At Nrth (formerly Elevate) in Toronto, Qasim Naqvi, Partner at Strategy&, gave a masterclass called "The AI Gold Rush: Where Is the Value?" One exercise has stayed with me since.
He put two AI investments side by side.
Investment #1: 96% accurate, 90% adoption, 30% faster, cost unchanged.
Investment #2: 80% accurate, humans still handle 20% of cases, capacity up 15%, cost down 12%.
Then the question: which one would you pick?
Most of us reach for the first. It has the better score on almost everything. But look at what changed in the business. In the first, people are using a faster tool and the cost line hasn't moved. In the second, the firm can take on more work for less money.
The takeaway: technical success and investment success are not the same test. The better investment is the system that changes the business, not the model with the highest score.
That is an operations idea as much as an AI idea. In a gold rush, the people who build lasting wealth aren't swinging picks at random. They work out where the value sits, and position themselves to capture it.
Two habits from the session are worth borrowing.
Start with where your own rush creates value. Pick your industry, your market, your clients. Where does value actually get made in your firm, and where does it leak? Then ask where AI can change that. The problems that remain when you take AI out of the picture are where your real struggles live. Most people reach for AI to solve the problems they already know about, and skip this step.
Write the one-page memo before the funding. Before a serious AI investment, Naqvi's framework asks leaders to answer six questions:
- Value: what valuable problem are we attacking? Put a number on it.
- Change: what exactly will AI change, and against what baseline?
- Capture: what has to change in the workflow for the value to appear?
- Adoption: what behaviour proves people rely on it?
- Impact: which operating or financial measure moves if we're right?
- Next bet: what's the cheapest experiment that tests our weakest assumption?
If the memo can't answer these, the proposal isn't ready.
For small specialist firms, question three is where most plans quietly fall apart. The tool is easy to buy. The workflow around it is the hard part, because nobody has written down how the work actually gets done. There is no baseline to measure against, no clear owner, and no way to tell whether people are relying on the new tool or working around it. You can't capture value from a process you can't describe.
That's why "fix the process first" isn't a delay tactic. It's what makes the investment measurable.
And Naqvi's closing idea reframes the whole thing. AI adoption isn't a project with a finish line. It's a capability that compounds. The question isn't "are we done adopting AI?" It's "is our rate of getting better still accelerating?"
Try it with one tool you already pay for:
- Name the problem it was meant to solve, and put a number on it.
- Write down the baseline from before you started.
- Check what moved: time, capacity, cost, or error rate.
- If nothing moved, ask whether the workflow changed, or only the software.
If you can't do step two, that's your first project.
Ideas shared from Qasim Naqvi's "The AI Gold Rush: Where Is the Value?" masterclass at Nrth (formerly Elevate) 2026. Follow Qasim Naqvi on LinkedIn.
What would your one-page AI memo say about the tools you already have?