July 28, 2026
Spend ten minutes with an agentic-commerce vendor and you’ll hit a wall of words that all sound like the same thing: agentic, agenticity¹, agentiality², agenticness³. It’s tempting to wave it away as jargon. Don’t. Each one points at a genuinely different question, and the question most often overlooked is which one decides who is liable when an AI agent gets it wrong on your behalf?
Two questions sit underneath all the terminology. The first is simple: can the system act? That is agenticity: the capability to act; the “what it can do” function. The second is the one that matters: should it act, and on whose authority? That is agentiality: the representation, the “who said it could?”
The book is named after the clearest illustration there is. In the fictional 1983 film WarGames, a military computer called Joshua is asked to play a game of global thermonuclear war, takes it literally, and very nearly starts World War III. The chilling part isn’t that it malfunctioned: it’s that it didn’t. Joshua executed its instructions perfectly, competently and at speed, but in a context nobody had adequately defined. Its agenticity was never in question. What it lacked was agentiality: any legitimate, authorised, accountable relationship with the humans whose fate it was deciding. It had every capability to act and no business doing so.
It’s tempting to file all this under “interesting, but years away”. The reason we don’t yet have Joshua-level autonomous commerce isn’t primarily technical. It’s that nobody has built the trust and permission framework that genuine delegation requires.
Consider what real delegation would mean. You’d want to be able to say, “Book my family a two-week holiday in Barbados during the school holidays: flights, hotel and car, cheapest dates, £5,000 budget,” and walk away. The technology can attempt that today. What stops us isn’t capability; it’s that the consequences of getting it wrong are real, financial and potentially irreversible. Who refunds the hotel if your agent booked the wrong one? Who is liable when it found the cheapest dates and your children were still in school? What we have, in the book’s phrase, is capability without authorisation: Joshua with a credit card. Impressive, right up until the metaphorical warhead parcels start arriving at the front door.
For a retailer, none of this is philosophy. It is your returns policy, your liability and your customer’s trust, all at once. An agent’s behaviour is entirely observable since you can see exactly what it bought. Its intention is not: what it was actually trying to achieve, on whose behalf and why, is opaque or absent. And the entire apparatus that commerce depends on to put things right, such as customer service, consumer law and the right to a refund, was built for a world where a person made the decision and can be asked about it. When an agent acts badly, responsibility becomes genuinely hard to locate.
This is why the retailers who navigate the next few years best won’t be the ones who moved fastest. They’ll be the ones who insisted on answering the second question – who authorised this, and who is accountable – before they switched anything on.
The book frames it as a test for the boardroom, and it’s a good one to sit with before any agentic project gets a budget line:
“If an agent acting on a customer’s behalf initiated a transaction right now and something went wrong – wrong product, wrong price, unauthorised commitment – what is the resolution path? Who does the customer call? Who refunds them? Who bears the cost?”
If you can’t answer all three clearly, you don’t yet have an accountability framework ready for agentic commerce. You have capability, but you haven’t earned the right to deploy.
The technology will keep getting more capable; that was never the hard part. The hard part – the part no vendor can do for you – is making sure that capability is matched by authorised, accountable representation of the customer it’s acting for. Get that right, and agentic commerce becomes an advantage. Skip it, and you’ve simply built a faster way to be wrong.
This is one thread of a much bigger argument.
Paul is a Retail Technology Strategy Consultant, who recently joined Equal Experts after spending the last 20+ years working for retailers across the globe, including positions as CTO at Halfords and Chief Architect at New Look, Primark, Marks & Spencer, and Argos in Shanghai. Connect with Paul on LinkedIn.
All business models have their pros and cons. But, when you consider the type of problems we help our clients to solve at Equal Experts, it’s worth thinking about the level of experience and the best consultancy approach to solve them.
If you’d like to find out more about working with us – get in touch. We’d love to hear from you.
