: responsible development and deployment of medical AI; global collaborations: Australia, New Zealand, US, Singapore, India, UAE
4 May 2026impact 60
Collaboration on AI Agents for Finance
CFOs: {'AI_usage': 'visibility into AI usage, token consumption, and projected spend', 'contract_processing': '5x more contracts processed using Codex'}
6 May 2026impact 65
Uber deploys AI assistants for operational efficiency
Uber integrates OpenAI's AI assistants and voice features to enhance operational efficiency and user experience for its vast global network.
Uber uses OpenAI to power AI assistants and voice features that help drivers earn smarter and riders book faster across a global real-time marketplace.
Uber operates at massive scale: 40 million trips per day, 10 million drivers and couriers across 15,000 cities in over 70 countries.
Uber designed Uber Assistant around three core principles: safety, trust, and low latency.
This deployment demonstrates the growing real-world application of agentic AI in large-scale, complex operational environments, highlighting its potential for significant efficiency gains and improved user engagement.
6 May 2026impact 55
Singular Bank's Singularity AI assistant improves banker efficiency
Singular Bank deployed an internal AI assistant, Singularity, powered by ChatGPT and Codex, to significantly enhance bankers' efficiency and client engagement.
Bankers using Singularity save 60-90 minutes daily on preparation and data analysis, allowing them to focus more on client interactions.
The system executed over 3,500 operations in a 30-day period, identifying key portfolio risks and imbalances such as concentration risk and elevated exposure.
Singularity is designed to enhance bankers' judgment, proximity, and impact on the client rather than replace them, signaling a strategic shift towards applied intelligence in private banking.
This case demonstrates a concrete, measurable impact of agentic AI deployment in a critical enterprise function, highlighting the ongoing shift towards AI-enhanced human roles and applied intelligence in the financial sector.
6 May 2026impact 65
Frontier firms deepen agentic AI adoption
Frontier firms are achieving significant competitive advantages by deeply integrating agentic AI into their workflows, moving beyond simple chat to delegated, complex tasks.
Frontier firms—those at the 95th percentile of usage—now use 3.5x as much intelligence per worker as typical firms, up from 2x a year ago.
Message volume explains only 36% of the frontier advantage; most of the gap comes from richer, more complex AI use.
The largest advantage shows up in advanced tools, with frontier firms sending 16x as many Codex messages per worker as typical firms.
Leading firms measure depth, build governance for production use, invest in enablement, scale what works, and move from chat-based assistance to delegated work with agents.
Travelers expects the assistant to handle approximately 100,000 first notice of loss calls in its first year.
This indicates a maturing phase of agentic AI adoption where strategic integration, rather than mere access, drives significant business impact and competitive differentiation.
7 May 2026impact 60
AlphaEvolve becomes core AI infrastructure
optimizations: {'effect': 'reduced write amplification by 20%', 'application': 'Google Spanner'}, {'effect': 'doubled training speed of transformer model', 'application': 'Klarna'}, {'effect': 'achieved 4x speedup in MLFF training and inference', 'application': 'Schrödinger'}
7 May 2026impact 65
Parloa's AMP automates enterprise customer service
Parloa launched an AI Agent Management Platform (AMP) to automate customer service interactions for enterprises.
AMP allows enterprises to design, deploy, and manage customer service interactions at scale.
A global travel company reduced human agent requests by 80% using Parloa's platform.
Parloa emphasizes a modular approach to improve agent instruction-following and foresees customer service evolving into a multimodal experience.
This demonstrates the increasing maturity and enterprise adoption of agentic AI for specific business functions, highlighting significant efficiency gains and the trend towards user-friendly agent management.
15 May 2026impact 70
GPT-5.5 achieves SOTA on OfficeQA Pro for enterprise agent workflows
GPT-5.5 sets a new state of the art on the OfficeQA Pro benchmark, significantly improving enterprise agent workflows.
GPT-5.5 achieved 50% accuracy on the OfficeQA Pro benchmark, establishing a new state of the art for complex enterprise document tasks.
The model demonstrated a 46% reduction in errors on OfficeQA Pro compared to its predecessor, GPT-5.4.
Significant improvements were observed in parsing older documents and scanned PDFs, which were challenging for earlier models like GPT-5.4.
GPT-5.5 also enhanced orchestration across multi-step tasks, avoiding inefficient search detours seen in previous versions and leading to more reliable workflow completion.
Databricks is making GPT-5.5 available for customer agent workflows through its AI Unity Gateway, leveraging AgentBricks and the Agent Supervisor API.
This progression demonstrates how advanced LLMs are enhancing the reliability and capability of agentic systems for complex enterprise tasks, pushing the frontier of practical AI deployment and making agents more robust for real-world applications.
19 May 2026impact 60
Integration of Street View Imagery
details: {'availability': 'Currently available for U.S. locations with plans for global expansion.', 'user_interaction': 'Users can explore and creatively reimagine real-world locations.'}; description: Project Genie enhances virtual environment simulation by integrating real-world imagery from Google Street View.
21 May 2026impact 65
AdventHealth achieves 80% administrative task reduction with enterprise-wide ChatGPT for Healthcare deployment
AdventHealth successfully deployed ChatGPT for Healthcare across its system, significantly reducing administrative burden and improving clinician workflows.
AdventHealth achieved an 80% reduction in time spent on administrative tasks by deploying ChatGPT for Healthcare.
