The Hidden Investments Behind AI Success
Lots of firms “try AI” and end up disappointed. They buy a tool, plug it in, and hope for magic. A few weeks later, little has changed.
Here’s the truth: AI isn’t plug-and-play. The big gains come when you change the system around it — the way people work, how decisions are made, even what you sell. Economists call this idea complementarities: when the right things move together, the payoff multiplies.
Complementarities: when 2 + 2 = 5
In the 1990s, Paul Milgrom and John Roberts studied modern manufacturing. Companies were buying flexible new machinery — but only those that also re-organised their production (just-in-time, new team structures, better training) saw real performance gains.
The message? Each change was modest on its own. Combined, they were explosive. That’s what complementarities mean: doing one thing helps a bit, but doing them together makes the whole worth more than the sum of the parts.
AI works the same way. A chatbot or dashboard added on its own will barely move the needle. Combine it with redesigned roles, cleaner data, and new incentives, and suddenly you get that “2 + 2 = 5” effect.
Lessons from the 90s IT Boom
Timothy Bresnahan, Erik Brynjolfsson and Lorin Hitt looked at the IT boom of the 1990s, a period when some firms leveraged technology to pull ahead while others were left behind. The difference wasn’t just who bought computers — it was who moved three pieces together:
Information technology — the new systems and digital tools.
Workplace reorganisation — job redesign, teams with more autonomy, decisions pushed closer to the work.
Product and service innovation — using IT not just to cut costs, but to improve what you already deliver and to create entirely new offers.
Firms that pulled all three levers grew faster and demanded more skilled staff. Those that only bought tech… didn’t.
The lesson for today is clear: AI is another general-purpose technology. To make it pay, you need to modernise how you work and what you deliver, not just buy software.
Why Your Best Assets Look Like Costs
When you buy something physical — a laptop, a server, a van — accounting treats it as an asset: it goes on the balance sheet and its cost is spread over years.
But the real work that makes new technology pay — training, process redesign, data cleanup, better routines, documented workflows — is usually booked as an expense. Profit takes a short-term hit even though these activities build capabilities that last.
Economists call this intangible capital. And because these intangibles are often undercounted in company accounts, they’re easy to underinvest in. The firms that treat them as genuine investments — whether by training their own people, documenting new workflows, or bringing in partners to help accelerate the build-out — are the ones that turn tools into durable advantage and long-term gains.
Evidence backs this up. One notable study found every $1 of computer equipment was linked to about $12 of market value, versus only $1 to $1.50 for traditional assets. The gap reflects the unmeasured organisational build — training, new processes, and better ways of working — that makes technology pay.
These benefits also take time. Early on, costs show while results are still bedding in. Later, performance climbs as skills, routines and data discipline click into place — the familiar productivity J-curve.
AI is like signing a star striker. Without the midfield, the tactics, and the training, they won’t score many goals. The supporting system is what unlocks their potential.
What This Means for Your Business
If you treat AI as a tool purchase, you’ll get tool-sized results. Treat it as a bundle of changes, and the effects multiply.
Five takeaways for business leaders:
Plan a bundle, not a bolt-on. Pick one workflow (e.g. invoice-to-cash) and change it completely — tool + process + roles + training + KPIs.
Shift decisions to the data. Empower frontline teams where the information lives, with simple guardrails.
Look outward too. Ask how AI enables better offers for customers, not just internal savings.
Budget for the invisible. Training, process work and data hygiene aren’t extras — they’re the core investment.
Expect the lag. The J-curve is real. Track leading indicators (cycle time, error rates, adoption) before the profit shows up.
AI rewards the firms that learn how to work differently. Change the tech and the system around it, and you don’t just add a good player — you build a winning team.
