Finding onboarding problems before they grow

We were good at hiring people and inconsistent at landing them.

Some managers didn’t meet their new hire until well into week one. Some hires were dropped onto complex client projects with no ramp. And nobody had a way to see trouble coming. By the time a struggling hire surfaced, it was a performance conversation, not an onboarding one.

Across regions, handoffs were not always clear. At times we reached a start date without the manager, project, or first-week expectations lined up. I wanted a way to catch those gaps sooner. With a very small talent and culture team, “just check in more” wasn’t a plan.

Two conversations at every step

I built a lightweight operating system on tools we already had. Every new hire gets a structured check-in at Day 1, Week 1, Day 30, Day 60, and the end of probation. Every check-in happens twice: once with the hire, once with their manager.

The two stories don’t always match, and when they don’t, that gap is where the useful information lives.

Each check-in produces two scores from 1 to 5, Performance and Engagement, which roll up to a single health status. Simple enough to score in a minute, specific enough to act on.

How it works

1. An offer is marked accepted, and the hire lands in the health tracker with their manager, start date, and project.
2. A missing manager or project flags red immediately, so gaps surface before the start date instead of on it.
3. Slack nudges go to the hiring manager from acceptance through month three, so prep happens before day one instead of after it.
4. At each milestone, the regional owner runs both check-ins, guided by a Claude Project holding the question guides, scoring rubric, and message templates in English and Spanish.
5. A Check-in Companion handles one hire at a time: suggested questions, notes, scores, and a recommended next action.
6. An Org Health Report pulls the whole cohort together: who’s At Risk, which managers have more than one hire on Watch, patterns by region, and what to do this week.

I run the Canada check-ins myself. Our Mexico culture lead runs Mexico and Latin America and reports back, so the system works the same way in both places without me in every conversation.

The details that mattered

The last check-in is Day 83, not Day 90. Mexico has a 90-day probation period. A check-in on Day 90 tells you what happened. A check-in on Day 83 still lets you change it. I made it universal so every region runs on the same calendar.

The dashboard got smaller, not bigger. My first version tried to hold everything. Real check-ins are messy and contextual, and a form couldn’t capture them. The fix was letting the conversation happen in Claude, where I can think out loud, and keeping the dashboard to scores and status. The tool got better when it did less.

What changed

New hires in Canada and Mexico now move through the same rhythm, from offer to the end of probation. Managers get prompted before day one instead of reminded after it. And a hire who is struggling can show up in a weekly list, while there’s still time to help, instead of in an exit interview.

We now have a consistent check-in rhythm and earlier visibility into how new hires are doing, so the People team can follow up while there’s still time to help.

What I’d do differently

I’d start with the rubric and the questions, not the dashboard. The interface was the fun part and the least important part. What matters is a manager hearing, on Day 30, that their hire feels lost in a way they hadn’t noticed.

The reminder isn’t the important part. The conversation is.