Onboard a Team in Weeks With an AI Shadow Engineer

Every engineering leader has run this play: a new product to adopt, a migration to execute, a platform the whole team must learn. The standard toolkit is a vendor workshop, a wiki page, a course-library license, and hope. Six months later, three engineers actually understand the new system, everyone routes questions through them, and the migration is late.

The failure isn't effort. It's architecture. Team onboarding is almost always run as a content problem — 'get the information in front of them' — when it's actually a training problem: retention, practice, feedback, and verification, at team scale.

We built miatz's engine for individuals. Companies kept asking for the same machinery pointed at their stack. So that's exactly what corporate team-grooming is: the same loop, the same shadow engineer, the same telemetry — aimed at your product, your migration, your topic.

The playbook

Step one: the topic becomes a program. Whatever the team must learn — an internal platform, a Kafka migration, a new compliance regime, your own product for new hires — gets decomposed into a gated curriculum: spaced-repetition decks for the facts that must be instant, daily hands-on reps against realistic environments, and incident drills for the failure modes that matter. Mastery gates replace the calendar: nobody is 'done' because the workshop ended; they're done when they demonstrate the competence.

Step two: Xook shadows every engineer. [Xook](/xook) — our AI shadow engineer — pairs with each team member inside the actual work, in the same propose-and-approve loop our learners use. It proposes steps, answers questions grounded in the program's corpus and each engineer's own history, and never acts unilaterally: every suggestion requires a human decision, and every decision is audit-logged. The effect is a patient senior engineer, per person, who never gets tired of the same question — which is precisely the resource your three overloaded experts can't be.

Step three: leaders read telemetry, not vibes. Every review, rep, decision, and 'this is still fuzzy' check-in flows into a dashboard. Not surveillance theater — training instrumentation, the same kind we run for [individual learners](/program).

Onboarding without telemetry isn't a program; it's a broadcast — and you find out who received it only when something breaks.

What the telemetry actually answers

The questions leaders ask us to answer are concrete:

• Readiness: who has demonstrably cleared the gates for the migration's critical path — and who is one gate away versus five?
• Bottlenecks: which concept is the whole team flagging as fuzzy? (That's a materials bug, not thirty people bugs — fix it once.)
• Risk: which decisions is the team getting wrong in drills, before they get them wrong in production?
• Progress: is the cohort trending toward the cutover date, on evidence rather than self-report?

Compare that to the traditional signal — a completion percentage on a video course — which correlates with almost nothing except whether the tab stayed open.

Why this beats the workshop

The learning science here is boring in the best way. Workshops fail for the same reason cramming fails: massed exposure, no retrieval, no spacing — the forgetting curve takes most of it within weeks. Distributed practice with retrieval and feedback is the most robust result in the field, and it's exactly what a daily loop delivers and a two-day offsite structurally cannot.

There's a second-order benefit teams notice within weeks: the expert-bottleneck unwinds. When Xook absorbs the long tail of 'how do I…' questions — with answers grounded in your program, not generic internet knowledge — your senior engineers stop being help desks and go back to being engineers. And because every question asked is logged telemetry, the questions themselves improve the curriculum for the next hire.

The audit log pays one more dividend: institutional memory. Six months later, 'why did we configure it this way?' has an answer with a name, a date, and a rationale — searchable, instead of buried in a departed contractor's head.

Where to start

The engine is topic-agnostic by design — the same machinery grooms a founding cohort on fundamentals and a platform team on your migration. What changes is the corpus and the gates.

If you have a migration with a date on it, a product with an onboarding problem, or a team that needs to be genuinely dangerous on a new stack, [talk to us](/for-companies). Bring the topic; we'll bring the loop, the shadow, and the dashboard.