How to map stakeholders and multithread a 6-month enterprise deal using AI

TL;DR: Pull every named person from your transcripts, emails and calendar into a stakeholder table, classify each by role and disposition using evidence only, then identify structural gaps — the roles that must exist in a deal this size but that you have never met. The output is a named, sequenced outreach plan, not a diagram.

The problem: one person is not a relationship

Enterprise cycles run six to 24 months, and the longest ones are long precisely because of the number of stakeholders and procurement gates involved. Over that span your champion gets promoted, reorganised, or leaves. If they were your only real relationship, the deal restarts from zero — except worse, because their successor inherits a project they did not choose.

Everyone knows to multithread. Almost nobody does it systematically, because it requires admitting who you have not met, and that is uncomfortable to write down. Which is exactly why handing the analysis to something that has no ego about it works well.

What you need

• Call transcripts from the account (all of them, concatenated).
• Your email thread history with the account — export or paste the substantive threads.
• Calendar entries showing who actually attended what.
• Whatever your CRM holds on the account.
• Claude or ChatGPT on a paid tier. A Project keeps the account context loaded across the whole deal — see the Projects comparison.

Step 1: Extract everyone who has been named (15 minutes)

“Below are transcripts, email threads and meeting invitations from an enterprise opportunity. Build a table of every individual mentioned, including people who were referred to but never met.

Columns: Name | Job title as stated | Function | Have I met them? (yes / no / email only) | Every verbatim mention of them with who said it | What they appear to care about, with evidence | Apparent stance (supportive / neutral / sceptical / unknown).

Rules: include people identified only by role, such as ‘our head of security’, as separate rows marked UNNAMED. Mark stance ‘unknown’ unless there is a direct quote supporting it — do not infer enthusiasm from politeness. Quote sources for everything.”

The UNNAMED rows are frequently the most valuable output. “Our head of security will need to look at this” mentioned once in April, never followed up, is the sentence that explains why the deal stalls in October.

Step 2: Find the structural gaps

Now ask what should exist but does not appear:

“For a [deal size] [product category] purchase at a company of roughly [headcount] in [industry], list the roles that would typically need to be involved for a deal of this type to close — including technical evaluation, security review, procurement, legal, finance approval, and any operational team who inherits the service afterwards.

Compare that list against my stakeholder table. Output three groups: (1) roles I have engaged, (2) roles that exist in the account but I have never spoken to, (3) roles I have no evidence exist at all. For group 3, tell me the question I should ask to find out.”

Group 2 is your work list. Group 3 is your risk list — the roles you do not know about are the ones that appear in month five with objections nobody prepared for.

A note on accuracy: the model is generalising about org structures here, and it will sometimes describe a textbook company rather than this one. Treat group 3 as hypotheses to verify with your champion, not as fact.

Step 3: Classify the relationships honestly

“For each person I have actually engaged, classify them: CHAMPION (actively sells internally when I am not there — requires evidence they have done so), SUPPORTER (positive but passive), COACH (gives me information but has no stake), BLOCKER (has raised objections), or UNKNOWN. Cite the evidence. Be strict about CHAMPION: wanting the product is not the same as advocating for it. If nobody qualifies as a champion, say so plainly.”

The strictness matters. Most deals that reps describe as having a strong champion actually have an enthusiastic user with no political capital. The test is whether there is evidence of them advocating internally in your absence — running a meeting, forwarding your material, chasing procurement.

Step 4: Build the outreach sequence

“Based on the gaps, build a multithreading plan. For each person I need to reach: who is the best internal route to an introduction and why, what is the specific reason for the meeting from THEIR perspective rather than mine, what would I be able to offer that is useful to them, and what is the risk of approaching them.

Sequence them — some introductions should happen before others. Flag anyone I should NOT approach without my champion’s agreement, and explain why going around them would be damaging.”

That last constraint prevents the classic error: going over a champion’s head to their boss, and converting your only advocate into a blocker. AI is happy to suggest aggressive multithreading; the sequencing instruction forces it to account for internal politics.

Step 5: Give each stakeholder their own reason to meet

The reason multithreading fails is usually not access — it is that the rep offers the same pitch to everyone. A security architect does not want your value proposition. They want to know how you handle their specific concern.

“For each person in the outreach plan, draft a short introduction request. Lead with something relevant to their role, reference the internal context accurately, and make the ask small and specific — a 20-minute conversation with a named agenda, not a demo. Under 120 words. No flattery, no ‘I hope this finds you well’.”

The principles here are the same ones in the cold outreach guide, with the advantage that these are warm internal introductions rather than genuine cold contact.

A worked example of what surfaces

On a deal I would have described as well-covered — six months in, four people engaged, regular meetings — the extraction produced this:

Engaged: IT Director (met 8 times), Infrastructure Manager (met 6), Network Architect (met 3), Procurement Officer (email only, once).

Named but never met: a Group CFO referenced once in February — “that’ll need to go up to group finance”; a Head of Information Security mentioned in April — “security will want to look at this before we commit”; a Site Operations Manager for the two largest locations, mentioned three separate times as someone who “would need to be comfortable with the cutover.”

Roles with no evidence of existing: nobody had ever mentioned legal, and for a multi-year multi-site contract that is not credible — they exist, we simply have not found them.

Champion assessment: UNKNOWN. The IT Director was warm across eight meetings, but there was no evidence in any transcript of him advocating internally when we were not in the room. No forwarded material, no internal meeting he ran, no chasing of procurement.

That is a deal with four contacts that is functionally single-threaded, with three known unmet stakeholders each capable of stopping it, and no evidenced champion. It felt like a strong deal. On the evidence it was a well-attended series of pleasant meetings, which is not the same thing — and the security review nobody had scheduled since April was the thing that eventually cost it a quarter.

Step 6: Re-run monthly

Add each new transcript to the Project and re-run steps 1-3. What you are watching for over time:

• Is the number of engaged stakeholders growing? Flat for two months on a large deal means single-threaded, whatever the meeting count says.
• Has anyone’s stance moved? Particularly, has a supporter gone quiet — silence after engagement is a signal, not neutral.
• Have new names appeared? New names late in a cycle usually mean the deal has been escalated, which cuts both ways.
• Has anyone left? The map should be checked against reality; people change roles and your CRM will not tell you.

What this cannot do

1. It cannot read the politics. Who dislikes whom, whose budget is threatened by your project, which director is on their way out — these determine outcomes and appear in no transcript. Your champion is the only source, and you have to earn the candour.
2. It will invent plausible org structures. Ask it who “typically” approves and you get a generic enterprise. Verify before acting.
3. It overstates warmth. Transcripts are full of professional courtesy that models score as support. This is why every stance classification demands a quote.
4. Privacy. You are building a file on named individuals. Keep it factual and professional, restrict it to business-relevant observations, and be aware that in some jurisdictions this is personal data with obligations attached. Nothing should be in there you would be embarrassed for the person to read.

Do this properly and the payoff is unglamorous but decisive: when your champion announces they are moving to a new role, you already know three other people who want this to happen.