Account Planning Template + a Claude Agent That Builds Every Rep's Territory Plan

Write the tiering rubric: fit, signal, history

Start with the customers you win. Take the last 20 closed-won deals and the last 20 closed-lost, and write down what the won ones had in common that the lost ones did not: headcount band, industry, a tool in the stack, a team that exists, a buying motion. That is fit, and if you have already built an ICP for the list-building play or a scoring model for the lead scoring play, reuse its fit component here. The territory plan is the same rubric applied to the accounts a rep already owns.

Then add what those plays do not have: history. An account with a closed-lost deal eleven months ago and a champion who is still there is a different account from one you have never spoken to, and a different one again from a lapsed customer. Give history its own points: prior opportunity, prior contact with a role you sell to, a lost deal with a documented reason, a lapsed contract. History is why territory planning is a rep's job and not a list-building job.

Signal is the third family and the one that makes the plan about this quarter: hiring for the role you sell to, a new leader in the buying group in the last 90 days, a funding or expansion event, a public complaint about the incumbent, a contract anniversary you know about. Set the tiers so that tier A needs fit plus at least one signal or a live history, and cap it. Twelve tier-A accounts per rep is a plan; forty is a list with a heading. The template below has the rubric I start from.

Pull the territory and fill the gaps

Connect HubSpot or Salesforce over MCP and pull each rep's accounts with everything the rubric reads: the fit fields, every deal open or closed with its outcome and reason, contacts with titles and roles, last activity date and direction. Expect the fit fields to be thin. In most CRMs half the accounts have no headcount and a third have no industry, which is why the old template's section four said 'unknown'.

Fill the gaps through Clay, Apollo, or ZoomInfo. Headcount, industry, tech stack, and funding are enough for the fit score; the buying-group people come later, and only for tier A, because enriching contacts for eighty accounts a rep will never touch is the fastest way to spend the credit budget on the C tier. If you already run enrichment in Clay, this is a table; if not, the agent can call the provider's MCP directly.

Then research the signal. For each account the agent runs a bounded web check: job postings for the titles you sell to, leadership changes in the last 90 days, a launch or expansion announcement, a public mention of the incumbent. It records the signal with its source and date, or records 'no signal found', and it must never invent one. The account brief play does this for one account before a call; here it runs across the territory once a quarter and once a month after.

• Reads: accounts by owner: fit fields, deals (stage, outcome, close reason, dates), contacts (title, role), last activity (date, direction)
• Reads: enrichment for missing fit fields (headcount, industry, stack, funding); buying-group contacts for tier A only
• Reads: web research per account for a dated, sourced signal
• Writes: tier, reason, and plan link to the account, only after the rep confirms; plans and the summary as documents; Slack posts

Tier the territory and draft the tier-A plans

The run scores every account against the rubric and writes a one-line reason beside the tier: 'A: 640 employees in logistics (fit 38), hiring a Head of RevOps posted Aug 22 (signal 20), lost deal Oct 2025 to incumbent on price, champion still in role (history 15)'. The reason is the whole product. A rep who reads 'A' argues with the rubric; a rep who reads the reason argues with the facts, and that is a conversation that ends in a better plan.

For each tier-A account, the agent then drafts the one-page plan from the template: why now, in one paragraph built on the signal; who to reach, listing the contacts we know with their last touch and the roles in the buying group we have nobody for; the play, naming which of your use cases or offers fits the signal and a first-message angle that references it; the risk, naming the incumbent, the last lost reason, and what would kill this deal; and three actions with owners and dates. Every claim carries a source, and any section the agent cannot support says 'no evidence' rather than filling the space.

Tell the agent whose voice the plan is in: the rep's, for the rep. It is a working document, not a deck. Short lines, the facts, the sources, and a space marked 'rep's read' at the top where the rep writes what they know that the data does not. The account brief skill is a good model for the compression; the plan is that brief plus history plus a play, once a quarter.

On {{QUARTER_START}}, build the territory plan for each rep in files/reps.csv.

For each rep, read from {{CRM}} every account they own: fit fields (headcount, industry, tech stack, funding, region), deals (stage, amount, outcome, close reason, dates), contacts (name, title, role tag, last touch), last activity (date, inbound/outbound).
Where headcount, industry, or stack is missing, enrich via {{ENRICHMENT}} and record the source. Do not enrich contacts yet.
For each account, run one bounded web check: open roles matching {{BUYER_TITLES}}, leadership changes in the last 90 days, launches or expansions, public mentions of {{INCUMBENTS}}. Record signal, source URL, and date, or 'no signal found'. Never infer a signal.

