If You Can't Do the Math, Don't Jump into AI Market Intelligence Yet

A couple of days ago, a friend of mine who works in foreign trade took me out to dinner.

Halfway through the meal, he suddenly asked me: Is this AI market intelligence thing actually worth it? Our company is already using it, but the moment anyone asks one more time, "You spent the money — so what did you actually get?", I'm at a loss for words.

I said, don't ask me. Let me ask you three questions.

After you started using AI, how much faster do you make a market decision than before?

How much money did that saved time earn you?

And if it didn't give you any of that, how much have you missed?

He looked at me, thought for a long time, and finally said: I've never actually added it up.

Good grief.

That only confirmed what I already believed:

Don't rush to argue about whether AI is good or bad. Ask yourself first whether you can actually add up what it's worth to you.

My friend isn't an exception. I've met too many owners who don't flinch when they buy AI, but who all fall silent the moment ROI comes up.

But today this question can't be dodged. The market having been speeded up by AI is nothing new. If you don't see it coming early, you'll only be trailing the crowd, eating the fruit that spoils first.

What Market Intelligence Means

What is market intelligence? In plain words: you know something before the whole market does.

You can't get that from the traditional way of doing things.

How does the traditional approach work? Send out surveys, lurk in forums, dig through historical reports — then sit through a month of meetings and put out an 80-page deck. By the time you've read it all, the market has shifted again.

"Explaining what happened after the fact" is the doom that traditional research can't escape. No matter how fast you are, you're still driving by looking in the rearview mirror.

AI, on the other hand, never gets tired:

• It swallows billions of pieces of information from across the web in seconds;
• It turns up patterns of correlation that the human eye can't see;
• It doesn't tell you "what happened" — it tells you "what's about to happen";
• It watches 24/7 without eating or sleeping.

The moment the supply chain hiccups, a policy shifts, or consumers quietly change their tastes overnight, getting those signals one step earlier is a whole business quarter of advantage.

Splitting the ROI Ledger into Three Books

What is ROI? Simply, you put in one dollar and you get back a few.

But in the AI market intelligence game, ROI can't be a single line. I generally split it into three books.

Book One: More Money Earned

Four openings here.

Finding new markets. A food company had AI scan consumer data across the web and surfaced a cluster of people "who wanted to buy plant-based pet food in the city." It charged into that market and did an extra £5 million of business its first year. Those people were tucked away in a corner you would never have thought to look.

Improving products. An electronics factory fed its returns and complaints into AI and had it pinpoint where failures clustered. After one round of fixes, product failure dropped 15% and customer satisfaction rose 10%.

Dynamic pricing. A SaaS company let AI adjust prices by demand elasticity, and its average revenue per user (ARPU) rose 7%.

Precision targeting. A retail brand split its audiences into fine slices, and its marketing ROI climbed 20%.

Notice — none of that money was ever on your records. AI's first book is exactly about drawing out money like that.

The Second Book: Money Saved

This one has two angles, and both are quite handsome.

Saving on research labor and time. A global consulting firm used AI to automate and cut the cycle time of a research engagement by 40% and the labor cost by 25%.

Avoiding pitfalls, the most handsome of all. An automotive-components supplier had AI predict, six months ahead, that raw materials were about to rise. It stockpiled early and saved £2 million in a year.

How would you have lost that money otherwise? You would have lost it because you "couldn't see it." AI's value is precisely turning "couldn't see it" into "saw it in time and dodged it."

Book Three: The One You Can't See

This book stays off the financial statement, but it's the one with the biggest picture.

Decisions get faster. The others are still assembling decks and calling meetings when you've sealed the decision on the spot. Speed may be hard to quantify — but three years down the line, it's the difference between first place and third.

Innovation keeps coming. If AI keeps feeding you new leads every day, your product team's new ideas never run dry, and the frequency of your launches beats the competitor.

The brand stands stronger. AI can smell a reputation about to turn sour in advance and deal with it before it blows up. Customer satisfaction, the Net Promoter Score, brand image — things that once could only be "felt" can now become numbers.

How do you keep this book up? Time-to-market, how many new products you ship each season, Customer Satisfaction (CSAT), Net Promoter Score (NPS), brand awareness, risk exposure. All of it is quantifiable, if you want it to be.

And there's one more "invisible" thing you probably hadn't thought of.

Market intelligence is how you "see AI." So what about the reverse? Do you want AI to "see" you?

People no longer open the search box the way they used to. They go straight to asking the AI. If your name doesn't come up in AI's answers, you're missing most of the market. That road already has a dedicated lane, called Answer Engine Optimization (AEO). Early players in that lane, like UltraScout AI, exist precisely to keep an eye out for you: whether your brand is mentioned or passed over, and where it lands, across the major AI platforms.

So What Number Does This Ledger Usually Land On?

