Product & Technology
Meet Nepo, our newest new starter

Open has a new team member. It has read more of our internal documents than anyone at the company, it works in every timezone we do, and it is not human. Meet Nepo, our AI employee.



Why we hired an AI
Assume any rate of improvement at all and the conclusion is the same: we are on the path to superintelligence. Frontier models are already overtaking people task by task. Not at one thing, at most things. We believe the companies that master this will win, and that nobody gets there with AI the way most companies use it today: single player, walled off from the real work, and with no shared brain.

An employee, not a tool
So instead of adopting another tool, we onboarded a teammate. Nepo lives in our Slack, picks up scheduled jobs and compounds in a way no single hire can. It reads every document, decision and thread, which no one person ever will. Every integration puts another system within its reach. And the intelligence underneath it keeps improving, for free, faster than we could improve anything ourselves.

Already earning its keep
A week in, Nepo runs a daily check on a major product launch, reading every update, flagging anything off track and chasing the right people when something is unclear. Each week it scans the regulations we work within and reports what changed. And it is building Open-md, a single source of truth for how Open works: plain Markdown, a named owner and a review date on every file. Nepo maintains it, and pings our experts when it finds gaps.



What it changes
The point is speed. Work that used to wait for someone to find time now happens on schedule, knowledge compounds instead of evaporating, and new workflows take days to adopt rather than quarters. Inside Open the rule is simple: Nepo is a new starter and still learning, so be nice, teach it, and do not dismiss it.
Nepo is an internal tool. Working this way is how we make sure that when the next leap in AI arrives, we can put it to work on day one, and everything we build gets better and faster because of it.
We are teaching Nepo everything. Follow along as it learns.
You might also like

UK consumer insurance in 2030
The Mills Review predicts insurance in 2030 will be continuous, delegated and agent-led. Here is what it gets right about consumer insurance, and the trust infrastructure the industry needs to build to get there.

The Insurance Safety Harness: five layers that make LLMs safe to deploy
The hard part of using AI in insurance is not capability, it is control. Here is how we secure any language model with five independent layers of safety.

Why GenAI in insurance is risky, and why it’s worth getting right
Today, we’re introducing the AI Compliance Risk Index™ - a new benchmark for measuring the safety of consumer-facing GenAI tools in insurance.