By Andy Lang, Founder, Munny Group

Every now and again, somebody does a piece of work that makes you stop and think. The recent testing by Which? on how AI answers everyday money questions is one of those moments.

And before I go any further, let me say this clearly: What Which? has done is proper consumer‑protection work. No drama. No hype. Just a simple question: “If people ask AI about their money, how good are the answers?”

They didn’t test AI on trick questions. They tested it on the stuff people ask when life is happening:

“Do I need to fill out a tax return?”
“How much can I gift without paying inheritance tax?”
“How do I pay off my debt?”
“How does equity release work?”
“How do I get a council tax reduction?”

Real questions. Real consequences.

And the scores? Most models landed somewhere between 60% and 68% across accuracy, honesty, clarity and ethical responsibility.

Now, 60‑odd percent doesn’t sound terrible. But here’s the thing: money isn’t a pub quiz. Being 60% right about tax or debt isn’t “pretty good”. It’s risky.

Why these scores matter

If an AI gets a football fact wrong, nobody loses sleep. If it gets a tax rule wrong, somebody could lose money.

Money questions aren’t just questions. They’re crossroads.

When someone asks, “How do I pay off my debt?” they’re not looking for a clever answer. They’re looking for a safe one. Same with pensions. Same with mortgages. Same with equity release.

It’s a bit like asking your sat nav for directions abroad. It’ll try its best. But if the maps aren’t built for that country, you’re going to take a few wrong turns.

General‑purpose AI wasn’t built for regulated financial guidance. So when you test it on regulated financial guidance, you get regulated‑financial‑guidance‑shaped gaps.

Fair enough.

The HELM framework — and why it’s useful

One thing I really liked about the Which? testing is that it draws on the HELM framework from Stanford. HELM doesn’t just ask:

“Is the answer right?”

It asks:

Is it honest?
Is it clear?
Is it relevant?
Is it ethically responsible?

In money, those last two matter just as much as accuracy.

A perfectly accurate answer delivered without context can still send someone down the wrong path. A confident answer delivered without honesty can create false certainty.

HELM is basically a reminder that financial guidance isn’t just information, it’s responsibility.

Why we built eM differently

When we started building eM, our AI‑powered Money Coach, we didn’t try to create a chatbot that could answer everything. I’ve never seen the point in that. It’s like asking your mate who’s handy with DIY to fix your boiler. He’ll give it a go, but that’s not what he’s trained for.

We built eM for one purpose:

Help people feel clearer, calmer and more confident with their money.

That meant designing her differently from day one.

She doesn’t guess.
She doesn’t bluff.
She doesn’t invent rules.
She doesn’t give regulated advice.
She doesn’t push products.
She doesn’t chase engagement.

Instead, she uses structured workflows, guardrails, and FCA‑aligned principles to guide people through money questions safely.

She’s not trying to be clever. She’s trying to be helpful.

And yes, I’m proud of her. In hindsight, maybe we were a bit early building her this way. But I’d rather be early than wrong.

Could HELM be used to evaluate eM?

Absolutely.

Not as a competition. Not as a “who scored higher?” moment. But as a way to show what happens when an AI is built specifically for financial wellbeing rather than general conversation.

HELM rewards honesty, clarity, relevance and ethical responsibility. Those are the exact principles eM was built on.

So if Which? or anyone else would like to apply the same framework to eM, I’d welcome it. It would help show the difference between AI that answers financial questions and AI that supports people with their financial lives.

Let’s not scare people away from AI

There’s one thing I’m very conscious of.

If we make people afraid of AI, they won’t stop looking for answers. They’ll just go elsewhere.

And “elsewhere” often means:

TikTok noise
Reddit threads
Telegram “experts”
Influencers who know more about engagement than economics

People deserve better than that.

AI can be brilliant for financial wellbeing, when it’s built responsibly.

That’s why we built eM. That’s why Which?’s work matters. And that’s why Talk Money Week is important.

The bottom line

Money is part of everyday life. Kids. Cars. Holidays. Bills. Mistakes. Plans. All of it.

People don’t need perfection. They need clarity. They need honesty. They need confidence. They need to understand the numbers well enough to make their own decisions.

Which? has shown us the risks. HELM has shown us the standards. And eM is our contribution to a safer, kinder, more practical future for financial wellbeing.

If you’re using AI for money questions, just make sure it’s built for money. And if you’re not sure where to start, eM’s a good place.

AI, Money and Why Getting It Right Actually Matters