Don't start with the AI... start with what you're trying to achieve.
In the 1770s, James Watt had a problem. He'd built a far better steam engine than anything on the market, and mine owners weren't buying it. Engineers kept pitching them on piston size and crankshaft length, but that isn't what a mine owner cared about. They wanted to know one thing: how many horses can I get rid of?
So Watt invented a unit to answer that question: horsepower. It was never a scientific measure. It was a marketing one, and we still use it to sell cars today. Then he went further. His firm would put the engine in and take a third of whatever the customer saved on coal. If the customer didn't win, neither did he.
I heard that story from Rory Sutherland at a conference recently and it's stayed with me. I can't think of a better description of what a good technology partner looks like.
The problem with being sold technology
Most businesses we speak to have been sold something at some point that didn't quite do what they hoped. A system that looked great in the demo. A licence bundle full of features nobody uses. A project that ticked every box on the spec but didn't change how anyone actually worked.
Right now, the thing being sold hardest is AI. To be clear, AI can save real time and money. But almost anyone can sell you a Copilot licence. Far fewer will sit down with you first, ask what you're actually trying to achieve and tell you plainly whether AI is the right answer.
Rory's main argument was simple: don't start with the numbers. If you begin with a fixed budget or a fixed product, you narrow the possible answers before you've understood the question. Technology works the same way. Start with the tool and you'll find ways to use the tool. Start with the outcome and you'll find the right solution, which is sometimes a much smaller and cheaper one.
Ask a better question
One of Rory's examples was electric cars. Manufacturers are spending billions on bigger batteries to cure range anxiety. His question was whether it would be cheaper to reduce the anxiety instead, with better information about chargers, clearer battery displays and reassurance on the dashboard. It's the same problem with a very different answer.
We see this all the time. A client asks for faster laptops because "everything's slow". Sometimes that's the right call. Sometimes the real issue is a shared drive so tangled that people spend twenty minutes a day hunting for files, and fixing it costs a fraction of a hardware refresh. You only find that out by asking what's actually getting in the way.
The doorman fallacy
Rory also talked about what he calls the doorman fallacy, an idea he explores in his book Alchemy. A consultant visits a hotel, decides the doorman's job is "opening the door", replaces him with an automatic door and books the saving. What nobody measured was everything else the doorman did: the welcome, the security, the sense of occasion, the local knowledge. All of it disappears, and none of it was ever on the spreadsheet.
This is the biggest risk with AI rollouts. If you look at roles one function at a time, you'll automate the obvious task and quietly lose the value around it. Rory suggested bringing AI in around how teams actually work together, rather than bolting it onto individuals one job description at a time.
That matches our experience. Our workflow optimisation work starts by mapping a process end to end, with the people who do it, before we recommend anything. Sometimes AI and automation transform the process. Sometimes the answer is a simpler change nobody had spotted.
Value that never shows up on an invoice
The last idea that stuck with me was reverse benchmarking. Will Guidara, who ran Eleven Madison Park in New York, took his team to one of the most highly ranked restaurants in the world. Instead of asking what they should copy, he asked what had been a bit disappointing. The answer was the tea, coffee and beer. So he hired a beer specialist and a world-class barista, and his restaurant became known for exactly what everyone else had ignored.
In IT, the things that matter most are often the things nobody measures. The engineer who spots a problem before it becomes one. The advice that stops a client spending money on something they don't need. Those moments never appear as a line on an invoice, but they're the reason a partnership works.
What to ask before your next technology decision
Whether you work with us or anyone else, these questions are worth asking:
What outcome are we actually trying to achieve, and how will we know we've got there?
Are we asking the right question, or is there a cheaper way to solve the underlying problem?
What value might we lose that nobody is measuring?
Is our provider's success tied to ours, or just to what they sell us?
Will they still be around to make sure it works six months after go-live?
Outcomes first, technology second
Rory closed with a line borrowed from a mathematician: maths isn't there to tell you what to think, it's there to sense-check your dreams. Technology should work the same way. It should serve what you're trying to build, not decide it for you.
If a process is costing your team more time than it should, we're always happy to map it out with you. We'll give you a straight view on what would help, even if the answer isn't AI. Visit our website for more information.
Written by Bhavesh Shah - Director of Technology at BITS Group