The Real Benefits of AI in Business: It's Not the Tools

Every shop in your area can buy the same CNC machine. Same make, same model, same tooling package. Two years later, one shop has doubled its throughput and the other uses it like an expensive drill press. Nobody blames the machine. The difference is that one shop rebuilt its workflow around what the machine could do.

A new experiment shows AI works exactly the same way. And it might be the clearest answer yet to the question “what are the benefits of AI in business?”

What did the experiment find?

Researchers at INSEAD and Harvard Business School ran a field experiment with 515 high-growth startups in a three-month AI accelerator (Kim, Kim & Koning, 2026). Every firm got the same package: API credits, access to frontier AI models and technical training.

Then they split the group at random. Half got case studies showing how other firms had reorganized their production around AI. The other half got workshops on general entrepreneurship. Same tools, same budget, different playbook.

The firms that learned to reorganize:

• found 44% more AI use cases, concentrated in product development and strategy;
• completed 12% more tasks;
• were 18% more likely to acquire paying customers (11 percentage points);
• generated 1.9x higher revenue;
• asked for 39.5% less outside capital, while their demand for labour stayed the same.

In the authors’ words, the key step “was not learning to use AI—all firms had the same tools and training—but recognizing where AI could reshape how value was created.”

So is it the tools or the way you use them?

It’s the way you use them. That’s the whole post in one line, but it’s worth sitting with.

Most businesses adopt AI like a new hand tool. They hand everyone a chatbot login and hope for the best. Someone uses it to tidy up an email. Someone else summarizes a PDF. The office feels a bit faster, but the shop’s output doesn’t move, because the work still flows through the same steps, the same handoffs and the same bottlenecks.

The researchers name this the “mapping problem”: the hard part isn’t getting AI, it’s discovering where and how it creates value across your whole process. A shop is only as fast as its slowest station. Speed up the welder and leave the fitter as-is, and you’ve built a bigger pile of parts in front of the fitter.

What does reorganizing around AI actually look like?

The paper gives real examples from the treated firms. Our favourite: a field services startup “rebuilt its full operations chain—from an intent classifier replacing a dispatcher to a payment-reminder engine replacing collections staff.”

Notice what they didn’t do. They didn’t ask “how can AI help our dispatcher type faster?” They asked “what does dispatching actually need?” and rebuilt the step.

Now picture it in a fab shop. The old way: an RFQ arrives, an estimator downloads the drawings, retypes the quantities, emails the GC for missing details and waits. The reorganized way: an AI agent reads the RFQ, pulls the takeoff, flags what’s missing, chases the GC and hands the estimator a draft quote. The estimator spends their time on the price and the risk, not the typing.

That’s the pattern: AI agents do the work that isn’t a decision. People do the deciding.

Do the gains show up for everyone?

No, and this is the part most summaries skip.

Many firms in the study had little or no revenue to begin with; at baseline the median firm’s monthly revenue was $0. The paper finds the revenue gains “small through most of the distribution but rise sharply in the upper tail, with the largest effects emerging at the 90th and 95th percentiles.” In plain terms, AI didn’t lift every boat a little. It let the best firms break out.

That’s a feature, not a bug. AI raises the ceiling. A great estimator with an AI agent doing the takeoffs doesn’t get 5% better; they can bid on jobs they used to pass on. A shop with sloppy processes just gets its mess delivered faster. Reorganizing comes first, and that’s why it matters so much.

One honest caveat: these were early-stage startups, not 40-year-old fab shops. But the lesson travels. A shop’s processes are older and more tangled, which means there’s more to untangle and more to gain.

Does AI mean fewer people?

Not in this study. Treated firms grew faster, but their demand for labour stayed the same, and they asked for less outside capital. They did more with the team they had.

That’s Jevons paradox applied to AI: when work gets cheaper to do, you don’t do less of it, you do more. More quotes out the door. More jobs tracked properly. More customers kept in the loop. The estimator who used to spend Tuesday retyping quantities now spends it winning the job.

What about the AI companies themselves?

The AI companies are growing fast too. At Stripe Sessions 2026, Stripe’s Maia Josebachvili said: “We saw the top AI companies grow by 120% in 2025. So far, in ‘26, by 175%” (Stripe).

To be clear, that’s companies selling AI, not shops using it. But it tells you where the money and the talent are flowing, and that the tools you can buy next year will be better and cheaper than today’s. The question isn’t whether the tools will be good enough. It’s whether your shop will be set up to use them.

Where should a manufacturer start reorganizing?

Don’t start with the tool. Start with the map.

1. Walk one workflow end to end, from the email that starts it to the invoice that ends it. Quote-to-cash is a good one.
2. Mark every step that isn’t a decision: reading, retyping, chasing, checking, filing. That’s the AI agent’s job.
3. Mark every step that is a decision: the price, the schedule call, the tricky detail. That’s your team’s job.
4. Set the limits. AI runs the routine work on its own within limits you set; people sign off where it matters, like quotes, big POs and anything unusual.
5. Rebuild the steps, not just the tasks. If the software you have can’t support the new flow, build the missing piece around it. Our guide to custom software vs off the shelf covers when that makes sense.

For the full playbook on scoring what to automate first, see how to automate your business with AI.

What’s the real difference between shops using AI and shops that aren’t?

In a few years, it won’t be “uses AI” versus “doesn’t.” Everyone will have the tools, the same way everyone has email. The difference will be shops that reorganized around AI and shops that bolted it on.

Picture the first kind. RFQs turn into draft quotes before the estimator’s coffee is cold. Mill certs match themselves to heat numbers. Customers get status updates without anyone writing them. AI agents hand work to each other and flag the exceptions. The team spends its day on the calls that need judgement and taste, and on the work they actually enjoy.

The other kind has a chatbot login nobody uses.

Want help mapping AI into your shop?

That’s exactly the work we do. We’ll walk through one workflow with you, find the steps AI agents can take over and set up workflow automation that fits how your shop runs. Book a free 30-minute consultation.

Sources

• Kim, H., Kim, D., & Koning, R. (2026). Mapping AI into Production: A Field Experiment on Firm Performance. INSEAD Working Paper No. 2026/20/STR. SSRN · Harvard Business School
• Stripe Sessions 2026, “Inside the AI economy: What Stripe’s data reveals.” Stripe

Frequently asked questions

What are the benefits of AI in business?

More work done with the same team, faster turnaround and fewer errors on repetitive computer work. In the INSEAD experiment, firms that reorganized around AI completed 12% more tasks and earned 1.9x the revenue, while asking for 39.5% less outside capital.

Does AI increase productivity?

Yes, but how you use it matters more than whether you have it. In the INSEAD study, every firm got the same tools and training; only the firms shown how to reorganize their work saw the gains in tasks, customers and revenue.

Why do some companies get more out of AI than others?

The researchers call it the mapping problem: finding where and how AI creates value across the whole operation. Firms that looked beyond one or two tasks found 44% more AI use cases, mostly in product development and strategy.

What are the benefits of AI in business operations?

In operations, AI agents take on the reading, typing and chasing: turning emails and drawings into jobs, updating schedules and sending status updates. People keep the decisions and sign off where it matters.

Does using AI mean hiring fewer people?

Not in this study. Firms that reorganized around AI grew faster, but their demand for labour stayed the same. The change was what each person could get done.