Your AI Is an Honest Broker: You Brief It to Be Recommended, and Pay It to Be Pushed
For two years I’ve described the AI engines as your Untrained Salesforce: seven employees you never hired, Google, ChatGPT, Perplexity, Claude, Copilot, Siri, Alexa, working around the clock and either selling for you or selling for your competitor. The frame has done a lot of work in keynotes and with clients, because it turns an abstract visibility problem into a concrete commercial one. But there’s a moment, when you watch an engine recommend a competitor over you, where the salesforce picture quietly breaks, because your own salesforce isn’t supposed to sell for the other side. Treating the engines as employees you train felt like a good idea, and for a while it served, but it turned out to be unsustainable as the universal frame, because employees don’t sell for the competition and these engines do it constantly. Finding one approach that held everywhere, in the boardroom, on the practitioner’s desk, and across the paid side, was a genuine struggle. I worked it through while writing my keynote for Google Marketing Live in Singapore, and I want to name Raja Narula at Google here, because the conversations we had in that period really helped me get to the frame that finally held. That break is the interesting part, and chasing it down led me to a sharper analogy that I now reach for first.
The salesforce is really an honest broker
The sharper picture is this: the AI isn’t your salesforce, it’s an honest broker. Think of a travel agent who carries every airline, a mortgage broker with the whole market on screen, an independent adviser paid to serve the client sitting in front of them. They don’t work for any one brand, and that’s precisely their value to the buyer, because the buyer trusts them not to be in anyone’s pocket. An AI engine is the same kind of creature. It carries every brand in your category, it answers to the person asking, and it recommends whoever it judges best for them. Which is why it will recommend your competitor without a flicker of disloyalty: it was never loyal to you in the first place.
This sharper picture came out of discussions with Raja Narula at Google, while I was writing my keynote for Google Marketing Live 2026 in Singapore. We were working out how to make the brand side of AI recommendation land for a room of marketers, and the broker is what made it click: it explains the competitor problem without the contradictions the employee picture carries, and it does it in a single, familiar character.
An under-briefed broker loses you the deal three ways, and you were out-briefed, not outbid
Picture yourself as an airline whose brochures the agent keeps leaving in the drawer. Your first instinct is that the agent is doing a poor job: they don’t understand your routes, they don’t appreciate your service, they won’t put you forward. Then the penny drops. The agent isn’t doing a poor job at all, they’re doing an excellent job for the airline that briefed them properly, which isn’t you.
There are three moments in the agent’s day where this costs you. At the top, when a client asks off the cuff and your name simply doesn’t come to mind, so you’re never in the conversation. In the middle, when the agent does the research and hands over a pile of brochures, and yours isn’t in the stack they push. And at the close, when the client has all but settled on you and the agent says, well, there might be a better option, and talks them out of it, which is the cruellest of the three because it’s a deal you’d already won. None of that is the agent failing you. It’s the agent faithfully passing on the thin, unconvincing picture of you that’s the only picture you ever gave them. You weren’t outbid, you were out-briefed.
You can’t bribe an honest broker, so you brief one, then you train one
Here’s the part that decides what you actually do about it. You can’t buy an honest broker’s recommendation, and you wouldn’t want one you could, because a broker who can be bought is a broker the client stops trusting, and a broker the client stops trusting is worthless to you. What you can do is brief them. You give them the most detailed, the most convincing, the most corroborated information about who you are, what you do and who you’re right for, better and more often than any competitor bothers to.
That’s why I keep the word train even though nobody’s on your payroll. Briefing is the floor: you tell the broker your story once and hope it holds against the next brand through the door, which is all you can ever do with a human broker. This broker is different, because it learns from a single source of truth it reads right across the web, so you can go one step past briefing and genuinely train it: not tell it once, but make your corroborated story the thing it reads everywhere, consistently, until it cannot ignore you. Briefing is a conversation, training is rebuilding the reality the broker reads from, and the second is only possible because the Algorithmic Trinity every engine runs on (search engines for information, language models for intelligence, knowledge graphs for validation) reads from that one web, so a single source of truth trains every broker at once. None of that buys loyalty or touches the broker’s independence: you are feeding an unignorable, corroborated truth, and the broker still recommends on merit. That is the opportunity the name points at. The salesforce is untrained by default, because most brands only ever brief it, or never manage even that, and it is uniquely trainable in a way no human broker allows. Brief it better than the competition and, on merit, you become the name it reaches for. Train it, and you become the name it cannot ignore.