I've done the math with a few companies that moved early. Over one to three years, the return basically lands between 150% and 400%. How high the ceiling goes depends on how finely you've broken down the books. The finer your books, the less the numbers can fool you.

You could put it that way: the ROI of AI market intelligence isn't calculated — it's designed.

Doing It on the Ground, in Five Steps

Enough theory. How do we make it real? Here are five steps — don't skip any of them.

Step one: write your goal in specifics. Don't write "improve efficiency," the bare phrase. Write "cut research costs by 20%," "bring new products to market 10% faster," "gain 3% share in a region." If you can't write this out loud, you haven't clearly thought about what you'd be buying it for.

Step two: keep a baseline. Before you put it to use, write down the numbers where they currently stand. No baseline, no frame of comparison — and you'll never be able to prove whose work the win was.

Step three: connect your data first. What AI fears most is dirty data and data silos. Do the data governance and data integration first, and only then talk about the intelligence — otherwise the AI becomes a mute that can't talk.

Step four: watch the numbers, track them constantly. If AI tells you there's an opportunity somewhere, commit a small tranche of working capital to the pilot and keep a separate ledger for that pocket of business. Where it's right, scale up; where it's wrong, adjust.

Step five: review quarterly, iterate over the long term. ROI is never calculated once and then done. Recompute it every quarter, swap out models, refill data sources, revise goals. The more often you compute, the more accurate the number, and the more the boss trusts you.

Three Stories from the Field

We've covered the theory; here are three true ones.

Story one: a fashion chain's inventory pain. Fashion is the industry that fears being squeezed from both sides — order too much and you sit on dead stock, order too little and you run out of supply. This global fashion retailer was already burning about 8% of its profits every year on the inventory wars alone. It fed historical sales, social-trend momentum, weather, and competitors' promotions into AI, driving its forecast accuracy up to 95%. Result: inventory carrying costs fell 18%, the number of discount runs dropped 10%, and sales went up 5% anyway. First-year ROI: 250%.

Wow.

Story two: B2B SaaS first-mover. When you work the B2B side, the thing you fear most is a competitor quietly developing in the shadows — by the time you notice it, its launches have already taken off. This company had an AI scan industry news, patents, funding announcements, and tech forums day and night, and it dug out two "future competitors" that hadn't even formed yet, nine to twelve months before any human team would have. Its product roadmap and go-to-market battle plan were revised before the competitors could get large. Result: it managed to protect 3% more market share than its original plan had allowed for.

Story three: a pharma company, the ascetic race against time. New-drug development typically runs on a ten-year cycle and costs billions. This pharma firm handed AI all of its scientific literature and data, plus the way patients discuss their condition, and let it help circle "targets that were worth dropping down the lift." It cut 30% of the time needed to find a viable next candidate. On every successful candidate drug, the early stage alone can save about £15 million.

Three industries, three moves. All follow the same logic in the end: ultimately AI's true role is to win you time. And time, in business, is money.

The Four Pitfalls, and Don't Miss a One

Numbers can be fun to watch; the pitfalls deserve an early look. These are the four that eat up your ROI.

Pitfall one: dirty data. Dirty data goes into AI, and polished nonsense comes out. Data governance is the foundation beneath the foundation.

Pitfall two: nobody can use it. A tool might cost whatever you like — if nobody in the organization can read and call it, it's just a showpiece that can calculate. Training is worth more than the purchase.

Pitfall three: department silos. If each department stores its own piece and marketing and product each go their own way, AI insights never flow out, and the value decays behind the department walls.

Pitfall four: attribution disputes. The money earned — did it come from AI, or from the business team merely getting lucky? If no one can say for sure, the boss will never trust what your ROI really equals.

So how do you break through?

Begin with a pilot. Move fast and small, pick one small problem and earn a small "ROI" number first, win the organization's trust through a small victory, then wrap it outward step by step.

Put the training first. Put the people who truly use AI, the ones who understand the business, in charge of running it. Don't leave the tool sitting on the showroom shelf.

Tear down the department walls. Put intelligence, marketing, R&D, and sales at the same table, so the insight can travel across departments.

Write good attribution. Use a control group, A/B tests, and econometric models to give every conclusion evidence.

Back to That Dinner

What AI is truly worth is not "shaving the procurement bill," but "keeping you from going down the wrong road and handing you the secret that lets you get the jump on the competition."

The cost of AI was never that procurement bill.

It's the murky ledger you've been keeping.

So my advice is only said once: don't rush to buy. Take two months to dig through the decisions you made over the past year, and turn the time you've lost and the tuition you've paid into a single page of Excel.

That page of Excel is where your baseline lives. With the baseline standing, "whether to buy AI" is no longer a question of faith — it's a problem you can solve with arithmetic.

Here's hoping you did the math right.