In its 2025 study, The GenAI Divide, MIT looked at 300 public enterprise AI deployments and found that 95% of them delivered no measurable business impact. Only one in twenty produced real value.

When I first read that, my instinct was the same as everyone’s: the models must not have been good enough. But the study is clear that this is wrong. The cause was not model quality. It was implementation. Generic tools bolted onto the side of real work, dazzling in a demo and useless in production a month later.

Lately, I have been reading everything I can find about one specific kind of enterprise, the regulated kind (also because I earn my bread through this): banks, fintechs, and payment processors. And the more I read, the more I think the 95% number hides a second, harder problem that almost nobody is talking about out loud.

Generic AI is built to always answer

Think about how a general-purpose model behaves. It is built to always produce something. Ask it a question it has never seen, and it will not stop and say, “I do not know this one.” It fills the gap with something that sounds right.

In most of life, that is a fine trade. A slightly wrong answer is a minor annoyance, and you move on.

Now put that same behaviour inside a bank.

A confident wrong answer about a regulatory deadline is not a typo. It is a fine. A misread threshold on a suspicious-transaction alert is not an inconvenience. It is a report that should have been filed and was not. A document that looks like it passes a check but quietly does not is not a rounding error. It is real money and real liability.

The work these institutions run on has a particular character that makes guessing intolerable:

• The rules are published: Scheme rulebooks, central-bank timelines, verification requirements, and compliance thresholds. There is a correct answer, and it is written down somewhere.

• The work is language-heavy and repetitive: Read a document, check it against a rule, decide, write the response, record why. Thousands of times over.

• Every case has money and liability attached, which is exactly why nobody has been willing to let a generic chatbot anywhere near it.

This is high-stakes but low-judgment work. It follows the published rules almost every single time. And it is still done by hand, at enormous cost, because getting it wrong is expensive, and a general model cannot be trusted not to invent its way through the parts it does not know.

So the question changes

Here is the shift that I believe most of the industry has yet to make.

Everyone is asking, “Can AI do this work?” And technically, the answer is mostly yes. The models are capable of reading the document and drafting the response.

But that is not the question the person who actually has to approve it is asking. The compliance officer, the risk lead, the person whose name goes on the outcome, is asking something else entirely:

Those are not the same question. And I have come to think the gap between them is the whole story. It is why 95% of pilots die. It is why the impressive demo never becomes the thing that runs on Monday morning. The demo answers the first question. Nobody in the building can answer the second.

The uncomfortable part

If the problem were just “the models need to get smarter,” it would solve itself. Every lab on earth is working on exactly that, and the models do get better every few months.

But I do not think that is the problem. A smarter model that still guesses is still a model that guesses. Making it more fluent does not make a risk team trust it with a filing. If anything, a more confident wrong answer is more dangerous, because it is harder to catch.

Which leaves a genuinely interesting question, and it is the one I have been sitting with:

If the fix is not a smarter model, then what is it? What would AI have to actually be, structurally, before the person carrying the liability would put their name on its output? What has to be true about how it works, not how clever it is, before “interesting demo” becomes “we can run this”?

I have some strong opinions forming on this, from following how this work actually happens and where the published rules leave off. There is a shape to the answer, and it has very little to do with the things the AI headlines are about.

I am going to write about it properly in the next few pieces: what “trustworthy” has to mean when a wrong answer is a fine, why the deployment question (”where does our data even go?”) quietly kills more projects than the technology ever does, and why the same boring, rule-bound work shows up in department after department wearing a different uniform each time.

If that is a thread you want to follow, stick around. Follow along here, and I will take it apart one piece at a time.

The models are not the hard part. The hard part is everything the demos skip.

Some of my other articles:

Let’s break this down without overthinking it.

1. The first drop-off: Documents

Most small merchants already don’t know the difference between TRN, MoA, license activities, or why a simple online business needs so many papers.

If they can’t understand why something is needed, they don’t submit it.

2. The long wait: Compliance back-and-forth

This is where most deals die.

Merchants get an email:

“Your document is blurry. Please re-upload.”

“We need additional info.”

“We are reviewing.”

