When the CFO Vetoes: The Cost Math That Survives Procurement

The ROI of IT automation is one division, and a CFO will do it in about eleven seconds: what does it cost to resolve one issue today, and what does the new thing charge to resolve the same issue? Published benchmarks put a Tier 1 contact at $20 to $30 of operational cost and a Tier 2 contact at $50 to $100. Dex is priced at $1.99 per resolved issue. That is the whole argument, and it is the only version of the argument that survives the room your champion was not invited to.

Everything else in the business case — the productivity narrative, the engineer-hours story, the transformation slide — is commentary that finance discounts by default. This post builds the block that isn't discounted: a per-resolution total cost of ownership comparison a champion can paste into an email, forward to a CFO, and have survive contact with procurement. It matters right now because the deal no longer dies at discovery. It dies after approval, on cost and trust.

Why the veto moved to the end of the process

The purchase decision and the funding decision are two different events, and only the second one belongs to finance. G2's 2026 Buyer Behavior Report, based on 1,038 B2B software buyers surveyed in June 2026, found that 49% of buyers had a CFO veto an already-approved software purchase in the previous 12 months. Finance involvement in software decisions climbed from 31% to 46% over the same period.

Two more numbers from the same study explain why AI purchases specifically are exposed. Among organizations with a dedicated LLM or token budget, the veto rate rises to 54%. And concerns about internal resistance to AI adoption grew from 16% to 29% in a single year — the largest single-year shift in the entire study.

Read those together and the picture is not hostility. It is a finance function that has been surprised by AI spend before and has responded by moving its review to the last possible checkpoint. Your champion is now relaying a case to a reviewer who has already been burned by a category, not by you. We mapped the full relay problem in what a CIO's AI buying committee actually asks; this post is the finance seat's block, written to be forwarded.

What the CFO is actually objecting to

A CFO rarely vetoes a price. They veto a number they cannot forecast.

The distinction is worth being precise about, because it changes what evidence helps. If the objection were price, a discount would fix it, and discounts routinely fail to unstick these deals. The objection is usually structural: the cost is fixed, the benefit is variable, and the reviewer has no way to bound the downside. A tool with a platform fee plus per-agent seats plus an AI add-on tier bills the same whether it resolves 8,000 issues or eight. Its cost per resolution is genuinely unknown until the contract year is over.

That is an uncomfortable thing to defend in a fifteen-minute review, and champions usually try to defend it with upside — productivity gains, deflection rates, hours recovered. Upside arguments lose to unbounded downside every time. The move that works is to change the shape of the cost, not the size of it.

The two cost shapes, side by side

This is a comparison of pricing structures, not of vendors. The seat-based shape is common across the category and it is not inherently wrong — it is simply built from three layers that all bill on a calendar rather than on output. A flat platform fee, charged before a single ticket is touched. Per-technician seats, which scale with headcount, a variable that has no relationship to how much IT work exists. And an AI add-on tier, which is where the automation you are actually buying lives, sold per seat on top of the other two.

None of that is a scandal. It is just a fixed floor. And a fixed floor is exactly the object a CFO has learned to veto, because it converts an uncertain benefit into a certain liability.

Usage-based pricing removes the floor. Dex bills $1.99 per issue it fully resolves. No seats, no subscription, no minimum. The line moves only when work gets done, which means the worst case is a zero rather than a stranded commitment. To be fair about the trade: at very high resolution volume against a very small team, a fixed floor amortizes and can win on absolute dollars. What it cannot win on is forecastability, and forecastability is what is being priced in that room.

One thing this comparison deliberately does not touch: your ITSM. The service desk platform you already own is the system of record, and its cost is not the line under review. What is under review is the automation layer bolted on top of it.

The per-resolution TCO block

Here is the artifact. Replace the four inputs with your own numbers, keep the structure, and send it as-is. It is built to be read by someone who will not open an attachment.