The second lever is paid, and it only pays once the briefing’s done
There’s a second lever, and it matters because the stages I work on most are hosted by the companies that sell it. You can pay the broker for prominence: the airline’s poster in the window, the premium brochure at the front of the rack, the sponsored slot. That’s advertising, and it works. But it works in a specific way that’s worth being honest about. Pay a broker who already rates you and they push you hard, because the placement and their genuine recommendation point the same way. Pay a broker who doesn’t rate you and you buy the slot, not the sell.
For me, that’s the part of the paid story most people get wrong. The broker places you because you paid, then quietly drags their feet, because pushing something they don’t believe in costs them the client’s trust, and the client’s trust is the only asset the broker actually has. You see the placement and assume the money’s working, while the conviction that converts is missing, and conviction doesn’t show up on a dashboard. So the two levers aren’t equal partners. You brief to be recommended, you pay to be pushed, and the spend only pays once the briefing’s done.
Third-party proof is what the broker actually believes
If briefing is the work, the obvious question is what counts as a good brief, and the answer isn’t what most brands assume. What you say about yourself on your own properties is claim and frame: necessary, foundational, and close to worthless on its own, because the broker knows you wrote it and you published it. What your clients and partners say about you is corroboration: a real step up, because the publication isn’t yours, though it’s still produced under your influence, so it carries some weight and not the most. What independent third parties say about you, the journalist, the analyst, the coverage written and gatekept entirely outside your reach, is proof, and it’s the heaviest thing the broker can read. The order is unforgiving. First party frames, second party supports, third party proves, and without proof nothing stands. You can’t write your way to the bottom of that gradient, you have to earn it, which is exactly why the broker trusts it when you do.
The broker is the Mirror Principle wearing a face
Underneath the whole analogy sits the principle I’ve held since 2012: what an engine shows about you is its opinion of the world’s opinion of you. The broker makes that principle physical. An honest broker forms their view of you from everyone else’s view, the reviews, the coverage, the corroboration scattered across the market, and where the world is quiet about you and loud about a competitor, the broker reads exactly that and recommends accordingly. You can’t lean on them and you can’t flatter them, you can only change what the world is able to see. Improve the genuine, visible reality, and the broker, doing what an honest broker does, reads the better reality and passes it on. That’s also the answer to anyone who calls this manipulation: the broker is reading proof that was always true and merely invisible, and reporting a truer picture of you is the job, not a trick.
So the question is who briefed it
The salesforce is still your salesforce, in the sense that it’s selling in your market every hour of every day. It’s just that it’s an honest broker, and the brand it recommends is the brand that briefed it best. Brief to be recommended, pay to be pushed, and the only question left is whether you did the briefing or your competitor did.
Strategy Sandbox. First publication: 7 June 2026. Status: original articulation.
The Untrained Salesforce and the Algorithmic Trinity are my coinages from 2024: the AI engines as a sales team you never hired, selling for you or your competitor around the clock, all running on the same three technologies (Search engines for information, Language models for intelligence, Knowledge Graphs for validation) that read from one web. This piece formalises the honest-broker analogy as the way to understand that salesforce, together with the two-lever model (brief to be recommended, pay to be pushed), the brief-to-train escalation (briefing becomes training when it runs through a single source of truth the broker cannot ignore), and the observation that a poor opinion isn’t amplified by ad spend but quietly throttled, because the broker protects the client’s trust before your budget. The broker is an analogy, not a new term: the named concept remains the Untrained Salesforce, and the analogy explains how it behaves. The analogy took its current shape while I was preparing my keynote for Google Marketing Live 2026 in Singapore, with thanks to Raja Narula at Google, whose conversations in that period helped me reach it. Jason Barnard, Kalicubeยฎ.