The organization treated AI adoption as a 'product' and framed AI as a tool to return time to clinicians, tracking usage via 'messages per user per business day' as a key performance indicator.
ChatGPT for Healthcare assists physician advisors in generating structured patient chart summaries and drafting rationales, ensuring data protection and compliance in regulated environments.
This case study demonstrates a successful model for enterprise-wide agentic AI adoption, highlighting the importance of strategic framing, measurable outcomes, and compliance in achieving significant operational and human impact.
1 Jun 2026impact 60
Enhanced AI Adoption through Agent Logic
agent logic performance: {'use_cases': ['legacy code understanding', 'test generation', 'incident response', 'compliance automation'], 'token_consumption': '30× lower'}
16 Jun 2026impact 60
AI Planning Prototype Development
description: The UK government is developing an AI planning prototype with Gemini to reduce processing times for homeowner applications.; expected outcome: 50% reduction in application decision times.
29 Jun 2026impact 60
Increased Confidence in Agentic AI
expert opinions: agents help streamline processes and improve performance; key application area: data workflows; confidence in agentic ai: surging for measurable tasks
29 Jun 2026impact 60
Impact of AI Framing on Human Performance
framing issue: Framing AI agents as coworkers negatively impacts human performance and accountability.; study reference: Wiles's study involving 1,261 managers.
1 Jul 2026impact 40
Review of Google Home Speaker and Gemini for Home
price: 99.99; features: {'microphones': 3, 'thread_border_routing': True, 'neural_processing_unit': True}; criticisms: {'ai_limitations': True, 'gemini_performance': 'unfinished'}
1 Jul 2026impact 60
Enhanced Real-Time Voice AI Integration
collaboration: Hugging Face and Cerebras; implemented in: over 9,000 Reachy Mini robots
2 Jul 2026impact 65
Orbital Reconnaissance Agent
True Anomaly's Jackal spacecraft successfully performed an autonomous orbital rendezvous and inspection mission for the U.S. Space Force, demonstrating advanced agentic capabilities in space.
The Victus Haze exercise involved Rocket Lab launching a Puma spacecraft on short notice (16 hours 42 minutes), which was then intercepted, identified, and imaged by True Anomaly's Jackal spacecraft.
The Jackal spacecraft used onboard sensors to find and identify the Puma from 2,000 kilometers away, then flew close, orbited it, and captured imagery before returning to its original orbit.
This mission demonstrated the close inspection of a space vehicle soon after launch, a capability deemed necessary for monitoring novel space weapons.
This marks a significant step in deploying autonomous agentic AI systems for critical physical operations in space, moving beyond software-only applications.
2 Jul 2026impact 60
Woodside Energy Advances AI Integration
goal: enhance decision-making and operational efficiency; focus: AI integration in industrial operations; tools: Startup Advisor; company: Woodside Energy
3 Jul 2026impact 65
AI-powered browsers integrate agentic capabilities
The browser market is shifting to integrate AI assistants that act on behalf of users, leading to a new generation of AI-powered browsers.
The browser wars have entered a new phase, focusing on which company's AI gets to act on the user's behalf inside the browser itself, moving beyond just search results.
New AI-powered browsers like Perplexity's Comet offer chatbot-based search, email summarization, and task automation.
OpenAI's Atlas integrates ChatGPT for search and allows browsing within the chatbot, while The Browser Company's Dia also features an AI chat tool.
Upcoming platforms like Aside are designed as AI-first, browser-native automation platforms for autonomous task completion, form filling, and data management.
Other browsers like Jatter and DuckDuckGo are also enhancing their offerings with AI features, including chatbots and personalized recommendations.
This trend signifies a major expansion of agentic AI capabilities into a ubiquitous user interface, transforming how users interact with the web and potentially challenging established browser paradigms.
6 Jul 2026impact 60
Vercel Enhances AI Agent Functionality
deployments: 6000000; tokens flow: 1000000000000; coding agents triggered: 3000000
7 Jul 2026impact 50
Microsoft's Deployment of In-House AI Models
new models: 7; model types: coder, text-to-image generator
7 Jul 2026impact 60
Integration of Hugging Face Models into Microsoft Foundry
managed compute: Managed GPU platform-as-a-service for model deployment; enterprise support: Operational support for enterprise use of Hugging Face models
7 Jul 2026impact 60
MUFG's AI Integration Initiative
deployment: {'employees': 35000, 'custom_gpts_created': 1800, 'training_participation': 100}; organization: MUFG
7 Jul 2026impact 60
Deployment of Autonomous Vehicles in Ukraine
miles driven: 2500; deployment count: 100; missions completed: 1100; casualties evacuated: 52
7 Jul 2026impact 60
Hugging Face and Amazon SageMaker Integration
integration features: one-click deployment, automatic domain provisioning, pre-configured permissions; workflow improvements: eliminates manual IAM role configuration, simplifies access for model customization
9 Jul 2026impact 60
Launch of ChatGPT Work
user adoption: More than 5 million people use Codex every week.
9 Jul 2026impact 60
Lyzr Raises $100 Million Using AI Agent
startup name: Lyzr; funding round: Series B; funding amount: 100000000
9 Jul 2026impact 50
Integration of Atlas Features into ChatGPT
new feature: integrated browsing in ChatGPT; discontinued feature: Atlas AI browser
Lineage
Descended from
Led to
This is one node. The map holds the whole field.
Watch Agentic AI — and everything it connects to — grow as real news threads onto the map every day.