Client Profile

 

This scenario reflects a regional property and casualty insurer with about 1,500 employees across several states. The finance organization of about 150 people spans FP&A, accounting and the monthly close, treasury, and regulatory reporting. It operates under SOX controls, external audit, and state insurance department examinations.

 

Nine months earlier, the company licensed Microsoft 365 Copilot for its corporate functions. IT ran the deployment well: licenses went out on schedule, training was available on demand, and an acceptable-use policy was published on the intranet. By every technology measure, the rollout was done.

Starting State: Deployed, Not Adopted

 

Usage data told a different story. A small group in FP&A used Copilot daily to draft variance commentary and summarize long documents. Across accounting, treasury, and regulatory reporting, usage was close to zero. Training completion was high.

 

Underneath the numbers, the patterns were familiar:

• Finance staff were afraid of getting it wrong. In a function where a wrong number can become a restatement or an audit finding, “use AI” sounded like an invitation to take on personal risk.

• The policy read as a prohibition. The acceptable-use policy said not to enter confidential data. Most of finance reasonably concluded that meant everything they worked on.

• Managers were handed a mandate without support. They received the same training module as their teams, plus a new expectation to “drive adoption,” with no time or guidance attached.

• Some use went underground. A few people were using personal AI accounts for low-risk drafting, out of view, because it felt safer than asking.

• Nobody owned the people side. IT owned the tool, finance owned the work, and the gap between them belonged to no one.

Inflection Point: The ROI Question Arrives

 

Two things brought it to a head. The CFO asked IT what the Copilot licenses were returning, and IT had no answer beyond seat counts. Then IT began planning the rollout to legal and compliance, and the IT leader didn’t want to repeat the same pattern in a function with even more reason to hesitate.

 

The IT leader’s diagnosis was accurate: this wasn’t a technology problem, and IT couldn’t fix it alone. Culture and manager behavior in finance sat outside IT’s authority. The IT leader needed the finance leadership team to own the change with them.

The Approach: Three Fixed-Scope Steps

 

Step 1: Manager Adoption Diagnostic (three weeks)

The work started by finding out where adoption was actually breaking down. Short pulse surveys went to finance’s roughly 20 managers and their teams. Confidential interviews were held with a sample of managers from the highest- and lowest-usage teams, and the existing Copilot usage data was reviewed alongside.

 

The pattern was clear. Team usage tracked manager usage almost exactly. The managers whose teams had opted out weren’t resistant to AI. They didn’t know where the lines were relative to SOX controls, had no time to learn, and didn’t want to be the person who asked a question that suggested they didn’t understand the policy. The findings report gave the sponsor a team-by-team picture, the two or three blockers that mattered most, and a 90-day plan.

 

Step 2: Leadership Alignment Workshop (half day)

The CFO’s staff, the IT leader, and the head of internal audit spent a half day working through what the diagnostic surfaced, using the ACT approach (Align, Commit, Track):

• Align: The group agreed on a shared narrative they could all repeat: AI drafts, finance owns the number. They also agreed on guardrails sorted by type of task: where AI use was encouraged, where it required documented review, and where it was off limits. Internal audit helped shape the guardrails, so they reflected control requirements rather than general caution.

• Commit: Each finance leader named two or three specific, visible actions to take personally within 60 days. Examples included using Copilot on their own monthly reporting and describing that use in staff meetings.

• Track: Commitments were scheduled for review at the monthly close retrospective.

 

Step 3: Change Activation (three-month retainer)

The work then moved to the people expected to change:

• Managers got their own enablement before their teams: the narrative, talking points for the questions they’d hear, and hands-on time with their own work.

• Use cases came from finance’s real pain points in the close cycle, including variance and flux commentary, responses to audit document requests, policy and contract lookups, and first drafts of board reporting narrative.

• A short monthly show-and-tell, held after the close, let people in similar roles demo what worked and what didn’t.

• The acceptable-use policy was rewritten in plain language for finance, with examples, and the sanctioned path was made easier than the workarounds.

• Baseline metrics from the diagnostic were set up for re-measurement at 90 days, from enablement through adoption, usage, and time reinvested.

Outcomes: What Changed

 

Within the first quarter, the shifts were visible:

• Managers became the multiplier. Managers began using Copilot on their own work and talking about it openly. Usage spread beyond FP&A into accounting and regulatory reporting.

• Fear turned into clear rules. Staff could say what was encouraged, what needed review, and what was off limits, and they knew internal audit stood behind those answers.

• Shadow use moved into the sanctioned tool. Once the rules were explicit, the reason to work around them went away.

• The CFO got a money trail. The function started tracking what time saved on commentary and reconciliations was reinvested into, which produced a credible answer to the ROI question in place of seat counts.

• IT got a repeatable model. The IT leader took the same sequence to legal and compliance with the finance leadership team as a reference.

What Made the Difference

 

Across situations like this one, a few patterns stand out:

• The champion and the sponsor were different people. The work succeeded because IT brought it in and finance leadership owned it.

• Managers were treated as their own audience. The fastest movement came from the layer that decides whether everyone else adopts.

• Guardrails came from the people closest to the risk. Involving internal audit turned “be careful” into rules people could act on.

• Usage was measured, not training completion. Completion had been high all along, and it explained nothing.

How I Use This Pattern in My Work

 

Because this is a composite, it doesn’t describe a single client. It reflects what I see most often in traditional and regulated companies: the tools are in place, the training is done, and adoption in the business functions has quietly stalled.

 

I usually work with the IT or AI leader who owns the rollout, and with the function leaders they need on board. Engagements start with a fixed-scope entry point (a Manager Adoption Diagnostic, an AI Readiness Assessment, or a Leadership Alignment Workshop) and continue into change activation when the findings call for it.

 

If you own an AI rollout and recognize your organization in this story, let’s talk.

 

Book a 30-minute conversation →

Return to case studies page →