The red flags nobody told me about

A lead took our deck, pasted it into an AI, and asked it to find the holes. Then he sent us the holes.

That is the whole story, and it is a good day, not a bad one. It is a good day because he told us. Most people who do this do not tell you. They read what the AI hands back, they see a page of red flags, and they just go quiet. You never find out you lost. You think the message did not land. The message landed fine. It got read by a machine that had no reason to be kind, and the machine won.

I want to be honest about our own side here, because that is where this started. We built a deck for the India region. Good deck. We had looked at it many times. A potential lead fed it to whatever LLM he uses for due diligence, and it came back with a list. Some of it we already knew. A couple of things we should have caught. Nothing on the list was a lie - it was our own material, read back to us cold, every loose end pulled at once. Sitting in the meeting reading his notes, the feeling was not embarrassment. It was: how many people got this same list and said nothing.

That is the part that should worry any founder more than the list itself. One person did us the favour of telling us. As my colleague put it in the meeting, it could be a silent killer if nobody mentions it. Plenty of people are probably running the same check in the background and just not contacting us. You only ever hear from the ones who report it.

The buyer changed, and did not send a memo

This is not a story about AI being scary. It is a story about a habit that has already formed while most sellers were still tuning their opening line.

Gartner surveyed B2B buyers and found 45% used generative AI during a recent purchase, primarily to gather information on vendors and products (Gartner, reported by Demand Gen Report, 2026). Not to write poems. To check you. Roughly one in two of the people you are pitching now has a research assistant that will read your website, your deck, your demo page and your LinkedIn in one go, hold them next to each other, and point at anything that does not line up.

A human reading your deck is generous by default. They skim, they give you the benefit of the doubt, they fill small gaps with goodwill. An AI does none of that. It reads all of it, it does not get bored on slide nine, and it treats every claim as something to check rather than something to nod along with. It is the most thorough, least forgiving reader you will ever have, and it shows up before you do.

So the old sequence is broken. It used to be: you send a message, they reply, you talk, then somewhere down the line procurement kicks the tyres. Now the tyre-kicking happens first, silently, by a machine, before anyone has typed a word back to you. By the time a buyer decides whether to reply, the AI has already given them a reason not to.

Why this fits what we already see

We run one LinkedIn account with about 18,000 messages on record, and the reply data has said the same thing for months. The corporate opener - we specialize in custom AI solutions - went to 838 people and replies 16%. A five-word note, congrats on the new role, replies 60% off 107 sends. A warm, useful thing sent to someone already connected replies 77% off 35. The median reply, when one comes, lands in about 0.8 hours.

Read together, those numbers were always about one thing: trust decides whether you get a conversation, not cleverness. The AI-due-diligence habit is the same rule wearing new clothes. The buyer is not reacting to your sentence anymore. They are reacting to whether everything you have ever published holds together when a cold reader lines it up. A great opener on top of a deck that fails the check earns you nothing. The check happened first.

What we are actually going to do about it

The fix is not to write for the machine. It is to stop being surprised by it.

Run the adversarial pass yourself, first. Before a deck or a page goes out, feed it to the same AI the buyer will use and ask it to tear the thing apart. Ask it what a sceptical buyer would flag. It will hand you the red flags for free, in private, while you can still fix them. We are now doing this on every deck, and it is going in as a standard step for every product we put out, not a one-off. As I said in the meeting, run an adversarial AI pass always. It gives you all the red flags up front, before a customer does.

Then fix what is fixable and answer what is not. Some flags are real gaps - close them. Some are just missing context the AI could not find, so put the context in. We are adding a plain FAQ-style slide that answers the obvious hard questions directly, so the facts are sitting right there when the buyer's AI goes looking. When the machine reads material and finds that every claim checks out, that is not a neutral result. That is the thing that builds trust in the conversation that follows.

And here is the part that should keep a salesperson calm. The same Gartner survey found 69% of B2B buyers still turn to sales reps to validate AI-generated insights at the critical moments. The AI does the first read. The human still closes it. So the goal was never to win the machine. The goal is to not get quietly eliminated by it before you are in the room, because the room is still where the deal happens.

The lead who sent us his AI's red flags did more for us than a polite reply ever would have. He told us what everyone else was seeing and not saying. The uncomfortable question is the one he left behind - if a stranger's AI read all of your material today, cold, with no goodwill, what would it hand back, and would you ever hear about it.

---

Sources

First-party (network + meeting)

1. Team meeting transcript, 2026-08-27 — Drive doc 1jhYAk3naIaEXjOKHinRCrAafOctkeBcKTU_GeyKn0Cg, Transcript section (read in full). - Rameshver Premdayal: "one of the leads kind of fed that [the India-region deck] to their LLM maybe something they used to do due diligence and ... there are a lot of red flags I mean things we already know and ... there are some good reasons uh mitigating factors for a lot of points but I'm wondering if we should make sure our deck kind of uh AI research proof." - Rameshver: "it could be a silent killer if nobody mentions it. Maybe many are running it in the background and running similar workflows in the background and they are just not contacting us. You only know the ones who report this to us." - Pravin Luthada: "run an adversarial AI pass always I think it kind of is a good practice so then it gives already all the red flags out already to us." - Pravin, on why it matters: "when the customer then passes it through any of the LLM they know okay hey all the data provided was actually correct then and such and I think that's the main bit ... it just builds trust in any discussion." - Only our own deck is described as flagged. The named-person legal/restructuring specifics discussed in the same meeting are private and sensitive and are deliberately excluded from the story.

1. Reach get_results, 2026-08-27 run — one LinkedIn account, ~18,490 outbound messages on record; 3,817 contacted, 934 replied, replyRate 0.245, median time-to-reply 0.8h. By opening line: corporate pitch "We at Linkenite specialize in custom AI solutions..." 0.163 (137 of 838); "Congrats on the new role!" 0.598 (64 of 107). By approach: shared_asset (a useful thing sent to already-connected people) 0.771 (27 of 35). Used only as corroboration that trust, not phrasing, earns the conversation.

Secondary (opened + verified)

1. Gartner B2B buyer survey (2026), via Demand Gen Report — "Gartner: AI Is Reshaping B2B Buying, but Human Sellers Still Close the Confidence Gap", demandgenreport.com. Opened and read. - "45% of buyers used Gen AI during a recent purchase, primarily to gather information on vendors and products." - "69% of B2B buyers prefer to validate AI-generated insights with sales reps" at critical decision points. - Analyst: Robert Blaisdell, VP Analyst, Chief of Research, Gartner Sales practice. - Why cited via the outlet, not the primary: Gartner's own press-release page (gartner.com/.../2026-05-20-gartner-survey-finds-sixty-nine-percent...) returned HTTP 403, and the BusinessWire wire copy of the same release reset the connection on three attempts. The Demand Gen Report piece attributes and quotes the survey and the analyst by name, and its 69% figure is corroborated by the search index. Only the two figures above are used.

Deliberately NOT cited

• Forrester 2024 "89% of B2B buyers adopted generative AI" — surfaced in search but the primary was not opened, so it is not used (per the 08-20/08-21/08-25/08-26 precedent: no figure that was not opened).
• Any floating vendor statistic on "AI due diligence" adoption — none had an opened primary.

House rules honoured

• No real company named as struggling; the only material critiqued is our own deck.
• No price anywhere.
• No competitor named.
• The private named-person legal matter from the same meeting is excluded entirely.