Vendor-supplied ROI.
The number is wrong. Both you and the vendor know this. Signing a contract on it is a career risk for whoever owns the budget. And an integrity problem for whoever owns the financial narrative.
Service A.03 · Strategic engagement
Quantify the AI benefit before the contract. Re-quantify after.
An independent, defensible, board-ready benefit case. Built around your baselines, not the vendor's. Refreshed against actuals after deployment.
Why this matters.
Every AI vendor will give you a benefit case. It will be inflated. They will compare against fictional baselines, count benefits multiple times, and assume scale that has never been observed in any deployment of their product.
Your CFO knows this. Your board knows this. Whoever signs the contract is liable for an ROI claim that everyone in the room privately doubts.
What you need is an independent, defensible, board-ready quantification, built around your operating data, your risk-adjusted assumptions, and your measurable outcomes. Built before you sign. Refreshed after you deploy.
That's the engagement.
The trap.
Hand-wavy '10x productivity' claims.
Productivity of what? Compared to what baseline? Producing what measurable outcome that lives on a P&L? Most AI ROI claims fail this test, and 'we'll figure out the metrics during deployment' is the polite version of 'we don't have a way to measure this.'
One-shot ROI.
The benefit case is built once, at contract time. Six months later, nobody has measured against it. The original claim becomes mythology, quoted in board decks but never validated. The vendor renewal happens on inertia, not evidence.
What you walk away with.
An independent benefit case with explicit, defensible assumptions tied to your operating data. Every input traceable to a source you can defend in a board meeting.
A side-by-side comparison: vendor claims vs. our analysis, with deltas explained line by line. The kind of artifact that ends the negotiation and gets the procurement team to a 'yes' or a 'no' cleanly.
A risk-adjusted ROI range (best case, expected case, worst case) your board can interrogate. Single-point ROI estimates are a rhetorical move, not an analytical one. Yours will have honest bounds.
A measurement plan: what to track, how, on what cadence, who owns the report. So that 90 days from now, somebody can answer 'is it working?' with data instead of vibes.
A re-quantification 90 days after deployment with actuals vs. projections, and a clear recommendation on whether to expand the deployment, hold it where it is, or sunset it. The honest answer often saves more money than the original benefit case did.
Who this is for.
- CFOs, Heads of Finance, or PE Operating Partners signing AI contracts above $100K ARR
- Boards reviewing AI investment requests
- CEOs who need a defensible ROI claim for the next earnings call, fundraise, or board meeting
- Anyone who has been burned once by a vendor's benefit case and isn't going to do that twice
Engagement structure.
Two phases. Phase 1 is pre-contract: a two-week fixed-fee analysis producing the benefit case, vendor comparison, and risk-adjusted ROI range. Delivered before the contract gets signed. Phase 2 is post-deployment: a one-week engagement at month 3 to re-quantify against actuals, with a clear "expand / hold / sunset" recommendation.
Both phases stand alone. Our incentive is the integrity of the number, not the close of the deal.
Frequently asked.
Won't the vendor be upset if we challenge their ROI?
Sometimes. Good vendors welcome an independent benefit case because if it confirms theirs, it accelerates the buying decision and the procurement conversation gets shorter. Vendors who refuse independent scrutiny are telling you something useful. Pay attention to it.
Can you do this for an AI initiative we're building internally, not buying?
Yes. Internal projects often need ROI defense more than purchased ones, because there's no vendor to share blame with if it underdelivers. Internal AI builds are also where the most fictional ROI tends to live, because there's no procurement counterparty to push back on assumptions.
What if we don't have good baseline data?
That's the first finding, and it's actually common. We work with what's available (operational metrics, financial actuals, time studies if needed) and explicitly mark assumptions vs. measured baselines in the deliverable. The benefit case will be honest about what's measured and what's estimated. Honest is what makes it defensible.
How does this compare to a Big Four ROI study?
Faster, cheaper, and the analyst doing the work has actually deployed AI in production. Big Four ROI studies are often conducted by analysts who have read about AI but never shipped it. We've shipped it. The difference shows up in the assumptions.
Do you reuse benefit cases across clients?
No. Every benefit case is built from your operating data and your assumptions. We bring methodology, benchmarks, and pattern recognition; the numbers are yours.
What's the typical ROI gap you see in your analysis?
In our experience, vendor-claimed ROI is materially higher than what's actually defensible against a client's baseline. We're not anti-AI; we're anti-imaginary-baseline. Real, defensible AI ROI exists. It just isn't usually what's on the vendor slide.
Bring us
the messy one.
The system that's been on the roadmap for two years. The migration that's already failed once. The AI strategy that didn't make it past the deck. That's the one we want.