Solutions / Onboarding assistant

Employee onboarding assistant (RAG)

A grounded assistant answers new hires from your own policies, runbooks and process docs, with the source attached — so the same twenty questions stop landing on your team.

HR Internal RAG
Answers with a source
linked to the actual document
Available immediately
including the first week
Gaps made visible
where docs do not answer
Who it is for

Companies hiring steadily enough that senior people spend their first weeks answering the same questions instead of working.

Short answer

New hires ask in plain language and get an answer built from your handbook, policies, runbooks and internal wiki, with a link to the paragraph it came from. Questions it cannot answer from your documents are logged rather than guessed at, which turns onboarding chatter into a prioritized list of the documentation you are actually missing.

The problem

Every new hire asks the same twenty questions, and each one costs a senior person an afternoon.

01

Onboarding load falls on your best people

The person who knows the answer is the person you least want interrupted. Multiply by every hire and the cost is real, but it never shows up in a budget line.

02

The wiki has the answer, and nobody can find it

Documentation exists across four tools with three naming conventions. Search returns forty pages, so new hires ask a human instead — which is faster for them and expensive for you.

03

Nobody learns what the docs are missing

Questions answered in a DM leave no trace. The same gap is filled by hand for every hire, forever, because nothing ever records that it is a gap.

How it works

From trigger to result, step by step.

01

Collect what a new hire actually needs

Handbook, policies, tooling runbooks, security rules, who-does-what. We start with the documents that answer real first-month questions, not the full archive.

02

Ground every answer, refuse the rest

Answers are built only from your documents and carry a link to the source. If the documents do not cover it, the assistant says so and routes to a person, because a confident wrong answer about policy is worse than silence.

03

Respect who may see what

Permissions follow your existing groups, so the assistant never surfaces a document to someone who could not open it directly. This is designed in first, not added later.

04

Put it where people already ask

Slack or Teams, so asking the assistant is easier than asking a colleague. An internal tool on a separate URL is one click too far and simply goes unused.

05

Turn unanswered questions into a doc backlog

Every unanswered question is logged and clustered. Your team writes ten missing pages instead of guessing, and coverage improves with each cohort.

Before / after

What changes on the ground.

Today, by hand
×Senior people answering the same questions
×Documentation spread across four tools
×New hires blocked waiting for a reply
×No record of what the docs are missing
With the automation running
›Answers in seconds with the source attached
›One place to ask, inside Slack or Teams
›Out-of-scope questions routed to a person
›A ranked list of documentation gaps
What you get

Delivered, not demoed.

›An indexed knowledge base built from your onboarding documents
›A grounded assistant that cites sources and refuses to guess
›Permission-aware retrieval matching your existing access groups
›A Slack or Teams entry point where questions already happen
›Logged gaps clustered into a documentation backlog with volumes
Built with

We build in your stack rather than moving you onto ours. The list below is what this solution most often connects to — other systems are a scoping question, not a blocker.

Claude OpenAI LangChain pgvector PostgreSQL Slack Notion Confluence
Time to production4–6 weeks
Build priceFixed price
First stepFree mini-audit
Honest limits

When this is not the right solution.

·If your documentation is thin or contradictory, the assistant will surface that rather than fix it. Sometimes the right first project is writing ten pages, and we will say so.
·With a handful of hires a year, a good checklist and a buddy beat any system, and cost less.
·If policy answers carry legal weight in your jurisdiction, the assistant should point to the document and to a person, not stand in for HR.

Questions we get about this one

Answers carry the source and its date, so a stale answer is visible instead of invisible. Contradictions between documents are flagged during indexing, which is often the most useful output of the first week.

Bring us the process that hurts.

The mini-audit is free: we take your version of this process apart and tell you plainly whether automating it pays. If it does, you get a scope and a fixed price.