← All writing

Essay · Sales enablement and reputation

Red team your sales narrative before an answer engine does it for you

Your buyers meet a version of your story that nobody on your team approved. Run the attack yourself, on a calendar, and fix the claims that break.

Essay · Published October 18, 2026 · Updated October 24, 2026

The hostile version arrives first

Your prospect has already heard your pitch. Not from your rep. From an AI summary, a competitor's comparison page, a peer who churned two years ago, a review site. That version is short, nobody on your team approved it, and it is rarely the one in your deck.

In Lead Quality Over Volume I wrote that prospects who find a company through ChatGPT can arrive nearly fully researched. That is good news for cycle time when the research is right. When it is wrong, the rep walks into discovery carrying a story the buyer has already discounted, and nobody in the room says so out loud.

The rep almost never sees the prior. It surfaces late, as an objection in a pricing conversation, a security questionnaire that asks about a weakness you thought was fixed, or a deal that simply goes quiet. In The Messaging Consistency Gap That's Killing Your Pipeline I argued that when marketing, PR, sales and customer success tell different stories, the AI descriptions of the company turn muddy. Inconsistency is fuel. The hostile version feeds on it.

So the question for a revenue team is practical: what does the version of your story that buyers meet first actually say, and which of your claims break when someone attacks them? The answer should come from a drill you run on purpose, on a calendar, before a buyer runs it for you.

What follows is a proposed method, a design I would hand to an enablement lead, not a report of results. The unit of work is the individual claim. The cadence is quarterly. The output is enablement material, not a crisis plan.

What carries over from reputation red teaming

In Reputation Red Teaming, written for Risk Management Magazine, I applied military red team methods to reputation risk: simulate hostile media narratives, competitor positioning, stakeholder pressure and incomplete information, then watch how a crisis team responds under stress. The core idea transfers intact. An organization cannot see its own blind spots, so it assigns people whose only job is to find them.

Three things change when the target is a sales narrative.

  • The target is a claim, not an incident. A crisis exercise rehearses one bad day. A sales drill audits the dozen sentences a rep says on every call, one at a time.
  • The tempo is scheduled, not event driven. Crisis teams red team before a plausible disaster. A sales narrative decays continuously, as competitors reposition and AI summaries drift, so the drill runs on a fixed cadence.
  • The output is an enablement asset, not a response protocol. The deliverable is a revised talk track, a proof request, a retired claim, or a public content brief. Nobody gets a holding statement.

Everything below is built for those three shifts.

Roles and inputs

The drill needs four roles and one rule: nobody plays two of them. A half day session is enough for a first pass on a single product narrative.

Drill roles
RoleWhoJob in the room
Red cellTwo or three people outside the deal team, at least one from customer successAttack each claim using only sourced material: a captured AI answer, a competitor page, a lost deal note, a review theme
DefendersTwo working repsPresent the pitch exactly as delivered today and answer each attack with materials that exist now, not promises
JudgesThe enablement owner plus one product or finance voiceScore each claim independently against the rubric; they do not argue with the red cell or coach the defenders
ScribeOne person, usually from enablement operationsOwns the claim register: every claim, every attack, every score, every assigned fix

Customer success sits on the red cell for a reason. They hear the gap between what was sold and what was delivered, which is the raw material of the most damaging attacks. The finance or product judge keeps scoring honest on claims about cost, roadmap and integration.

Inputs get assembled a week before the session:

  • The pitch as delivered. Current deck, talk track, and the ten claims reps make most often, written as single sentences. The scribe numbers them; that list becomes the claim register.
  • Lost deal notes from the last two quarters, especially the stated reason and the competitor named.
  • Competitor positioning. Their homepage language, comparison pages, and the narratives they push in coverage. In Competitive Intelligence Mastery I described tracking competitor narratives to spot positioning openings. The red cell uses the same discipline in reverse: where are they positioned to hit you?
  • Captured answer engine responses. Write the questions a buyer would actually type (category, comparison, pricing, risk) and save the full responses from more than one engine, with the date. In Voice, Speed and Trust I recommended auditing how AI describes the brand. Here that audit becomes evidence for an attack.
  • Review themes. The recurring complaints and praise on public review sites, grouped, not cherry picked.

Six attack categories

A red cell left to improvise attacks whatever annoys it. Categories make the drill repeatable and make one quarter comparable to the next. Each claim faces every category that applies, and each attack must be phrased as a question a real buyer could ask.

  1. Answer engine drift. What does the AI summary omit, garble, or attribute to a competitor? The red cell reads the captured response aloud and asks the defender to reconcile it. Example: "The summary I got says your platform lacks a native integration with our CRM. Is that wrong, and where would I see that it's wrong?"
  2. Competitor reframe. Your strength, restated as a weakness in the competitor's words. Example: "You call it configurable. Your competitor calls it a six month implementation. Which is it for a team our size?"
  3. Former customer story. The red cell plays a buyer who talked to someone who left. Customer success supplies the plausible version. Example: "A peer told me support got slower after year one. What changed, and what would stop it happening to us?"
  4. Procurement and risk skeptic. Proof, security, total cost. This is the attack reps escalate most, so it tests whether the claim can stand without a sales engineer in the room. Example: "You say customers see value in the first quarter. Show me a named reference with a comparable environment, and tell me what the second year costs."
  5. Internal contradiction. The red cell lines up what marketing publishes, what sales says, and what customer success tells renewals, and asks the defender to pick one. This is the consistency gap turned into a live test. Example: "Your website says this is built for enterprise. Your onboarding guide talks about teams of ten. Who is it really for?"
  6. Silence. The buyer question no material answers at all. These are the most dangerous, because an answer engine will fill the gap with someone else's content. In Why Sentiment Alone Can't Explain Reputational Risk I argued that tone scoring misses where a brand is actually vulnerable. A silence never shows up in sentiment data, because nobody said anything. Example: "How does pricing change if we cut seats mid contract?"