No timelines | No clarity.

And after 2–3 loops, the intent is gone.

3. “What’s the status?”

Every PSP has this problem - the merchant submits everything but still has no idea: who’s reviewing | how long it’ll take | what’s pending |what’s approved

When the process is opaque, the merchant mentally checks out.

4. Integration friction

Even after approval, a lot of merchants drop off here.

Why? unclear API docs | missing examples | SDKs that don’t match real use cases | support that replies after 24 hours

It’s not that merchants don’t want to accept payments - they just don’t want to feel lost.

5. Risk rules that reject without explaining

If you block a category, say it.
If the merchant doesn’t qualify, explain it.

Silence = frustration.
Frustration = churn.

The reality

Most PSPs don’t lose merchants to competition.
They lose them to confusion, slow loops, and zero visibility.
Fix onboarding, and you’ll see growth jump without touching pricing, marketing, or sales.

🛍️ And for merchants in the UAE

If you’re facing:
- slow approvals
- unclear requirements
- inconsistent support
- or issues going live with your payment provider

There’s a simpler option.

➡️ Switch to Comera Pay

Fast onboarding, clean flow, predictable timelines.

Because accepting payments shouldn’t feel like opening a bank account in 1998.

And if you still face any issues, you know who to reach out to!

When you travel abroad and swipe your card, sometimes the POS asks if you’d like to pay in your home currency instead of the local one. That’s what’s called Dynamic Currency Conversion (DCC).

Take an American tourist in Italy, buying shoes for €100.

The system might offer:

Pay €100, and let your bank handle the conversion later.

Or pay ~$120 right away, converted on the spot with DCC fees built in.

For the customer, the benefit is clarity — they immediately know how much they’ll be charged in their own currency.

For merchants and banks, DCC can mean additional revenue from the conversion spread.

What happens in the background is simple but clever.

1. The POS or online checkout detects that your card is foreign.

2. It then gives you the option to transact in your own currency.

3. If you accept, the transaction doesn’t just flow through the acquiring bank and card network like usual, it goes via a DCC provider, who handles the conversion and ensures your card is billed in your home currency.

For a merchant who wants to offer this, the setup isn’t with Mastercard or Visa directly. You’d typically go through your acquiring bank or payment processor, who can enable DCC on your terminals or e-commerce checkout.

With global travel and cross-border payments growing rapidly, tools like DCC are becoming increasingly relevant — even if customers still debate whether it’s better to accept or decline.

Image From: www.fintechwrapup.com

🎬 Netflix vs Amazon: Netflix had profits but negative FCF for years due to content spend. Amazon’s thin profits hide strong cash generation thanks to AWS and efficient operations.

💡 How You Can Use FCF:

1. Track it over the years - consistent growth is a good sign
2. Compare FCF yield (FCF / Market Cap)
3. Be cautious if profit rises but FCF falls

Think of it like personal finance:

💵 Salary = Revenue

💳 Bills & EMI = Expenses

💰 Savings = Free Cash Flow

That’s why investors like Warren Buffett love businesses with strong FCF — it shows real strength and staying power.

Next time you look at a company, ask: What’s their Free Cash Flow telling me?

It’s a win-win:

Buyers preserve their cash and optimise working capital

Suppliers get immediate liquidity and stay stable

4. How SCF Complements B2B Payments

Think of B2B payments as the rails, and SCF as the train carrying value across those rails.

Without digital, secure, and reliable payment systems, SCF programs can’t operate effectively.

Now, modern platforms are embedding financing within payment flows -Suppliers can opt for early payment → financing happens automatically.

That’s embedded SCF, and it’s where the future lies.

5. The Future: Real-Time and Data-Driven

Over the next 5 years, expect:

- Real-time payments + AI-driven credit models to assess risk instantly
- Fintech and bank partnerships powering embedded finance within ERPs
- Cross-border SCF solutions that eliminate FX friction

Together, B2B + SCF are creating an ecosystem where cash moves smarter, not just faster.

💭 If you’re building or working in payments, it’s worth understanding this bridge between how money moves and how liquidity flows.

That’s where real innovation (and value creation) happens.