Want to become a partner or sponsor, click here to schedule time to talk with the team.
FinOps & Beyond is what engineering, finance, and IT leaders read to understand FinOps, and what it means for operating models, accountability, and spend decisions.
I wrote a short post on LinkedIn Friday and said I would come back to it here. So here we are.
What got me excited was OpenAI's economic research. I do not want to rehash the whole argument, but a core idea stuck with me: AI will expand the economic pie, but the real question is how that expansion happens and who gets what slice.
That is not only a technology question. It is a people question.
And it is the part of the AI value conversation that FinOps keeps skipping.
Work is changing. So is who gets to do it.
Two signals are getting clearer every week.
First, the work itself is changing fast. Not just what people do, but how they do it. The day-to-day of a finance analyst, engineer, support rep, or marketer already looks different than it did eighteen months ago. That gap is only going to continue to widen.
Second, a divide is opening between the AI haves and have-nots. Some people and organizations are learning how to use these tools well. Others are unwilling, unable, or simply unaware of how to adopt them. That divide is going to become a real business problem.
New roles are slowly starting to show up. Existing roles now require skills that are, by its nature, in short supply. And you cannot hire your way out of a skills gap that the whole market has at the same time. That second point is the one I keep coming back to in my mind, because it does not show up neatly in the bill. It is easy to count tokens. It is much harder to count whether your people can actually use the thing you are paying for or learning the new skills.
More work, not less
There is a comfortable story where AI quietly does the same work with fewer people and the line item shrinks. However, the evidence keeps pointing somewhere else. As The Atlantic argued in its recent profile of Erik Brynjolfsson, the more interesting effect may not be less work. It may be more work.
When the cost of doing something drops, we usually do more of it. Cheaper cognition means more analysis, more iterations, more tests, more drafts, more products, more workflows, and more ideas that were not worth pursuing before but suddenly are.
The work expands to fill the new capacity.
That is why the companies winning with AI are not simply doing last year's work with fewer people. They are taking on work they could not touch before.
For FinOps, this is an important point. If AI mostly expands the work, then measuring success as "spend went down" is measuring the wrong thing. The cost can go up while the value goes up faster. Throughput by itself is still a vanity metric. Cost per unit of valuable output is still the number that matters. AI did not change that math, but rather it made the math harder to ignore.
The people strategy is becoming visible
The haves-and-have-nots divide is not lost on industry and public leaders. Last week, a new effort called RAISE US launched to deal with exactly this. (Side note: I called it "Raise Up" in my Friday post. The real name is RAISE US.)
What is RAISE US? Well, it is a bipartisan effort led by Gina Raimondo and Eric Holcomb, with more than $500 million committed and anchor partners that include Amazon, Anthropic, Microsoft, Bank of America, and the OpenAI Foundation. According to Axios, the initiative will start with state partnerships in Arkansas, Connecticut, Maryland, and Utah, testing things like wage insurance, retraining incentives, AI-powered career coaching, and short-term credential programs.
You can be skeptical about whether a $500 million coalition moves the needle against a workforce of roughly 160 million people. I certainly am, but the signal is the point. The same companies selling the tools are now funding the reskilling (although the skeptics out there will think this is self-serving, which is a fair point). These companies and others can see the divide forming. They also know an economy of have-nots does not buy much in the long term. The human cost of this shift is being treated as a first-order problem by people who rarely agree on anything.
FinOps should take note.
What this means for FinOps
Most of the FinOps conversation about AI has been about tokenomics: the token as the unit of measurement, the run rate, the cost of an agent doing a task, the attribution model, the forecast. That work matters. I have discussed in previous issues.
But I believe it is one leg of the stool, and too many teams are trying to stand on one leg. The fuller job has three parts.
Count the tokens.
This is the part we already know. Meter the spend. Attribute it to a team, product, workflow, or customer outcome. Give the bill an owner and a denominator. Nothing below works without this.Track the human effort.
This is the leg most teams are missing are haven’t been including. When a team adopts an AI tool, something changes in the work. A task that took three people takes one. A report that took a week takes a day. An analyst handles questions she used to escalate. A support team resolves more issues without adding headcount.
That change is the value. Almost nobody is measuring it next to the cost.
You do not need a perfect productivity model. You need the spend sitting beside an output number finance already trusts, so the token bill has meaning and context.See where the capacity goes.
Once you can see cost and capability together, you can point the freed-up capacity somewhere useful.
More analysis that changes a decision. More product iterations customers actually feel. More support coverage. More experimentation. More work that was not worth doing before and now is. This is the upside of more work, not less.The practitioner who can steer leadership toward that upside is doing real FinOps.
Then comes the part nobody wants to talk about.
Do not over index on the extra work. More output is not automatically more value.
Cheaper cognition means you will generate more of everything, whether or not it was worth generating. More dashboards no one opens. More analyses that change no decision. More reports that exist because they got cheap to produce, not because anyone needed them.
The same Jevons effect that expands valuable work expands busywork right next to it. Now that busywork has a token bill attached. This is Visibility Theater's louder cousin.
The old waste was effort spent producing insights nobody acted on. The new version lets you produce ten times the insights at a fraction of the effort and call it a win.
It is not. It is faster waste with a usage meter running. So the practitioner's job is also to monitor the expansion. Ask the same questions of AI-generated work that you ask of any other spend:
What decision does this change?
What outcome does it move?
Who uses it?
What would stop happening if we stopped producing it?
If the answer is nothing, you found waste. It does not matter that it was cheap to make. Cheap waste at scale is one of the most expensive habits in the building.
Strategically, this widens the mandate. FinOps becomes the function that translates AI spend into value the business recognizes. And value now means three things at once:
What did it cost?
What did it make people capable of?
Was the extra output worth producing?
A team that can hold all three is worth ten teams that can only produce a cost dashboard.
Tactically, here is where I would start Monday:
Pick one team that adopted an AI tool and ask what changed in the work, not just what it cost.
Put the spend next to an output number finance already trusts.
Find your have-nots: the teams paying for licenses nobody knows how to use.
Find your over-producers: the teams generating more because they can, without knowing whether anyone consumes it.
Then take the story to leadership. Show what people can now do. Show the cost behind it. Be honest about the output that should not exist.
Cost without value is just a number. Value without cost is a fairy tale.
FinOps is the function that helps to hold both.
The job
The tokens are the easy part. They show up on an invoice, and we can argue about the unit price.
The harder part is what the meter does not show: whether your people can use what you bought, whether the work is getting better or just getting bigger, and who in your organization is quietly being left behind.
Count the tokens. Track the human effort behind them. Watch where the new capacity goes. And be willing to say when more work is just more work.
Count the tokens, then count what they changed. And make sure the change was worth having.
Written with the help of AI. All the ideas expressed are mine and mine alone.
FinOps Company Spotlight
If you would like your company included in the Spotlight, contact the CloudXray AI Team
Category: Managed Services & Consulting
What They Do: FinOps Consulting & Advisory Services; Owners & maintainers of the single largest FinOps company directory (finops.cloudxray.ai)
Why It Matters: Companies still need guidance on implementing FinOps and understanding the landscape of companies that exist