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

Every conversation we have with a CFO or finance person about AI starts with the same question: "What's this going to cost us?" Fair question. The honest answer has more to it than most companies tell you.

There are two kinds of cost, and they look very different:

1. The cost to build it. Real money. A one-time project cost you can plan for.
2. The cost to run it. So small it changes how you think about everything else. We mean that literally.

This post is about the second one — the running cost — because that's where most CFO worries go away once you see the numbers written down.

What the AIT weekly summary actually costs to run

We built a weekly summary for the American Indigenous Tourism organization that arrives every Monday morning. It pulls together information from their office tools, their email marketing, and their grants system. It also looks at current news and federal policy from that week.

The total cost each week: about two cents.

The breakdown:

• Four quick research checks against current news = about 2¢
• One AI call to write the summary itself = less than a penny
• Sending the email = $0 (already paid for)
• Reading their own data = $0 (it's their data)

Even being generous and rounding up, it's about a nickel per send. That's $2.60 a year, per person who reads it. For a summary that replaces about three hours of someone’s Sunday night.

The worry that doesn’t hold up to the math

A common worry we hear:

"We can’t keep adding monthly subscriptions. AI is going to be another recurring expense, just like every other piece of software we pay for, and they add up."

That worry comes from thinking about AI the same way we think about software like Microsoft 365 — where you pay $10 or $20 per person per month, every month. AI works differently. You pay only when you actually use it, and each use costs a fraction of a penny.

In real numbers:

• A traditional reporting service that gives you a weekly summary: $5,000 to $15,000 a year
• The same kind of summary built with AI: $50 to $200 a year to run
• The difference goes into building it once instead of paying every month

For most groups, building it yourself is 30 to 100 times cheaper to run. The trade-off: you pay more upfront, but you own it.

What it costs to build

We’re open about this on our services page. For a weekly summary like the one we built for AIT:

• First week, figuring out what you need: about $5,000
• Three weeks of building: about $15,000 to $25,000
• One more week of fine-tuning: about $5,000
• Total to get it working: $25,000 to $35,000

After that:

• Monthly help keeping it useful: $2,500 to $5,000 a month
• Cost to actually run it: $50 to $200 a year

Now compare that to the other choices:

Choice 1: Hire a junior person to do this by hand

• Salary plus benefits: about $80,000 to $110,000 a year
• Quality depends on the person
• Most burn out and leave every two years or so
• Five-year total: $400,000 to $550,000

Choice 2: Buy a big reporting service

• Yearly cost: $20,000 to $50,000
• You have to use their format, their dashboards, their way of doing things
• AI summaries are usually an extra fee
• Five-year total: $100,000 to $250,000, plus the time to learn their system

Choice 3: Custom-built AI summary (us)

• Year 1: $25,000 to $35,000 to build + $30,000 to $60,000 for ongoing work = $55,000 to $95,000
• Years 2 to 5: $30,000 to $60,000 a year for ongoing work
• Cost to run: almost nothing
• You own it — if you stop working with us, you keep the system
• Five-year total: $150,000 to $320,000, and it keeps getting better over time

A custom build ends up costing about the same as a big reporting service, but with two important differences:

1. It does what your team actually needs. It answers your questions, in the format your team uses, instead of forcing you into someone else’s template.
2. It gets smarter on its own. When the AI companies (like OpenAI or Anthropic) release better tools, your summary gets better too. With other services, they collect that improvement and charge you more for it.

What about the risks?

The worry most CFOs have but don’t say out loud: "What if we spend $30,000 on this and it doesn’t work?"

Here are the real risks, from most to least likely:

Most likely (but small impact): the first version needs adjustments

This almost always happens. It’s normal. The four-week build includes a final week of adjustments for exactly this reason — the first version is never quite the right format, the right timing, or the right level of detail. Plan for it, and it’s not a problem.

Sometimes happens: your information isn’t organized enough yet

This is the most common reason AI projects don’t work — but we find out in week one, not month six. If your information isn’t ready for what you want to build, we tell you in the first week. We either change the plan or pause the project. Much better to find out then.

Rarely happens (but bigger impact): the AI tools change a lot

This could happen if AI companies change their pricing or what their tools can do. We protect against this by building things that can switch between different AI providers easily. We can swap to a different one in days, not weeks.

Very rare: your staff doesn’t use it

This is actually rare when staff is part of the design from the start. The way this fails is when leadership decides on an AI tool without asking the people who would actually use it. We don’t take projects where the day-to-day users haven’t been included in the conversation.

What to ask before you commit

If you’re looking at an AI proposal from anyone — us or anyone else — these are the questions that’ll tell you what you need to know:

1. What does it cost to actually run each time you use it? If the answer is in dollars instead of cents, ask why.
2. What do you own versus what the company owns? If you stop paying them, what keeps working?
3. What’s the total cost over five years? Get it in writing. Our services page is straightforward about ours.
4. How do you stop if you need to? Every project should have a clear way out.
5. Where does your information live? Especially important for tribal organizations — see our questions for boards.

If you want to walk through the numbers for your specific group, the free 15-minute call is the right next step. Bring a rough idea of how much time your team spends on reports right now, and we’ll show you the math together.