Hiring outruns documentation.
The handbook describes a company a third the size, and the people who could update it are the ones shipping the product.
Because of AI, more and more companies are growing headcount faster than they ever have. Roles that never existed before have to be filled — and filled quickly.
Vanav reads a company’s real Slack, docs and tickets, works out what a brand-new role actually is, and drives the new hire through their first two days of real work.
The handbook describes a company a third the size, and the people who could update it are the ones shipping the product.
“Legal Engineer” was invented a week before the req opened. Nobody has written it down, because nobody has done it yet.
Everyone senior enough to onboard this hire is the reason the company is growing.
Every company in this position tries both, in this order, and then quietly gives up and lets new hires figure it out.
A handbook is stale the day it merges. At this pace it is stale before it merges, and the roles it would need to describe are invented faster than anyone can write them down.
The only people who know how the role works are the bottleneck that made the role necessary. Every hour they spend onboarding is an hour of the thing you hired them to do.
So the onboarding has to derive the role itself — from the evidence the company has already produced — run the first two days of real work, and interrupt a human only when it truly cannot proceed.
They open one link. Their name, their role, and — worked out from what the team has actually been writing to each other — what the job is here. Nobody had to write it down first.
Real work, not a reading list. A live task with a real deadline: why it matters, what finished looks like, and the context to do it without interrupting anyone.
Asking a colleague takes nerve. Asking this doesn't — so they ask the fortieth question as easily as the first, and get answers out of the company's own messages. When nobody has written the answer down, it says so instead of inventing one.
If they are about to do something the team already decided against, it says so unprompted — and shows them the message where it was decided, by name and date.
A manager normally catches a confident wrong turn by reviewing the work. That holds at six reports. It does not hold at sixty, and the correction arrives from a customer instead.
“I’m going to fix the Italian miss by adding cessione del contratto and a few other synonyms to the playbook keyword list.”
Marta wrote this in #customer-escalations on 19 Aug:
“NOT shipping: a keyword list. If anyone adds ‘cessione’ to a keyword list I will find you.”
They didn’t ask a question. They stated a plan, and the agent volunteered the decision that already ruled it out — with the date and the person.
When the answer genuinely isn’t written down, the agent asks the one person who would know — once — and writes the confirmed answer back into the company’s own record, attributed and dated. Everyone who hits the same gap afterwards gets it from there.
Your senior people are interrupted once per unknown, not once per hire. With three people starting the same month that is one question instead of three; at thirty it is the difference between a tool and a second job.
The manager screen shows blockers and nothing else. A surveillance dashboard gets killed by the culture it is sold into.
What the company hasn't decided is shown as undecided. Confident filler is the failure a hiring manager catches in ten seconds.
Every quote is checked as a literal substring of the message it cites, and the ones that are not there are dropped rather than shown. The reading is the agent's; the evidence under it is the company's own, and you can open it and check.
We’re rolling out to a limited number of fast-scaling companies first. Book a 20-minute call and we’ll add your company to the list.