Keelson · Blog

09
Strategy

How to Build an AI Business Case for Mid-Market Leadership · Keelson

How ops and finance leaders quantify time savings, frame ROI for the CFO, scope a pilot the organisation can actually run, and get cross-functional buy-in before a single dollar is spent.

A working note for the ops leader who has to win the internal meeting before the vendor conversation starts — one metric, one pilot, one page the CFO signs.

01 · Quantifying time savings

How to turn anecdotal time loss into a number the CFO can verify.

“Our team spends hours on this” is not a business-case number. It becomes one when it is reproducible — a metric the CFO can read on a Monday dashboard before and after the engagement, without the vendor in the room. The method is simple: pick a single workflow, count the hours per week across every operator who touches it, convert at the average fully-loaded rate, and confirm the CFO can pull the underlying data from a system finance already reads. If the hours are not in a system the CFO reads, the hours are not a metric.

Counting the right hours
Count only the hours the AI will displace.

The business case counts only the hours that the AI will displace, not adjacent coordination time. Claims triage, contract review, settlement matching, defect classification: each of these has a countable unit and a countable operator. A workflow with a countable unit is a workflow the business case can prove. A workflow without one is a strategy memo that will not survive the CFO meeting.

Making it verifiable
The CFO must be able to verify it without calling you.

A number only becomes a business-case number when the CFO can verify it without calling the ops team. Pick the metric that already lives in a system the finance team reads on a Monday morning. If it is not in that system before the engagement starts, it is not a business-case metric — it is a promise the vendor makes in a slide.

02 · Framing ROI for the CFO

The one metric, the one page, the one meeting where the case is won or lost.

CFOs at mid-market companies do not have time for a 30-slide deck. The business case that gets signed is one page: a metric the CFO already reads, a before number the team can verify right now, an after number the vendor is willing to lock in writing, and the time to get there. Before, after, time-to-result: three numbers and one sentence. The sentence is: if the number does not move, the invoice does not go out. That is the business case.

What the CFO actually needs
Three numbers and a sentence.

Before, after, time-to-result. The sentence is: if the number does not move, the invoice does not go out. That is the entire business case. A CFO who reads those three numbers and that sentence has everything needed to sign the lock doc. A business case longer than one page is a business case that has not yet found its metric.

What kills the internal case
Projected gains the CFO cannot verify.

A business case built around projected efficiency gains the CFO cannot verify does not survive the meeting. A vendor who will not put the metric in writing is not the vendor. The CFO’s job is to read numbers already in the system, not numbers the vendor is promising to put there — and a CFO who knows this will ask the question the slide deck does not answer.

03 · Identifying the right pilot scope

The one workflow the business case can prove in 90 days.

The commonest mistake in an AI business case is scoping the pilot around what the operator wants to build, not what the business case can prove. A pilot that touches four workflows and three systems cannot produce a clean before-and-after in 90 days. A pilot on one workflow, one metric, one named operator can. The pilot scope is the business case: if it cannot be scoped to 90 days, the case is not ready to go to the CFO.

One workflow
Name it by the word the team uses on the floor.

The pilot that wins the internal case names one workflow by the word the operator uses on a Tuesday morning — not “operations efficiency” but claims triage; not “back-office automation” but settlement matching. The name is the scope. If the name covers four things, the pilot covers four things and the business case cannot be proved in 90 days.

One metric
Lock on the number the CFO can verify before and after.

The pilot that wins the internal case locks on one number the CFO can verify before and after without the vendor in the room. The metric is the scope. A pilot that cannot be evaluated against a single verifiable number is a pilot the CFO cannot sign off on — because the CFO cannot read it on a Monday morning.

90 days
A result the CFO can read by end of next quarter.

A pilot that takes longer than 90 days to show a result will not survive a budget cycle. The business case needs a number the CFO can read by the end of the next quarter — not a progress update, a dashboard, or a quarterly readout. One number, before and after, in the same system the finance team reads every Monday.

04 · Getting cross-functional buy-in

The one meeting where ops, finance, legal, and the workflow owner sit together.

An AI business case that ops built without finance will not survive the CFO meeting. One that finance built without ops will not survive the first exception fire. The buy-in meeting is the meeting where all four functions — ops, finance, legal, and the workflow owner — sit in the same room, read the same one-page case, and each sign the same lock doc. If any of the four cannot sign, the business case is not done.

Ops
The workflow description and the exception taxonomy.

Ops owns the workflow description, the exception taxonomy, and the runbook. Ops is the function that can name the workflow by its Monday-morning name — the name the team uses before the vendor arrives, not the name that appears in the vendor deck. A business case that ops has not co-written cannot be named correctly, and a business case that cannot be named correctly has not scoped the pilot.

Finance
The metric and the verification method.

Finance owns the metric and the verification method. Finance is the function that can confirm the number already lives in a system they read on a Monday morning. A business case that finance has not co-written has a metric finance cannot verify — and a metric finance cannot verify is a promise, not a business case.

Legal
The data-residency and escalation lines.

Legal owns the data-residency and escalation lines. Legal is the function that confirms what data the pilot is allowed to touch, which records leave which environment, and what happens when the exception queue crosses the escalation threshold. A pilot that has not cleared legal has not confirmed scope — and a business case that has not confirmed scope has not scoped the pilot.

Workflow owner
The named operator who carries the runbook after handoff.

The workflow owner is the named operator who will carry the runbook after handoff. This is the person whose name goes on the escalation contact sheet on day one — not a title, a named human. A business case that does not name the workflow owner has not planned for the Monday morning three months from now when the vendor is off the account.

05 · The lock doc

The artifact that closes the internal case and opens the vendor conversation.

The business case ends in a one-page artifact the four functions sign: the workflow by name, the metric and its before number, the 90-day time-to-result, the escalation contact, and the sentence that says the invoice does not go out if the number does not move. That artifact is the internal business case and the vendor brief. A vendor who will not sign that sheet is not the vendor.

The self-serve audit at /audit walks the eight questions that produce the lock doc in twenty minutes. Or talk to a build operator directly at /start if you already know which workflow the business case is built around.

Run the 30-day auditTalk to a build operator
5 min read
Published 2026-08-24

Continue reading

More working notes for operators building AI systems that hold up after handoff.

How mid-market teams scale AI workflows after the first build: strengthen governance, change operator behavior, assign durable ownership, write runbooks and escalation paths, and expand a portfolio without losing metric discipline.

How mid-market operators choose one workflow, establish a baseline they can defend, measure implementation ROI, set pilot exit gates, and hand a production system to the person who will run it on Monday morning.

A practical operating checklist for mid-market teams: inventory every AI workflow, assign owners, set risk tiers and approval gates, control data and vendors, define human review and exception escalation, log what matters, and review the system monthly and quarterly before drift becomes an incident.

The practical AI ops digest

Join the practical AI ops digest — one short email a fortnight.

Working notes on outcome-locked pricing, the 90-day build, and what AI Ops actually does after handoff. No marketing — just the operator notes we send to the people who already read this blog.

Bring us the workflow

Build the case in twenty minutes, then bring us the workflow.

The self-serve audit ends in the same one-page lock doc — workflow by name, metric the CFO can verify, 90-day time-to-result, escalation contact on cover. No NDA required to start the conversation.

Replies within one business day.