Apparently, AI can do my shopping now.

Excellent news.

I've been waiting for technology to tackle the really important problems in life. Like remembering I've run out of coffee, finding a reasonably priced pair of trainers and explaining why the supermarket has decided that six strawberries are worth £4.50.

Google has announced new capabilities for Gemini Tasks, its AI assistant that can do more than answer questions.

It can work through tasks, keep an eye on prices and, with its new shopping features, help buy things on your behalf.

Actual shopping. With actual money.

Although the rollout is starting in selected countries, with some of the cleverer features arriving in America first.

Naturally.

The Americans get an AI personal assistant. Here in Britain, we're still trying to get a GP appointment before Christmas.

But putting my geographical jealousy aside, this is genuinely interesting.

Because we're moving beyond AI that tells us what to do.

We're entering the age of AI that actually does it.

And I'm not entirely sure we're ready.

First of all, I absolutely want this

Let's establish something.

I am not one of those people who thinks every technological advancement is a sign of civilisation collapsing.

Quite the opposite.

If AI wants to handle my boring admin, compare prices, organise appointments and remind me that I've forgotten someone's birthday, I'm listening.

Very enthusiastically, actually.

Imagine telling your assistant:

"Find me a decent pair of black boots, under £100, preferably on sale, and don't show me anything that looks like it belongs in a Victorian workhouse."

And it just... does it.

No scrolling through seventeen websites. No suspiciously personalised adverts following you around Instagram for the next six months.

Just a useful result.

Beautiful.

In theory.

Because there's a rather important difference between an assistant recommending something and an assistant making decisions on your behalf.

And that difference usually involves your bank account.

My AI has expensive taste. Apparently.

Now, to be fair, Google isn't suggesting Gemini should have unlimited access to your money and wander off on a shopping spree.

Its purchasing features involve safeguards, spending limits and user direction.

Which is reassuring.

But it does raise some interesting questions about how these systems will work as they become more independent.

Let's say I ask my AI assistant to find some new trainers.

It finds three options.

One is £65. Perfectly sensible.

One is £110. A little expensive, but lovely.

And one is £240, apparently designed using technology developed by NASA.

Which does my assistant recommend?

The sensible pair?

The fashionable pair?

Or the pair that happens to be promoted by whichever company has the best commercial relationship with the platform?

I'm not suggesting that's what's happening here.

I'm asking how we'll know the difference.

A personal assistant should work for me, not quietly become the world's most efficient salesperson.

We already live in a world where algorithms influence what we watch, buy, read and believe.

Giving those algorithms the ability to act on our behalf feels like a rather significant next step.

Especially when I struggle to trust the self-checkout machine at Tesco.

Unexpected item in bagging area.

Yes, it's the item I literally just scanned, Sandra.

Convenience is lovely. Trust is complicated.

Here's what I find fascinating.

For years, we've been happy to let technology make recommendations.

Netflix suggests what we watch.

Spotify decides what we'll listen to.

Amazon seems to know when we're thinking about buying something before we've even admitted it to ourselves.

But recommendations are one thing.

Actions are another.

What happens when your AI books the wrong appointment?

Orders something you didn't really want?

Misunderstands an instruction and confidently does exactly the wrong thing?

And who takes responsibility?

The company that built it? The retailer? Or you, because technically you gave it permission?

There's also the slightly uncomfortable question of privacy.

To be genuinely useful, an AI assistant needs context.

Your preferences. Your schedule. Your purchases. Perhaps your emails, messages and financial information.

The more it knows, the more helpful it can become.

But the more it knows, the more important it becomes to understand what information you're sharing, where it goes and who controls it.

I'm all for convenience.

I'm considerably less enthusiastic about giving a digital assistant the keys to my entire life without knowing where the spare set is kept.

The real question isn't whether AI can do it

I think we're asking the wrong question.

Every time a company announces a new AI capability, the conversation seems to be about what the technology can do.

Can it shop?

Can it book?

Can it negotiate?

Can it organise your entire existence while you sit on the sofa watching rubbish television?

Probably, eventually.

But just because something can be automated doesn't automatically mean it should be.

I'd happily let AI compare insurance quotes.

I'd be rather less comfortable letting it choose my next relationship.

Although, given some people's dating histories, perhaps an algorithm couldn't do much worse.

Sorry. That was unkind.

Slightly.

The point is that having an AI assistant shouldn't mean surrendering every decision.

Sometimes making choices is part of being human.

Even the silly ones.

Especially the silly ones.

I want an assistant, not a replacement brain

Despite all my questions, I'm genuinely excited about where this is going.

The idea of AI handling repetitive jobs, navigating complicated systems and saving people hours of their lives is brilliant.

For busy parents, people juggling multiple jobs, or anyone drowning in everyday admin, this technology could make a real difference.

And that's worth celebrating.

But I think the best AI assistants will be the ones that know when to act, when to ask and when to leave us alone.

They should be transparent about their decisions, clear about their limitations and very good at understanding the word "no".

A quality surprisingly absent from some human salespeople, now that I think about it.

I don't want AI to make every decision for me.

I want it to give me more time and freedom to make the decisions that actually matter.

The smartest assistant isn't necessarily the one that does everything. It's the one that knows what it shouldn't do.

So yes, Google. I'm interested.

Bring on the AI shopping assistants.

Let them find the discounts, compare the prices and remember that I've run out of coffee.

Just perhaps ask before ordering twelve pairs of shoes.

And if my AI ever develops a taste for designer handbags, I'm blaming the algorithm.

Although, knowing my luck, it'll order me a £900 handbag and congratulate itself for finding free delivery.

Priorities, apparently.