The $0.02 Executive Brief: Cost Math for the Skeptical CFO

Every conversation we have with a CFO or finance lead about AI tooling lands at the same question: "What's this going to cost us?" Fair question, and the honest answer is more nuanced than most vendors make it.

There are two costs and they have very different shapes:

1. The build. Real money. Project-based. A meaningful number that a CFO needs to evaluate against the deliverable.
2. The running cost. So small that it changes the framing. We mean this literally.

This post is about #2 — the running cost — because it's where most CFO objections evaporate when the math is actually written down.

What the AIT executive brief actually costs to run

Every Monday morning, the CEO of the American Indigenous Tourism organization receives an executive brief that we built. It pulls from their internal CRM, their email marketing platform, their grant tracking system, and runs four real-time intelligence queries against current news and federal policy.

The total weekly running cost: about two cents.

Specifically:

• 4 Perplexity Sonar queries × ~$0.005 each = $0.02
• 1 OpenAI / Claude API call to generate the narrative summary = $0.001 to $0.003
• Email send via existing infrastructure (already paid for) = $0
• Database queries against existing systems = $0

Round up generously and call it $0.05 per brief. That's $2.60/year per recipient. For a brief that replaces three hours of someone's manual report-pulling every Sunday night.

The CFO objection that doesn't survive the math

A typical version of the cost objection we hear in proposals:

"We can't keep adding subscription services. AI tools are going to be a recurring monthly expense like every other SaaS, and they add up."

This framing comes from the SaaS-vendor model, where you're paying $99/seat/month for the tool. AI infrastructure works completely differently. The dominant model for the kind of work we do is pay-per-API-call to the underlying model provider (OpenAI, Anthropic, Perplexity), and those calls cost fractions of a cent.

To put it in concrete CFO terms:

• A traditional SaaS reporting tool that produces a weekly executive brief: $5,000–$15,000/year
• An equivalent custom-built AI brief running on API calls: $50–$200/year in operational cost
• The difference goes into the build cost (one-time) instead of recurring fees

For most organizations the math favors the custom build by a factor of 30-100x in ongoing operational cost, with the trade-off that you pay more upfront and own the result.

What the build does cost

We're transparent about this in our services pricing. For an executive brief at the AIT scale:

• Discovery week (1): roughly $5,000
• Build sprint (3 weeks): roughly $15,000 to $25,000 depending on complexity
• Iteration through month 1: roughly $5,000
• Total first-month cost: $25,000 to $35,000 for a working production system

After that:

• Monthly retainer for evolution: $2,500 to $5,000/month
• Operational cost (API calls): $50 to $200/year

Compare against the alternative scenarios:

Scenario 1: Hire a junior analyst to do this manually

• Loaded cost (salary + benefits + overhead): ~$80,000-$110,000/year
• Quality is variable depending on the analyst
• Burns out and turns over every 18-24 months
• 5-year cost: $400K-$550K

Scenario 2: Buy enterprise reporting platform

• Annual subscription: $20,000-$50,000/year (after multi-year discounting)
• Required to use their data model, their visualizations, their workflow
• Doesn't include the AI summary layer; that's an upcharge
• 5-year cost: $100K-$250K, plus internal ramp time

Scenario 3: Custom AI brief (us)

• Year 1: $25K-$35K build + $30K-$60K retainer = $55K-$95K
• Years 2-5: $30K-$60K/year retainer (assuming evolution continues)
• Operational cost: trivial
• Owned by you — if you stop paying us, you keep the system
• 5-year cost: $150K-$320K, with ongoing capability improvements

The custom build comes in at roughly the same range as enterprise SaaS, with two material differences:

1. The capability is bespoke — it answers the questions your team actually asks, in the format your team actually uses, instead of the generic dashboard a SaaS provides.
2. The tool gets smarter as the underlying AI models improve. When OpenAI or Anthropic ships a better model, your brief gets better automatically. SaaS vendors capture that improvement and charge you for it.

The risk math

The CFO concern that's usually unspoken: "What if we spend $30K on this and it doesn't work?"

Real risks, ordered by likelihood:

High likelihood, low impact: the first version isn't quite right

Almost certain. This is normal. The 4-week build includes a final iteration week specifically because the first draft is never the right format, the right frequency, or the right level of detail. Budget for it and it's a non-event.

Medium likelihood, medium impact: the data foundation isn't ready

This is the most common reason AI projects don't deliver — but it's discoverable in week 1 (discovery), not month 6. If we find your data doesn't support what you're trying to build, we tell you in week 1 and we adjust scope or pause the engagement. Better to find out then than after the build.

Low likelihood, high impact: the AI model behavior changes radically

Could happen if OpenAI or Anthropic radically shifts pricing or capability. Mitigation: the build uses provider-agnostic APIs where possible, and the system is portable. We can swap models in days, not weeks.

Very low likelihood: the tool is rejected by staff

This is actually rare when staff is involved in design from week 1. The failure mode is when AI tools are imposed by leadership without the end-user team's buy-in. We don't take engagements where the day-to-day users haven't been consulted.

What to bring to the conversation

If you're a CFO or finance lead reviewing an AI proposal — ours or anyone's — the questions that produce signal:

1. What's the operational cost (not subscription cost) per output? If it's dollars not cents, ask why.
2. What's owned by us vs. owned by the vendor? If you stop paying, what continues to work?
3. Total 5-year cost projection? Get this in writing. Our services pricing page is honest about ours.
4. What's the off-ramp? Every project should have a documented exit.
5. Where does the data live? Critical for tribal organizations specifically — see our board approval framework.

If you'd like to walk through the math for your specific organization, the free 15-minute call is the right step. Bring a current operational cost (a person-hours-per-week number is enough) and we'll show the math live.