On looters and operators.

A CMO friend told me last month that her enterprise had spent $2.4 million on AI strategy work over the last year and shipped exactly nothing. Not nothing of consequence. Nothing. A series of decks, three frameworks, two governance committees, and a partnership with a Big Four firm whose final deliverable was a roadmap that ended in another roadmap. When I asked what was running in production, she laughed. The laugh said the part out loud.

I have been thinking about Francisco d'Anconia.

For readers who skipped Atlas Shrugged or wisely set it down at page 200, d'Anconia is a copper magnate who delivers a long speech defending money as the proxy for productive work. The line that earns its keep: money cannot exist unless something has been produced. Wealth is the material trace of effort, intelligence, and the willingness to ship. The corollary, which Rand borrowed from older arguments, is that there are two kinds of people in any economic system. There are producers, and there are second-handers, the people who consume what producers create without producing in their own right.

I am not here to defend Atlas Shrugged. The book has problems. The argument it borrows does not.

In the AI consulting business as it stands in 2026, that distinction has become legible. There is a producer pattern and there is a second-hander pattern, and they are increasingly easy to tell apart.

The producer's pattern

A producer in this work ships systems. The output is something that loads in a browser, runs in a pipeline, scores a record, decides a next-best action, sends an email, retrieves a customer, recovers a churned account. The output exists. It can be measured. The customer can switch it on Monday and switch it off Friday and observe the difference.

A producer's invoice describes a thing that happened. The work being charged for can be pointed at. Not "we ran a discovery sprint." A specific platform module, in your stack, doing a specific job, against a baseline.

The producer's economics are honest because they have to be. If the system does not work, the customer notices on the first invoice cycle. If it does, the customer scales it. Pricing becomes value-for-value because the value is observable.

Innovative Group ships this way. Our Next-Best Action work is not a slide. It is a deployed pipeline that we built for ourselves first, ran for one quarter, measured (a 40% lift in qualified pipeline, 75% reduction in manual segmentation effort, eight-week deploy), and then offered to the market. We call this Customer Zero because that is what it is. Before we sell a thing, we use it ourselves.

The second-hander's pattern

A second-hander in this work ships decks. The output looks like work. It has a logo, an executive summary, a maturity model, a five-by-five matrix, a heatmap of pain points, and seventeen pages titled "next steps." It does not load in a browser, score a record, or decide anything in production.

The second-hander's economics depend on opacity. The customer cannot tell whether the work is good because the work is internal to the customer's own head. The deliverable is "alignment" or "framework" or "readiness assessment." All of these things may be real. None of them are testable in the way a deployed system is testable.

This is the structural reason the AI strategy industry is so much bigger than the AI deployment industry, and why the gap is widening. Strategy work survives evaluation only by avoiding it. Deployed systems get evaluated every business day.

Why 2026 is exposing the gap

For a long time the gap was hidden. AI was new enough that buyers could not tell the difference between a pilot that did not ship for technical reasons and a pilot that did not ship because the consultancy lacked the engineering chops to ship anything. The two looked identical from the outside. The CMO who paid $2.4 million for nothing had no clean way to say what specifically should have been delivered.

That has changed. The bar moved. Production AI is now reachable in eight weeks for most marketing and sales workflows. The platforms have matured. The models have matured. The integration patterns are documented and repeatable. The technical excuse is gone.

The remaining excuse, when an engagement produces no shipped artifact after a year, is that the engagement was always going to produce no shipped artifact. The work was performance, not production. The fee paid for the appearance of strategic motion. d'Anconia would call this transferring wealth from producers to second-handers. I would call it agency.

What this means if you are a buyer

Two questions. Ask both at the start of any AI engagement, before the SOW is signed.

The first. At the end of this engagement, what specifically is running in our environment that was not running before? Make them name it. Make them describe the inputs, the outputs, the system of record it touches, the user-facing surface where the output lands. If they describe a deck or a roadmap, they are selling you a roadmap. That can be valuable. It is not deployed AI.

The second. How will I know in 90 days whether this worked? Make them define the metric, the baseline, and the measurement window before the work starts. The producers will give you a number. The second-handers will give you a process. The number is the thing that survives contact with reality.

What this means if you are an operator

Stop apologizing for the production focus. The market is finally rewarding it. The CMO who spent $2.4 million on nothing is now in budget review with her CFO, and the conversation is not about strategy. It is about what shipped.

The question for the operator practice in 2026 is not whether to position around shipping. That is settled. The question is how loudly to say it. My answer, after watching this market for thirty years, is louder than feels comfortable. The buyers who matter are exhausted. They will reward the firm that says the thing they have been afraid to say out loud, which is that most of what they bought last year was performance, and they need the next thing to be production.

That is the operator's claim. It is not a sales pitch. It is the moral high ground d'Anconia tried to defend in fiction sixty-eight years ago. It happens to be the same ground that wins in 2026.

← More from The Shift