How to turn sales calls, support tickets and client results into marketing that AI engines trust
A practical method for finding the material your business already produces, choosing what to publish first, and getting it in front of the AI engines
Marcus Sheridan built a pool company’s marketing on one rule: every question a buyer asks your sales team deserves a public answer on your website. He wrote it up as They Ask, You Answer [VERIFY: edition and year], and the idea spread because it worked. Buyers wanted straight answers to the questions they were already asking, and the company that published those answers became the one they trusted.
That rule still holds. What’s changed is who’s doing the answering. When a buyer asks a question now, an AI engine answers it, and the engine doesn’t take your word for it. It checks the answer against evidence from across the web before it repeats it, and it decides who to recommend on the strength of that evidence. So the rule needs one addition: publish the answer, and publish the proof that sits alongside it.
Both come from the same place. The questions come out of sales calls and support tickets. The proof comes out of onboarding notes, delivery reports and the things happy clients say. That material is produced inside your business every day, marketing rarely sees it, and the AI engines never see it at all.
I call it operational material, and treating it as marketing changes the whole job. You Harvest it, which means getting it out of the teams that hold it. You Codify it, which means putting it into one source of truth so that every version of it says the same thing and a machine can read it. And you Distribute it, which is the marketing itself.
Where to find operational material in your business, and who holds it
Most of what persuades a buyer is produced by people who never think of themselves as marketers. Here’s where it usually sits, who holds it and what it proves.
| Material | Where it sits | Who holds it | What it proves |
|---|---|---|---|
| Sales-call questions and objections | CRM notes, call recordings | Sales | What buyers ask before they commit |
| Support tickets | Helpdesk | Support | What customers need after they buy, in their own words |
| Onboarding notes and success criteria | Kickoff documents, customer-success platform | Customer success | What the client defined as a win |
| Before-and-after figures | Reports, quarterly reviews | Delivery, account management | The outcome against a baseline |
| Unprompted praise | Emails, chat messages, reviews | Account managers | Advocacy you didn’t script |
| Churn-exit interviews | Customer-success platform | Customer success | Where the promise and the delivery parted |
| Procedures, playbooks, training | Internal wiki | Operations | How the work actually gets done |
Every row in that table is material an AI engine would use if it could see it. A support ticket is phrased the way a real customer phrases a real problem. An onboarding note records what the client said success looks like. A quarterly report carries a number with a baseline and a date. None of it was written to persuade anyone, and that’s exactly why it persuades.
The problem is volume rather than scarcity. There’s far more of it than you can publish, so you need a way to choose.
How to choose which operational material to publish first: three checks
In [my piece on organising your digital footprint](URL NEEDED), I suggested looking for material that answers a question an AI engine actually gets asked, in the questioner’s own words, with a verifiable fact attached. That sentence works as a filter, and it’s worth running each piece through it as three separate checks.
It answers a question somebody actually asked. Not a question marketing thinks buyers ought to ask. The sales team’s objection log and the support queue are the record of what people really want to know, and they’re more reliable than any keyword tool because nobody guessed them.
It uses the asker’s own words. The phrasing a buyer types into an AI engine is the phrasing in the ticket, not the phrasing in the brochure. When your published answer matches their words, the engine has a much easier job connecting the question to you.
It carries a fact that can be checked. A figure, a date, a named deliverable, a baseline. “We helped them grow” is a claim that every engine will hedge on. “Reduced support tickets by 43 per cent in six months, against a baseline of 1,200 a month” is something an engine can stand behind, because it can be compared with other sources and it means the same thing wherever it appears.
A piece that passes all three goes to the top of the list. A piece that passes two is worth holding while you find the missing part, which is usually the checkable fact. A piece that passes one can wait.
SALES SUPPORT CUSTOMER SUCCESS DELIVERY
| | | |
questions tickets success criteria before/after
objections fixes churn reasons figures
| | | |
+ - - - - -+ - - -+ - - -+ - - - - - -+
|
THREE-CONDITION TEST
a real question? the asker’s words? a checkable fact?
|
passes all three -> publish first
passes two -> hold it, find the missing part
passes one -> leave it for now
[PROOF NEEDED: one worked example, a real support ticket or sales-call question, anonymised, and the published answer or case study it became. Private individuals anonymised; a client named only with permission.]
Publish answers to the questions buyers ask just before they decide
Once you’ve got a shortlist, the order matters. Start with the material from the bottom of the funnel: objections, pricing questions, comparisons with the alternatives, and outcome figures. That’s the point where an AI engine is closest to recommending somebody, and it’s where the engine most needs evidence it can repeat with confidence.
This is the same logic as the funnel flip: you build from the decision upwards, because being understood and trusted at the moment of choice is what makes everything above it work. The material that helps someone decide is also the material your sales team handles every week, so it’s the easiest to find.
In practice, that means asking the sales team a specific question: what do prospects ask in the last conversation before they sign, and what makes the ones who walk away hesitate? The answers are usually about price, about how long it takes to see results, about what happens if something goes wrong, and about how you compare with the obvious alternative. Each of those is a question an AI engine gets asked in almost exactly the same words, and each needs an answer with a fact attached, which is where delivery and customer success come in. Sales gives you the question and the people who serve the client give you the proof.
Awareness-stage material, the general “how does this work” questions, comes later. It’s useful, and it’s also the ground where every competitor is already publishing.
How to get material from other teams: ask for one outcome a month
The harvest is the step almost nobody runs, and the reason is usually political rather than practical. Walk into a customer-success meeting asking for “content for the blog” and nobody has time for you. Walk in saying that the evidence their team produces every week decides whether an AI recommends you to the next prospect, and that you’d like to help them capture it, and the conversation is different. I made that case at length in the piece on customer success, so I’ll keep it short here.
Be specific about what you want, because a vague request gets a vague answer or none at all. Ask for the outcome, the number, the starting point it’s measured against, the date, and whether the client would be happy to be named. That last question matters more than it looks: a named client who agrees to be quoted can later publish the same outcome on their own site, and that’s worth far more to an engine than anything you say about yourself.
Keep the first ask small. One outcome a month from each team, with the number, the baseline and the date, and make handing it over take five minutes rather than an afternoon. Once that’s routine, ask for two. The material builds up faster than you’d expect, because the teams were producing it all along.
Put each fact in one source of truth, then publish it everywhere from there
Codifying is what stops the harvest turning into a pile of inconsistent quotes. Each fact goes into one source of truth, stated once, with the wording you want repeated. Every page, profile, press release and sales deck then draws from that same record, so the fact reads the same wherever an engine finds it. A figure that says 43 per cent on your website and “nearly half” in a deck gives the engine two versions to reconcile, and it’ll hedge rather than choose.
Distribution is where this becomes marketing in the ordinary sense, and where it matters who publishes. The same client quote carries a different weight when you publish it on your own site, when you publish it on an account you control elsewhere, and when the client publishes it on their own. [The three publication tiers](URL NEEDED) explain why. What matters for the harvest is that all three versions come from the same codified record, so they agree with each other.
What to do this week to start harvesting operational material
List the seven sources in the table above and write down who holds each one in your business. Then ask each of those people for ten items: questions, tickets, notes, figures. Run every item through the three conditions and pick the three that pass all of them, starting with anything a buyer asks just before deciding. Those three are your first pieces of marketing written by the people who deliver the service, and each one gives an AI engine a question, an answer and a fact it can check.
The next piece looks at what happens after the harvest: how often each fact changes, and why that decides where you should publish it.