The Copilot Business Case That Wouldn't Close
A 1,200-person professional services firm is evaluating a generative AI writing assistant for its consultants. The vendor's reference customer reports '30% productivity gains.' The firm's own 4-week pilot with 80 consultants shows a 9% reduction in time spent on first drafts, but also a 12% increase in senior-review time because junior consultants submit drafts that are fluent but miss firm-specific nuance, requiring more rework. Seat cost is $40/user/month for all 1,200 people. Leadership wants a 12-month payback. Utilization data shows only 35% of pilot users opened the tool in a given week after week 3. The firm bills clients hourly, so 'time saved' only converts to value if it is either billed or redeployed to new work - and utilization is currently near capacity.
1. Build the honest value case. Which numbers in the setup are real value, which are illusory, and why does the 30% reference figure not transfer?
Read the worked answer
Real vs. illusory value:
- The 30% reference figure is not transferable. It comes from a different customer with different work, different review norms, and almost certainly a different measurement method (self-reported time savings is the usual source, which overstates). It is a marketing input, not a planning input.
- The 9% first-draft time reduction is real but small and partly offset. It is the one measured effect in the firm's own context.
- The 12% increase in senior-review time is a real cost and it is senior time, which is more expensive per hour. Net effect on delivery cost may be close to zero or negative until juniors learn to use the tool well.
- 'Time saved' is illusory unless converted. The firm bills hourly and is near capacity. Saving an hour of a consultant's time does not create revenue unless that hour is billed to a client or redeployed to billable/new work. If utilization is at capacity, saved time may simply become idle or absorbed - no P&L impact.
- The 35% weekly active use after week 3 is a warning, not a footnote. Low, declining adoption means the measured 9% applies to a shrinking base, so firm-wide value is far below 9% × 1,200.
Honest read: the pilot does not yet support a 12-month payback at full seat cost. The value case is unproven, not disproven.
2. Leadership insists on a 12-month payback. Propose a scoped deployment and the specific conditions under which the full rollout would be justified.
Read the worked answer
Scoped deployment (do not buy 1,200 seats yet):
- Target where the tool actually helps. The 9% gain and the review cost are concentrated in first-draft work. Deploy to the roles and document types where drafts are the bottleneck, and not to the whole firm. Start with the pilot cohort plus a matched expansion - maybe 150–250 seats - to get a real base rate.
- Attack the review-time penalty directly. The 12% senior-review increase is a training and template problem. Pair the tool with firm-specific prompt templates, examples of 'good' drafts, and a short review rubric. If review time returns to baseline, the net gain becomes real. Track this as the primary success metric.
- Fix the value-conversion question. Decide how saved time becomes money: either bill it (requires client agreement on scope/rates) or redeploy to business development or new offerings. Without this, no amount of time savings closes a payback.
- Address adoption. 35% weekly use means the tool isn't sticky. Find out why (workflow friction, trust, no perceived benefit) before scaling spend.
Conditions for full rollout:
- Net delivery time (draft + review) is down, not just draft time.
- Weekly active use stabilizes at a level that justifies per-seat cost (e.g., majority of target users).
- A credible path converts saved hours to billed or redeployed value.
- Payback modeled on measured net effect and actual adoption, not the vendor's 30%.
Bottom line: a smaller, instrumented deployment can hit a 12-month payback on the seats that work; a 1,200-seat commitment on current evidence cannot be defended.