16 companies · 27 sourced backers · 97 milestones on record
This frontier centres on AI agents designed to carry out multi-step tasks rather than only answer questions. Companies are working to make these agents useful in software development, customer service, legal work and everyday business operations, while building the computing, identity and payment systems needed to support them.
Where the 16 companies stand today, from their profiles and 10-year milestones. Tap a phase to see who is there.
- ▸Reliably completing multi-step tasks
- ▸Working within existing business systems, data and policies
- ▸Managing computing costs and energy needs
- ▸Verifying agent identities and making payments accountable
Publicly listed: 3 companies (owned by public shareholders).
- ›Every sourced story on all frontiers, filterable by company and news type
- ›10-year milestone records for each company: demo vs shipped
- ›Explained: plain-words readiness diagrams for real breakthroughs
- ›Signal Trends of 39 frontiers covering 378 companies over time.
- ›The events calendar: launches, approvals and deadlines ahead
- /dev/agents
The company's stated aim is a dedicated operating system that makes personal AI agents as standard and reliable as mobile apps became on smartphones.
Trials · backed by DoNotPay, Stripe
- Adaption (Adaption Labs)
To make AI systems that learn and adapt from real-world use rather than depending entirely on repeated retraining.
Lab · backed by Emergence Capital
- Cognition AI
To make AI capable of carrying out substantial software-engineering work rather than only suggesting snippets of code.
Everyone · backed by Founders Fund, Khosla Ventures
- Cursor (Anysphere)
To make AI an active collaborator in software development through an agentic code editor.
Everyone · backed by Andreessen Horowitz, Thrive Capital
- Decagon
To make AI agents capable of resolving customer-support tasks rather than simply routing questions.
Some users · backed by Accel
- Glean
To make company knowledge usable by AI assistants and agents across everyday work.
Some users · backed by Wellington Management
- Harvey
To make AI agents useful for substantive legal work.
Everyone · backed by Sequoia, OpenAI
- Imbue
To create AI agents that reason through tasks and carry out software work autonomously.
Trials · backed by Astera Institute, Khosla Ventures
- Microsoft (Copilot Agents)
To make AI agents useful workers across Microsoft's business software.
Everyone · publicly listed
- OpenAI (Operator/Agents)
To make AI agents capable of using computers to complete multi-step tasks.
Everyone
- Paid
To make AI-agent services commercially manageable through outcome-based billing and cost tracking.
Some users · backed by Founders Fund, CRV
- Salesforce (Agentforce)
To make autonomous AI agents useful across customer and enterprise workflows.
Some users · publicly listed
- ServiceNow
To make AI agents useful within enterprise workflow automation.
Everyone · publicly listed
- Sierra
To make customer-facing AI agents useful representatives of a business.
Some users
- Skyfire
To give autonomous AI agents verifiable identities and a practical way to make payments.
Trials · backed by Mucker Capital
- VESSL AI
To make large-scale AI computing resources easier to obtain and operate.
Some users · backed by VESSL AI
Keep following breakthroughs like these. Every Monday, get up to 5 major breakthroughs and launches explained in plain English.
Free. No account needed. One email a week. Unsubscribe anytime.