The agent buildout timeline
102 agents plotted by public launch date, grouped by macro category. Click any row for the agent detail.
Every catalogued agent placed on the timeline by its public launch date
The ramp is visible in the plot and it starts in 2023: 34 agents launched that year against 8 the year before, with 27 more in 2024. Everything earlier is a thin tail running back to 2011. Plenty of capability existed before that point; what changed was that it became reliable enough to ship.
Show more
Which sectors moved first says something about them. Finance is the earliest entry in the catalogue by a wide margin, in 2011, with SaaS following in 2016 and health in 2017, all well ahead of the general wave. For finance and health that fits industries already running automated decisioning, with the compliance scaffolding to hang it on. Developer, security and embedded agents only appear from 2023, once the underlying capability caught up. The open question is what the next buildout looks like: whether finance, health and legal compound into many specialised agents, or consolidate into a few the market trusts, with the governance and enterprise integration that would imply.
When did AI agents start appearing?
Coding assistants and frontier chat products came first. The wider wave followed from 2025, as agents spread into customer support, security, legal, health and finance. The launch rate has climbed steeply since.
What was the first AI agent?
It depends where you draw the line. Rule-based bots and automation scripts long predate large language models. What changed was models being able to plan a sequence of steps and call tools, which turned assistants into something that acts.
Which industries got AI agents first?
Software development and customer support led. Both involve high-volume text work where the output can be checked quickly and a mistake can be undone. Regulated fields such as health, law and finance followed later and much more cautiously.
What is the difference between a preview and general availability?
A preview or beta is public but unfinished: access may be limited, behaviour changes, and there is no service guarantee. General availability means the vendor supports it for production use and will hold it stable.
How many AI agents are there?
Hundreds of named commercial products, and the number keeps climbing. Any count is a snapshot. New agents launch every week, and others get absorbed into larger platforms as features rather than staying separate products.