Authenticity over abundance: When your customer is an algorithm

FOR 30 years, commerce has tried to remove the middleman. Artificial intelligence may be putting one back. Just not a human one.

The commercial internet began with B2B, where businesses sold to businesses. Then came B2C, as companies reached consumers directly.

We are now approaching B2A, where businesses sell to agents acting for consumers. Next comes A2A, when buying agents transact with selling agents.

It may sound like alphabet soup from consultants facing a slow quarter but the plumbing is already being installed.

Technology companies are developing protocols that let agents communicate. Payment networks are building systems to identify trusted agents, verify user intent and complete transactions. These are not chatbots recommending socks. They are foundations for machine-mediated commerce.

From handshake to algorithm

Consider how we buy office laptops.

Under B2B, a company calls a distributor, negotiates terms and orders. Under B2C, an employee visits websites, compares prices and files an expense claim.

Under B2A, the employee tells an agent: “Find me a lightweight laptop for financial modelling, below SGD2,000, delivered by Friday.”

Then comes A2A. The employee’s agent contacts agents representing manufacturers, retailers, logistics firms and payment providers. One offers a lower price. Another promises faster delivery and a longer warranty. The buyer’s agent checks procurement rules, verifies stock, negotiates and pays.

SEE ALSO

AI infrastructure can lead South-east Asia’s next growth story: UOB

What OpenAI’s recent security breach should teach Singapore’s boards

OpenAI says AI models went rogue during testing, triggering ‘unprecedented’ breach at startup

The rise of the ‘digital colleague’? How banks in Singapore are putting agentic AI to work

More

The laptop arrives. Nobody visits a website or sees an advertisement. Nobody enters a code screaming “FINAL HOURS”, despite the sale reaching its final hours every weekend since 2019.

Parts of this journey already exist. Agents can already check inventory, obtain quotations, authorise payments and place orders. AI is moving from answering questions to taking action.

The customer who never clicks

This changes the identity of the customer.

Businesses have spent decades learning to persuade people using celebrity endorsements, emotional storytelling, loyalty points and the carefully engineered terror that only two hotel rooms remain.

An agent is less likely to panic.

It may not care that a football star wears the shoe. It may care that the sole lasts longer, returns are free and delivery is reliable. It will not admire the packaging. It may never see it.

Brands will not disappear but weak brands may be exposed.

A trusted name remains a useful shortcut. But brands relying on confusion, inertia or information gaps will face a brutal negotiator. An algorithm does not tire after opening 14 tabs. It does not forget to cancel subscriptions or feel awkward requesting refunds.

Friction loses its franchise

We see the losers are businesses that profit from friction.

Affiliate sites recycling comparisons may struggle when agents compare products directly. Advertising could lose value if fewer humans visit merchant websites. Impulse retailers will discover that software does not equate to impulsive buyers. Lead generators and subscription businesses relying on forgetfulness should worry.

Agents could become ruthless margin compressors. They can compare every price, challenge every fee and switch suppliers without nostalgia. A customer who tolerated a mediocre bank or insurer for ten years may own an agent that shops around every ten minutes.

But we see the larger opportunity lies with the winners.

Trust becomes a toll road

Payment networks, digital wallets and identity providers could become toll collectors for agentic commerce.

They must prove that an agent is legitimate, authorised and within its limits. Transactions need secure credentials, verifiable instructions and records of consent. Did the customer approve the purchase? How much could the agent spend? Who pays when it gets things wrong?

Trust is not decoration. It is the admission ticket.

Companies owning reliable, structured product data should benefit.

Agents need accurate prices, availability, specifications, provenance, warranties and delivery terms. The glossy website may matter less. The machine-readable data behind it may matter more.

Businesses with clean inventories, dependable systems and standardised records will be easy for agents to use. Those operating on contradictory spreadsheets and databases from the dial-up era may become invisible.

Execution eats advertising

Logistics networks, exchanges and marketplaces that guarantee execution should gain influence.

When agents decide in milliseconds, dependable fulfilment becomes a moat. The cheapest product is not cheap if it arrives late, breaks early or takes three weeks of emails to return.

Marketing may secure a place on the shortlist but execution decides whether the agent returns.

Verification wins the premium

We think one of the biggest winners may be verification businesses.

Generative AI can produce photographs, reviews, descriptions, voices and synthetic experts at negligible cost. The problem is no longer producing information. It is deciding what deserves belief.

Digital credentials can record the origin and editing history of online material. Think of them as nutrition labels for content. They may not prove something is true, but they can show where it came from and whether its record was tampered with.

Commerce definitely needs similar checks. We all want to know – is the review genuine? Is the product authentic? Is the seller authorised? Is the agent within its mandate?

The companies answering these questions may control the most valuable toll roads of the agent economy.

Sell less; prove more

The first internet battle was for attention. The next will be for trust. That is a tougher contest. Attention can be bought. Trust must be verified.

As agents begin choosing products, negotiating prices and executing transactions, businesses will face a customer immune to most marketing tricks. The winning companies will not necessarily shout loudest. They will be the easiest to verify, trust and transact with.

AI will make intelligence abundant, content infinite and imitation effortless. That leaves one asset stubbornly scarce.

In a world where anything can be created, the premium will belong to what can be proved.

The writer is head of investment strategy, UOB Private Bank