Score each account against rubric/territory.md (fit / signal / history, with caps). Assign tier A / B / C and write a one-line reason citing the points and the sources. Cap tier A at {{TIER_A_CAP}} per rep; if more qualify, rank by signal recency then history, and list the overflow as 'A-watch'.

For each tier-A account, enrich the buying group ({{BUYER_TITLES}}) via {{ENRICHMENT}} and write plans/{{REP}}/{{ACCOUNT}}.md using template/account-plan.md:
1. Rep's read: [leave blank for the rep]
2. Why now: the signal, its source and date, in one paragraph.
3. Who to reach: contacts we know (name, role, last touch); roles in the buying group with nobody known.
4. The play: which of our use cases or offers fits the signal; a first-message angle that references the signal, not our product.
5. The risk: incumbent, last lost reason if any, what would kill this.
6. Three actions with owner and date. [rep to confirm]
Every claim cites a source. Where there is no evidence, write 'no evidence', never a guess.

Write summary/{{REP}}.md: tier counts and open pipeline by tier; tier-A accounts with no known contact (coverage gaps); accounts whose tier changed since last quarter and why; which plays the tier-A set leans on.
Write summary/manager.md across reps: the same, by rep, plus territories where tier A is under 6 or over the cap.

Post each rep their plans and summary in Slack for review. Write nothing to the CRM until the rep confirms; then write tier, reason, and plan link to the account.

Monthly (first Monday): re-run signals and history for every account, re-tier, list what changed and why, and for each tier-A plan check the three actions against the activity log and mark done / not done / no longer relevant.

Write the manager's territory summary

The summary is one page per rep and one across the team, and it answers the questions a manager asks in the territory review: how many tier-A accounts, how much open pipeline sits in each tier, where the coverage is thin, and what changed. Coverage gaps are the most useful line: tier-A accounts where we know nobody in the buying group are the accounts the SDR pairs with the AE on, and a list of them with the roles missing is a prospecting plan without a meeting.

The team view adds the comparisons a rep's view cannot: a territory with four tier-A accounts against another with twenty is a territory design problem, and the summary shows it in the first week of the quarter rather than in the forecast in week ten. The plays column shows which of your use cases the tier-A set leans on, which is the enablement plan for the quarter: if fourteen tier-A plans lead with the same signal, that is the message to rehearse.

Keep the summary read-only and factual. It reports tiers, pipeline, gaps, and changes with their reasons. It does not rank reps. A manager who wants to compare rep effort has the activity log; the summary's job is to show where the territory is strong and where it is thin, so the conversation is about accounts.

Review with the rep, then write back

Book a 45-minute review with each rep in the first week of the quarter, with the plans open. The rep's first job is the 'rep's read' line on every tier-A plan: what they know that the data does not, the champion who is a friend, the deal that died for a reason nobody logged, the account that looks perfect and will never buy because of a parent company. That knowledge is the reason a rep owns a territory, and the plan gives it a place to go instead of asking the rep to also type the headcount.

Then the rep argues with the tiers. Some tier-A accounts move down because the rep knows something; some tier-B accounts move up for the same reason. Every override is logged with the rep's reason, so the rubric can learn from it at quarter end. The rep confirms the three actions per plan, changes the dates to real ones, and commits. Only then does the agent write the tier, the reason, and the plan link back to the account.

Write nothing to the CRM before the review. A tier that appears on eighty accounts on Monday morning without the rep's say is a number the rep will route around, and the plan is dead again by week two, for the same reason the template was.

Refresh monthly and check the plan against reality

The first Monday of each month, the run re-researches signals and history for every account, re-tiers, and lists what changed and why: a new leader in the buying group at a B account moves it to A; a tier-A champion who left the company drops the account and says so; a signal that expired 90 days ago stops counting. The rep gets the changes as a short list, not a new plan to read, and the manager's summary gains a 'changed this month' section.

For each tier-A plan, the run checks the three actions against the activity log. 'Intro to the new Head of RevOps by Sep 15' is either a meeting on the calendar, an email in the log, or not done. The plan marks each action done, not done, or no longer relevant, and the rep's review meeting in month two is fifteen minutes on the not-dones instead of an hour rebuilding the plan.

At quarter end, lay the tiers from week one beside what happened: which tier-A accounts produced a meeting, an opportunity, a deal; which tier-B accounts did, and what they had in common that the rubric missed; which overrides were right. The rubric is a file. The calibration is an hour with the data on the table, and the next quarter's plan is better for it before the agent writes a word.