What was FinEdge's starting position?

A 280-person fintech, Series B, People team of seven, two heavy ChatGPT users, three occasional, two avoiding it entirely. No shared workspace, no policy, no measurement. The CPO had approved budget and no plan. We see this exact split on most first calls: money moves faster than method.

What did the 90-day roadmap actually deliver?

Nine production workflows, a published one-page AI policy, a shared workspace with reusable context, a weekly demo cadence, and a People metrics pack tied to revenue and risk. Time saved across the team: roughly 38 hours per week, sustained, by day 90, and for the first time, a CFO who could say what the People AI spend was actually returning.

What were the near-misses?

Two, both caught within weeks because a weekly evaluation review already existed by month two. First, a comp-letter automation that would have leaked salary band logic in error states, caught in a critique pass before it reached production. Second, a candidate-screening agent that began encoding a hiring-manager preference unrelated to the role, caught in the weekly evaluation review and removed.

What would you do differently?

Two things. Build the evaluation habit on day one, not day thirty, we added it in week five and both near-misses happened in weeks six and seven. And resist demoing every new workflow company-wide too early, the wider the audience, the harder it is to retire something that isn't working.

Is this template repeatable?

The sequence is repeatable. The specific workflows are not. Every team's hot cells on the workflow heatmap are different. Use the sequence (diagnose, prove, harden, scale) and let the workflow choices come from your own diagnostic.

AI roadmap case study: FinEdge's first 90 days

This is one case study, lightly anonymised, told in enough detail that another team could run the same play. Names, numbers, and one sequencing detail have been changed. The shape of the work is exact.

Starting position

FinEdge is a 280-person fintech, Series B, payments infrastructure for mid-market merchants. The People team is seven: a CPO, two HRBPs, two TA partners, a People Ops lead, and a learning lead. The CEO had approved a six-figure People AI budget at the previous board meeting. The CPO had a budget and no plan.

Before we started: scattered individual ChatGPT use across the team, no shared workspace, no policy, no measurement. Two of the seven were heavy users, three were occasional, two were avoiding it. Classic uneven adoption.

Days 1 to 14: diagnose

We ran the People Ops diagnostic toolkit. Three sessions over two weeks. Output: a one-page read of the function with five sections.

The two findings that shaped everything that followed:

The team expected the hottest cells on the workflow heatmap to be in TA. The heatmap put them in onboarding (high frequency, high pain, high AI fit on the document and Q&A steps) and in HRBP request triage (medium frequency, very high pain, very high AI fit on the routing and first-response steps).
The AI readiness score was 6 out of 10 on tooling, 3 on data, 2 on governance. The team had been planning to spend month one on tools. The diagnostic redirected month one to data and governance, with a single tooling decision.

Days 15 to 60: prove value

Three workflows in six weeks, in this order, with a single champion on each:

Onboarding Q&A. A retrieval-augmented assistant grounded on the onboarding handbook plus the last 18 months of People Ops Slack answers. Replaced 40% of week-one new-hire questions to People Ops. Saved roughly 6 hours a week across the team. Live by day 28.
HRBP request triage. A workflow that classifies inbound HRBP requests, routes them, and drafts a first response for the partner to approve or rewrite. Cut median response time from 26 hours to 4. Live by day 42.
Interview-loop scheduling. Less novel, more leverage. Reduced TA partner scheduling time by half. Live by day 56.

A weekly Friday demo cadence started in week three and never stopped. Fifteen minutes. One champion shows one thing. This single ritual did more for adoption than every training session combined.

Days 61 to 90: harden and scale

Six more workflows, smaller and more specialised, owned by individual team members rather than dedicated champions. The pattern from the first six weeks was now legible enough that the rest of the team could replicate it.

The harder work in this phase sat outside the workflows, in three pieces of infrastructure:

A published one-page AI policy covering data, vendors, evaluation, and what is forbidden.
A shared workspace with reusable context and standardised prompts, following the workspace setup pattern.
A People metrics pack tied to revenue and risk, refreshed monthly, presented at the exec team's monthly business review.

By day 90: nine workflows in production, an estimated 38 hours per week of sustained team time freed, and a CFO who could answer the question what is the People AI spend returning? without flinching.

The two near-misses

Both happened in the second month. Both were caught by evaluation, not by luck.

The comp-letter automation drafted near-final language for promotion letters. In an error state, where the underlying salary-band data could not be retrieved, an early version fell back on a static template that hard-coded the band ranges. A critique pass caught it. We rebuilt the workflow to fail closed, with a human escalation, before it reached anyone outside the team.

The candidate-screening agent was helping summarise interview feedback. Within two weeks, the weekly evaluation review picked up that it was over-weighting one specific signal that mapped to a hiring manager's stated preference, but had no validated link to performance in the role. We pulled it. The lesson: evaluation is not a quarterly exercise.

What we would do differently

Two things.

Build the evaluation habit on day one, not day thirty. We added it in week five. Both near-misses happened in weeks six and seven. The lag was almost the cost.

Resist demoing every new workflow company-wide too early. Once the wider company has seen something, retiring it is political. Keep the audience small until the workflow has survived a month of weekly evaluation.

What this case is and is not

This is one team's 90 days. Copy the sequence: diagnose, then prove, then harden and scale. Don't copy the workflows, your hot cells will be different. Use the diagnostic toolkit to find yours, then run this same shape against them.

What this connects to

Auto-recommended next reads in the People Ops cluster, ranked by shared concepts and headings:

Common questions

What was FinEdge's starting position?: A 280-person fintech, Series B, People team of seven, two heavy ChatGPT users, three occasional, two avoiding it entirely. No shared workspace, no policy, no measurement. The CPO had approved budget and no plan. We see this exact split on most first calls: money moves faster than method.
What did the 90-day roadmap actually deliver?: Nine production workflows, a published one-page AI policy, a shared workspace with reusable context, a weekly demo cadence, and a People metrics pack tied to revenue and risk. Time saved across the team: roughly 38 hours per week, sustained, by day 90, and for the first time, a CFO who could say what the People AI spend was actually returning.
What were the near-misses?: Two, both caught within weeks because a weekly evaluation review already existed by month two. First, a comp-letter automation that would have leaked salary band logic in error states, caught in a critique pass before it reached production. Second, a candidate-screening agent that began encoding a hiring-manager preference unrelated to the role, caught in the weekly evaluation review and removed.
What would you do differently?: Two things. Build the evaluation habit on day one, not day thirty, we added it in week five and both near-misses happened in weeks six and seven. And resist demoing every new workflow company-wide too early, the wider the audience, the harder it is to retire something that isn't working.
Is this template repeatable?: The sequence is repeatable. The specific workflows are not. Every team's hot cells on the workflow heatmap are different. Use the sequence (diagnose, prove, harden, scale) and let the workflow choices come from your own diagnostic.

12 min

When reading turns into doing

The Grain Audit maps one People Ops process end to end, ranks the highest-return automations, and hands you a 90-day plan you keep whether or not we work together.

Two weeks. GBP 2,000, credited in full against a programme. Three slots a month.

Book a Grain Audit

If this resonated, there's more.

Subscribe to receive new Intelligence pieces as they're published. No noise, just the work.

By subscribing you agree to our Privacy Policy. Unsubscribe any time.