Can AI Sell an Experience?

Part 2 of 3

In Part 1, I wrote about the agentic presence problem: the challenge facing service SMEs who need to exist in AI systems before they can ever be discovered, recommended or chosen.

But let's say we solve that.

Let's say every physio clinic, every family restaurant, every independent hairdresser has clean, accurate, machine-readable information that AI agents can find and trust. Presence problem solved.

Does that mean an AI agent can sell what they offer? I'm not sure it does.

There's a version of agentic commerce that works beautifully for products.

You ask your AI assistant to find the best noise-cancelling headphones under $200. It searches, compares, reads reviews, checks your purchase history and places the order. You never open a browser. The transaction is clean, fast and probably better than what you would have done yourself.

If it gets it wrong, you file a return and drop them in a locker. Now ask it to find you a good barber.

An AI agent can find barbers near you. It can filter by rating, price and availability. It can book the appointment.

What it cannot do is tell you whether you'll trust the person holding the scissors.

I know this firsthand. I've been living in Vietnam for the past nine months, and I've been on a relentless hunt for a hair stylist who can do a proper cut using only scissors. I've done all the Googling. I've fallen down TikTok rabbit holes. I've translated reviews.

The algorithms did their job perfectly. They got me into the chairs.

Over nine months, averaging two haircuts a month, I have sat in 18 different stylist chairs. I have not returned to a single one of them a second time.

Eighteen initial transactions. Zero retained trust.

This is the tension at the heart of agentic commerce for service businesses.

Products can be evaluated on specs. Experiences have to be felt. A good meal isn't just ingredients and a rating. A great physio isn't just someone with a license and an open slot at 2pm. It's the person whose hands you trust to fix your neck.

You can't verify that in a schema markup.

And there's an asymmetry of risk here that matters. If an AI orders you the wrong headphones, you file a return. If it books you a surgeon who doesn't listen or a therapist who isn't right, the cost of that failure is written on your body or your wellbeing. The higher the stakes of the experience, the less comfortable we are delegating the decision entirely to a machine.

For services, there is an un-automatable step between discovery and transaction: vulnerability. You have to let a stranger touch your hair, adjust your spine or cook your food. No API can establish that comfort level for you.

Trust has to be built in person.

So where does that leave service SMEs?

I don't think the answer is that agentic commerce doesn't apply to them. I think it applies differently.

AI agents will almost certainly become the primary way new customers discover local service businesses. Not word of mouth, not Google, not walking past a storefront. An agent that knows your preferences, your location, your schedule and your past behavior will surface the right physio or the right restaurant at the right moment better than any search bar ever did.

That's a genuine opportunity for service SMEs who show up in that layer correctly.

But the transaction still ends at the door. The agent books the table. The meal earns the next booking. The agent schedules the consultation. The physio earns the relationship.

Discovery becomes agentic. Trust remains human.

This creates an interesting inversion.

For product businesses, agentic commerce compresses the entire customer journey. Discovery, evaluation, purchase, all delegated. The human is almost removed from the loop. The business that wins is the one with the cleanest data and the lowest friction.

For service businesses, agentic commerce hands off to the human at exactly the moment it matters most. The agent does the finding. The business owner does the convincing.

That's not a limitation. That's actually the point.

The haircut, the meal, the adjustment after the physio session: these are not problems to be automated. They are the product.

The agent's job is to get the customer through the door with enough confidence to try. Everything after that belongs to the business owner.

In a world where AI is intermediating more and more of commerce, service SMEs may turn out to be the last businesses where the human relationship is structurally protected. Not because AI can't reach them. But because the value they deliver begins precisely where AI's role ends.

Which brings us back to where Part 1 left off.

If AI agents are going to become the primary discovery channel for local service businesses, then the quality of information those agents have access to becomes the single most important factor in whether a business gets found at all. Not the best ad. Not the loudest social presence. Not the most reviews.

The cleanest, most accurate, most trustworthy data.

A business that an AI agent can confidently understand and recommend will get the booking. A business that is missing, wrong or inconsistent in the data layer will not. My 18 stylists all existed. Most of them were good. But the ones I never found weren't failing at their craft. They were failing at being findable.

The agent gets customers to the door. But first, it has to know the door exists.

In Part 3, I want to look at what happens to service SMEs who don't solve that problem. Not as a warning. As a practical question about what the next few years actually look like for the majority of small businesses on the planet.

This is Part 2 of a three-part series on what agentic commerce means for service SMEs, the businesses that sell experiences, not products.