Stop putting every AI project in the same budget

AI is clearly making people faster. It is much less clearly making companies better.

Sam and I went pretty deep on this. We read dozens of reports, datasets, surveys, and studies, then compared them with what we have seen over two years of implementation work inside mid-market companies. We made a report and, because apparently one explanation was not enough, two different videos.

I think the picture is fairly simple. People are using AI. The people who are good at it are getting a lot faster. Very few companies have turned that into a new operating model.

Sam's way into it

Sam follows a manufacturer working a quote from inquiry to cash. AI can draft the quote in seconds. Great. Someone still has to confirm the specs, material costs, engineering approval, capacity, margin, and a delivery date the company can actually hit.

That seems obvious. It is also where a lot of the ROI goes to die. A faster answer is useful. It is not a return until the customer accepts it and the company can make and deliver the thing.

This is where Sam's answer and mine split a little. His video is about getting the work all the way through the business. Mine is about how a CEO should decide which bets to make in the first place.

The three portfolios

The most useful thing we made is probably not either video. It is the three portfolios.

Solved: implement it. Controlled knowledge search, extraction, coding assistance, straightforward integrations, and replacement of thin, expensive SaaS.

Conditional: underwrite it. Core ERP or CRM replacement, end-to-end workflow redesign, customer-facing agents with system access, and workforce redesign.

Exploratory: fund it like R&D. Unbounded high-stakes agents, chatbots acting as the operating system, irreversible decisions without controls, and bets tied to one model staying dominant or cheap.

The report

We put the charts, sources, full initiative lists, and ROI math in one place. If you want to argue with us, the receipts are there.

If you would rather watch

Sam made the operator version. He starts with that quote-to-cash workflow, shows where the handoffs eat the gain, and gets into what the high performers do differently. He also walks through Smith Mechanical, which saved more than $250,000 in first-year license costs after replacing ServiceTitan. That is one real field case, not a market benchmark.

I made the CEO version. Mine is shorter and more direct: what is basically solved, what needs a real company-specific business case, and what should stay in the R&D budget.

We can tell you what looks solved now. We cannot choose the right conditional bet for your company from the outside. We have opinions, obviously. But you need to find yours.