Six hours per monthly client report. One delivery in three comes back with corrections and gets redone from the top. Four accounts.

That is 6 × 4 × (1 + 1/3), which I ran and which comes to 32 hours a month, or 384 a year. Multiply by what an hour of that person costs your company and you have the number any AI proposal has to beat. The six hours, the one-in-three and the four accounts are made up. The mechanics are not.

The mechanics come from Andrew Amann, CEO of the AI development agency NineTwoThree. On October 16, his State of AI newsletter ran a piece titled “Steal our AI ROI formula”. It has no formula. It has four things to write down, and the multiplication above is my addition, not his.

What Amann asks you to write down

His starting complaint is that “make us more efficient” gives an AI project too much room to drift. So he picks one recurring workflow and records:

• What an hour of the work costs.
• How long one completed pass takes.
• How often it needs to be done again.
• How many times it runs in a month.

Then comes the line I would put on the wall of a marketing office: “get the baseline before you ask someone to estimate the build.”

In his words, the baseline lets you compare a prototype with the process people already use, and it helps decide whether the opportunity deserves a custom system or a simpler fix.

It also works for the other workflows on a marketing team’s plate. Building a campaign has a pass, a rework rate and a monthly rhythm. So does following up on a new lead. Each one needs its own four numbers, because one workflow’s rework rate says nothing about another’s.

Amann sells AI development. I sell a discovery program. Whichever of us you talk to, the strongest position you can take into the meeting is a number that is already yours. A 3-hour monthly task does not earn a custom system. A 32-hour one earns a quote, and you can push back on that quote because you know the starting point.

Where the numbers go on Lieberman’s grid

On October 15, Alex Lieberman posted the AI Portfolio Grid on X. It is a two-by-two matrix with two questions: can we execute this today, and how big is the prize? Three of the four boxes are Accelerate now, Quick wins and Deprioritize. The fourth holds big prizes the team is not ready for yet. I read his X post, not the longer article it summarizes.

The two ideas fit together neatly. The prize axis is the baseline: 32 hours against 3.

The readiness axis is trickier. Among the readiness questions in Lieberman’s post is “is the process understood and measurable?” If you could not fill in Amann’s four numbers, you just answered it. That initiative belongs in the big-prize, not-ready box, and the work to do first is on the process, not on the tool.

Lieberman also makes room for small fixes. Automating a recurring report or checklist rarely ties to a company goal, he says, but these fixes add up, and if one is cheap, has an owner and will show results within weeks, you do it. A monthly report is close to the textbook case.

What this looks like inside AI Maestro

AI Maestro runs for two months. The first deliverable is the Process Reality Map, which records how the work actually gets done, who does it and how long it takes. The four baseline numbers come out of that map. Every opportunity then gets an AI Opportunity Score, and the whole thing ends at a Go/No-Go gate.

Sometimes the answer is that the task needs a classic automation or a redesigned process, not an AI agent. I made that argument in “Marketing Doesn’t Need More AI. It Needs Less Sediment.” The baseline tells you which situation you are in.

The same measuring gap shows up on the invoice side. I wrote about marketing teams spending more on AI without measuring it.

One afternoon this week

Pick a workflow. Sit with the person who does it, not their manager. Ask how long a full pass takes, how many times it came back last month and how many times it ran. You supply the hourly cost from your own payroll.

Then read the next AI proposal with that sheet next to it. If you also want to tie the baseline to campaign decisions, IQ Source’s data-driven optimization service starts from the same place: measure what happens today before changing it.