RevOps + AI: From Prompt to P&L
How generative-AI engineering connects to conversion, CAC, and revenue.
Most AI conversations inside a company stall at the same place: a demo everyone agrees is impressive, and nobody can say what it is worth. RevOps is the discipline that answers that, and it is why I think the two belong in the same job.
Revenue operations is a measurement discipline
RevOps exists because marketing, sales and success each optimise a local number and the company optimises none of them. Its job is one model of the funnel — one definition of a lead, one attribution story, one place the number comes from.
That is exactly the missing piece in most AI programmes. Not model quality. Measurement.
Where generative AI actually touches the P&L
In my own work the connections have been unglamorous and specific.
- Cost per acquisition. An AI ad-copy workflow across 600+ ad groups moved Quality Scores from the 3–5 band to 7–9. Quality Score is a cost multiplier, so that is a CPA change, not a creative one.
- Implementation time. Prompt infrastructure cut implementation time substantially, which converts directly into how many engagements a fixed team can deliver.
- Conversion. Reducing abandonment in a checkout is a content and sequencing problem before it is a model problem, but it is measurable in the same week you ship it.
Every one of those is a line somebody already tracks. That is the point — if your AI initiative needs a new metric to look good, it probably is not working.
The sequence that works
Instrument first. Pick the number. Ship the smallest system that could move it. Measure against the pre-period, not against a projection.
It is slower than the demo path and it is the only version that survives a budget review.
Why the combination is rare
Most AI architects cannot sell, and most sellers cannot build. The gap is not talent, it is that the two disciplines are learned in different rooms. I learned the commercial side closing transactions and the technical side building systems, and the only real advantage it gives me is that I am unwilling to ship something I cannot attach to a number.
Keep reading
A Run Produces Data. A Loop Produces Decisions.
The signature thesis: the data you scrape should plan the next scrape. Loops beat runs.
Read →What Is an AI Systems Architect?
A working definition — from someone who designs, codes, and operates the systems.
Read →MCP for Operators: Data as Tools
Why the Model Context Protocol turns every product and dataset into a callable tool.
Read →