Onboarding new hires with an internal AI assistant

New hire onboarding generates the same handful of questions dozens of times over — a pattern almost perfectly suited to AI assistance, if it's scoped and built correctly.

Why onboarding questions are unusually well-suited to this

Unlike open-ended customer support, new hire questions during the first few months follow a highly predictable pattern: how do I set up my email, what's our PTO policy, who do I contact for X, where's the process documentation for Y. This predictability, combined with the fact that the answers genuinely exist in company documentation (HR policies, IT setup guides, org charts), makes it one of the cleanest use cases for a retrieval-grounded AI assistant — high repetition, well-documented answers, low ambiguity.

What this actually saves, concretely

The direct time savings — fewer interruptions to managers and HR for questions with a documented answer — is real but often secondary to a less obvious benefit: new hires get answers instantly rather than waiting for a colleague to have time to respond, which measurably affects how quickly someone becomes productive and how confident they feel in their first weeks. We track "time from question to answer" as a metric specifically because the delay in getting unblocked, not just the interruption cost to the person answering, is a real productivity factor for new hires.

What needs to exist before this works

The assistant is only as good as the underlying documentation. If your onboarding knowledge lives primarily in informal, undocumented tribal knowledge — things every existing employee just "knows" from being told once — there's nothing for the assistant to retrieve from. We typically find the process of building an onboarding assistant surfaces gaps in documentation that needed fixing anyway, independent of the AI tooling — a valuable side effect, but one that adds real time to the project if documentation gaps are extensive.

A concrete example

A company growing quickly (adding roughly 8-10 new hires a month) had onboarding questions consistently overwhelming their small HR team, with new hires reporting frustration waiting for responses to what were often simple, previously-answered questions. Our discovery phase found their actual documentation was reasonably comprehensive but scattered across an HR system, an IT wiki, and several onboarding-specific Google Docs, with no unified way for a new hire to search across all of it. We built a knowledge assistant indexing all three sources, with a scoped focus specifically on onboarding-relevant content, and a clear escalation path to actual HR staff for anything outside its documented scope (compensation questions, anything requiring individual judgment).

Post-launch, HR reported a substantial reduction in repetitive Slack/email questions from new hires during the first 90 days, and a new-hire satisfaction survey specifically flagged "getting my questions answered quickly" as notably improved compared to the prior cohort's feedback.

How Ndakum approaches it

Onboarding assistants are one of the highest-confidence use cases in our AI Knowledge Assistant work — high repetition, well-documented answers, and a clear boundary for when to escalate to a human.