Chapter 04 · Phase 1 of 8

Assess — Sentiment, Baseline, Stack Audit

Before you deploy anything, find out where your team actually is. Most AI rollouts fail because they skip the assessment phase and dump a tool on a team that wasn't ready for it.

1 · Assess 2 · Charter AI COEE 3 · Buy-in 4 · Pilot 5 · Deploy 6 · Train 7 · Govern 8 · Scale

Kristina Tsys leads AI implementation across a 15-person talent acquisition team at Fortive, hiring for 10 different companies. She had every reason to move fast. Instead, she stopped and surveyed her team.

This has actually turned out to be the whole change management process — because some people are familiar with AI, they use AI for a number of years. But some people they just start exploring, and it can be some resistance.
Kristina Tsys · Technical & Engineering Recruiter, Fortive Listen →

What she found was wildly uneven adoption. Some recruiters used AI daily. Some had only used ChatGPT for personal tasks. A few hadn't touched it. That data — not a vendor demo — became the starting point for everything that followed.

What Phase 1 produces

By the end of Phase 1, the AI COEE should have three artifacts in hand. Each one is a deliverable, not a deck.

1. The AI Sentiment & Usage Survey

Anonymous, org-wide, takes 5 minutes to complete. Run it before any tool decision. The survey has to capture four things:

Map results against McKinsey's four archetypes — Zoomers, Bloomers, Gloomers, Doomers — and you have your adoption strategy by population segment.

2. The Workflow Map

Matt Neylon put this principle in a single sentence: "You can't automate what you can't write down." Before you evaluate any tool, map the actual workflow you're trying to improve. Sticky notes. Whiteboard. Whatever. But document the steps, the handoffs, the decisions, and the friction points.

Document what's burning out your team — then automate that. Before you automate anything, understand the process well enough to know if it's worth automating. Automating a broken process just breaks things faster.
Matthew Neylon · HR Leader, The Mount Vernon School Listen →

Recruiting workflows worth mapping for AI candidates in Phase 1:

For each workflow, document: who owns each step, how long it takes, where the friction is, and what's already partially automated. The friction points are your candidate AI use cases.

3. The Stack Audit

Most TA orgs already have AI capabilities buried in tools they're underutilizing. Before you buy anything new, audit what you have. Most ATSes shipped AI features in the last 18 months. Most CRMs have AI scoring. Most calendar tools have AI scheduling. The AI COEE's first procurement decision is often: turn on what's already paid for.

Inventory in two columns: AI capability available in current stack, and actual usage rate today. Anything in the first column that's at zero or near-zero usage is your fastest win.

Pro tip

Capture the baseline metrics before anything else

You cannot prove ROI without a starting point. Lock in your time-to-fill, time-to-slate, cost-per-hire, recruiter-hours-per-hire, and offer acceptance rate at the start of Phase 1. If you start measuring after a tool is deployed, you've lost the comparison.

The two-week Phase 1 sprint

This phase shouldn't take more than two weeks. The AI COEE moves quickly here because the entire program depends on the data this phase produces.

Common Phase 1 mistakes

Do

  • Survey anonymously — you want honest data, not safe data
  • Include hiring managers, not just recruiters
  • Capture baseline metrics in the same document as the sentiment results
  • Share the survey results with the org before the next phase begins

Don't

  • Skip the survey because "we already know what people think"
  • Audit only one tool category — audit everything
  • Treat Phase 1 as a milestone you complete and forget — re-run the survey every 6 months
  • Let Phase 1 stretch past 3 weeks — momentum matters

Once you have the data, the AI COEE has its mandate. Phase 2 is where you formalize the team and its operating model.