AI Agents

Definition

An AI agent is a system that uses a model to plan and execute multi-step tasks by calling tools, reading and writing data, and taking actions in software environments, typically with a human setting goals and reviewing outcomes.

How it works

1. 01A goal is specified in natural language, often with constraints and examples.
2. 02A reasoning model decomposes the goal into steps and decides which tools to call.
3. 03Tools expose actions: search, code execution, database queries, API calls, browser control.
4. 04The agent observes results, updates its plan and loops until the task is complete or blocked.
5. 05Guardrails, permissions and review checkpoints determine what it may do without a human.

Latest developments

• 2026-09Enterprise deployments increasingly connect agents to systems of record with scoped permissions.
• 2026-06Agent-to-agent and tool protocols are consolidating around a small number of open standards.
• 2026-03Coding agents are widely used for autonomous multi-file changes with test-based verification.

Use cases

• Software engineering and code review
• Customer support resolution
• Financial operations and reconciliation
• Sales and revenue operations
• IT and security operations
• Research and reporting

Market

Agent platforms are sold by frontier labs, cloud providers, systems-of-record vendors and specialised startups. Pricing is shifting from seats toward usage and outcomes.

Risks

• Prompt injection and data exfiltration through tool access
• Over-permissioned agents acting on bad inputs
• Silent failures in long-running tasks
• Accountability gaps when outcomes are harmful
• Vendor lock-in through proprietary tool ecosystems

Future outlook

Agents will become the default way enterprises consume AI. The winners will be those who own the system of record, the identity layer or the best verified reliability. Expect the interface of work to become a review queue.

Sources & references

1. 01Research on tool-using language models and agent evaluation — Various (see Sources page)researchCompiled bibliography maintained by the research desk.
2. 02Research on tool-using language models and agent evaluation — Various (see Sources page)researchCompiled bibliography maintained by the research desk.