Categories one and six are what separate this drill from an objection workshop. Both come from outside the team's own deal history.

Scoring each claim

Without scoring, a red team session becomes a debate won by the most senior person in the room. The fix is a short rubric that two judges can apply separately and mostly agree on. In Beyond Press Releases I made the case for a decision ready reputation scorecard. The same logic applies here: the score exists to produce a decision about each claim.

I propose four dimensions, each scored 0, 1 or 2. A three point scale is coarse on purpose. Finer scales invite false precision and judge disagreement.

Claim survival rubric (proposed)
DimensionTestScores 2 whenScores 0 when
ProvableDoes public proof exist that a buyer can check without asking us?A named reference, published data, or documentation backs itProof exists only in a rep's memory or an NDA deck
ConsistentDo marketing, sales and customer success state it the same way?All three sources in the inputs match in substanceTwo functions contradict each other
Survives compressionDoes the claim come through intact in the captured AI answers?The summary states it accurately, attributed to usThe summary omits it, garbles it, or credits a competitor
Defensible liveDid the defender handle the attack without escalation?Answered with existing material, buyer question closedDefender deferred to a specialist or conceded

A 1 is anything in between, and judges note why. Each judge scores alone; the scribe flags any dimension where the two judges are two points apart, and only those get discussed. The register then sorts by total, lowest first. A tie goes to the claim that appears earlier in the deck, because reps lean on it more.

Two patterns deserve attention beyond the totals. A claim that scores 2 on defensible live and 0 on survives compression is a claim your best reps carry in person that the market never hears. A claim that scores 2 on provable and 0 on consistent has proof nobody is using. Both are cheap to fix and easy to miss.

Outputs that feed enablement

The drill fails if its findings stay in a slide titled "learnings." Every low scoring claim leaves the room with an owner and a destination asset.

  • Updated claim register. Owner: the scribe, handed to the enablement lead. Destination: the single source of truth reps and marketers both draw from, with scores and dates.
  • Revised objection handling. Owner: enablement. Destination: the talk track and battlecards, rewritten around the actual attack language from the session, not a sanitized version.
  • Proof library requests. Owner: product marketing. Destination: the reference program, case study queue, or security documentation, with the specific claim each new asset must support.
  • Retired claims. Owner: enablement and product marketing jointly. Destination: removal from the deck, the website, and any outbound templates. A claim that scores low on provable and defensible live, quarter after quarter, costs more credibility than it earns.
  • Public content briefs. Owner: the content lead. Destination: published material that closes answer engine drift and silence findings.

That last output matters most, and it is the one sales teams skip. Fixing the deck does nothing for the buyer who reads the AI summary before the call. The brief goes through RISE, the framework I laid out in The RISE Framework and in The Future of PR Measurement: Research, Integrate, Scale, Evaluate. Applied to a drill finding, most of the Research is already done, since the captured answers show the gap. Integration puts the same corrected claim across the site, documentation and spokesperson language. Scaling gets it published in places answer engines draw from. Evaluation happens at the next drill, when the red cell recaptures the same questions. The external record changes, not only the deck.

Cadence and retrospective

A single drill is an event. The value comes from the rhythm.

  • Quarterly, plus triggers. Run the full drill each quarter. Run a short version on any trigger: a launch, a pricing change, a competitor repositioning, or a review theme that suddenly spikes.
  • Recapture before every drill. Ask the answer engines the same question set each time and keep the dated responses. The comparison quarter over quarter is the only honest way to tell whether public content briefs worked.
  • Retrospective immediately after. In Retrospective Meetings I recommended a one to two hour retrospective right after a project, so the team converts success and failure into reusable practice. Here it covers the drill itself: which attacks were weak, which inputs were missing, where judges disagreed and why.
  • One leading indicator, reviewed weekly. Track the count of claims still scoring 4 or below with an open fix. In Track Business Success with Leading Indicators and Dashboards I argued that leading indicators let teams act before results arrive. Win rate is a lagging result. A falling count of weak claims is something the enablement lead can move this week.

We already do this, and the answer to that

The fair objection: win/loss reviews, battlecards and objection workshops already exist, so this is old practice with a new name. I have owned sales enablement (at PR Newswire, as VP, Public Relations Products), and I respect those tools. The overlap is real. The drill reuses lost deal notes and produces battlecard revisions.

Three gaps remain. Those practices start from deals the team saw, not from the story buyers met before the team showed up. They rarely test whether a claim survives AI compression, so the answer engine drift and silence categories have no home in them. And their outputs stay internal; none of them is built to change the public record that buyers and answer engines read. The distinct parts here are those two attack categories, the compression score, and routing findings through RISE into published content.

The first step fits in a week. Write down the ten claims your reps make most often, one sentence each. Ask two answer engines five questions a buyer in your category would ask, and save the responses. Read them against your ten claims. Any claim that is missing, wrong, or credited to a competitor goes to the top of the agenda for your first drill. Then put the drill on the calendar and name the red cell.

Sources