INPUTS (replace with your own)
  Employees                       500
  Tickets per employee per year     4      → 2,000 tickets/year
  Tier mix                        60% T1 / 30% T2 / 10% T3
  Loaded cost per contact         $25 T1 / $75 T2 / $200 T3
                                  (published benchmark range)

TODAY — cost of the resolution path
  1,200 Tier 1  ×  $25   =  $30,000
    600 Tier 2  ×  $75   =  $45,000
    200 Tier 3  ×  $200  =  $40,000
                            ────────
                            $115,000   operational cost/year

WITH AUTONOMOUS RESOLUTION
  Resolved by Dex:
    90% of Tier 1   =  1,080 resolutions
    50% of Tier 2   =    300 resolutions   (conservative)
                       ─────
                       1,380 resolutions

  Cost:      1,380  ×  $1.99   =    $2,746   billed only on resolution
  Avoided:   1,080  ×  $25     =   $27,000
               300  ×  $75     =   $22,500
                                  ────────
                                   $49,500

  NET, INVOICE COST ONLY                $46,754
  Effective cost per resolution           $1.99  vs. $25–$75 today
  Downside if nothing resolves                $0
  Cost to start                               $0  ($100 free credit
                                                   ≈ first 50 resolutions)

Three properties make this block survive procurement. The unit price is fixed and stated, so the reviewer can forecast. The floor is zero, so the downside case is bounded. And every input is a number the organization already has, so nothing depends on trusting a vendor's model.

Deliberately absent: the loaded-cost multiplier. The real cost of a ticket runs three to five times the invoice cost once user wait time and engineer interruption are counted, and we have made that case at length in the IT helpdesk math. Leave it out of the forwarded block. A CFO who has never seen that argument will read it as inflation, and it costs you the credibility of the numbers that are load-bearing. Bring it to the second meeting, if there is one.

The three questions that come back

When the block lands, the reply is usually one of three questions. Your champion should have all three answers before they hit send.

"What am I committed to?" Nothing. No seats, no minimums, no annual commitment. Every new account starts with $100 in free credit, roughly the first 50 resolutions, with no credit card. The pilot's true cost is the integration time, not the software.

"What counts as resolved, and who decides?" A resolution is an issue investigated, planned, and executed end to end against the real backend — Entra ID, Exchange Online, SharePoint, Intune — and closed. Not a deflection to a knowledge base article. Not a better-formatted ticket. Billing triggers only on full resolution, and every action lands in both native Microsoft 365 logs and Dex's own Activity Log, so the invoice is auditable against the work.

"What stops it doing something expensive?" Every action must match an explicit policy, enforced in the execution layer rather than in a prompt. No policy, no action. Dex never grants admin roles and never bypasses MFA, and Dex Pro runs on delegated permissions — the admin's own OAuth token, scoped to what that admin can already do — rather than a broad application key.

The scope assumption the whole model rests on

One line in the block does more work than the rest: the Tier 2 row.

Dex autonomously resolves L1 through L3, not L1 only. Routine Tier 1 work — password resets, MFA recovery, group and license access, provisioning — plus the deeper Tier 2 and Tier 3 troubleshooting, configuration, and engineering-adjacent tasks that used to need a senior technician. Genuine architectural and judgment calls escalate to a human with full context attached.

That scope is the difference between a good model and a marginal one. A Tier 2 resolution avoids $50 to $100 of operational cost against the same $1.99 unit price. An L1-only tool — which is what most of the category ships — caps its own return at the cheapest tickets you have. When a reviewer asks why the savings number is what it is, the answer is scope, and it is worth checking that any tool you compare against actually clears the same bar.

What to do Monday morning

Rebuild the block above with your own four inputs. Ten minutes with last year's ticket export gets you volume and tier mix; use the published benchmark range for cost per contact and say that you did.

Then send it to your champion with one instruction: forward this, don't summarize it. The summary is where the forecastability disappears, and forecastability is the entire product being sold to that room.

A CFO does not veto software. They veto a number they cannot forecast. Give them one they can, and the meeting you were not invited to stops being the one that kills the deal.