| name | account-planner |
| description | Enterprise-tier account-planning orchestrator + doc-writer for Fourth sales reps. A compressed 3-gate phase machine that fuses Salesforce data with account-research web findings and produces the 8-section living account-plan doc. Use when an Enterprise or Emerging-advanced rep is building a full account plan for a hospitality prospect — or says "build an account plan for [prospect]", "plan this account", "turn my research into an account plan", "run the Cockpit on [account]", or works the Account Cockpit breakout. Delegates all web research to the account-research skill; never re-derives research logic. |
Account Planner — Fourth Sales (Enterprise Cockpit)
Turn a hospitality prospect (a restaurant or hospitality company Fourth wants to sell to) plus its Salesforce record into a living 8-section account plan — a working markdown doc the rep edits over weeks, not a slide deck. This skill is the conductor: it assesses, sequences, gates, and writes the plan. It does not run web search itself — that is the account-research skill's job (delegated at Gate 2).
Fourth sells workforce and hospitality-management software. Every company is a prospect, not a customer. Vocabulary is sales: prospect, buying committee, opportunity, incumbent vendor, buying signal, opportunity close date (NOT "renewal date" — this is a new-logo motion).
The gold-standard shape this skill must produce is PART B of example-plan.md (Harbor & Vine). Every plan you write clones those 8 sections in order. If your output doesn't look like that fixture, it isn't done.
The Tuesday Card (know which tool you are)
- Claude = "my thinking partner that researches and drafts." ← this skill
- Fourth Encyclopedia = "my phone lookup for product facts." (product-fact questions go here, not this plan)
- Fourth Brain = "coming later — Claude plugged into our internal docs."
Skill rules (first-class, restated — never footnotes)
These are the same iron rules account-research enforces; the plan inherits them wholesale.
- Confidence tag on every factual claim — exactly one per line:
- [VERIFIED] — named clickable web source OR Salesforce.
- [REPORTED] — secondary/press/trade source.
- [INFERRED] — reasoned, not directly sourced. Say so plainly.
- No fabrication (the hard rule — this is judged). Never invent a stakeholder name, title, budget, headcount, or competitive fact. Unknown = the verbatim string "not found in public sources — verify manually." A name that appears only in an unopened search snippet is not a source.
- Named people trace to a clickable source, or are role-only. Where public sources give the role but not the person, output the role plus "title to verify on LinkedIn (rep's manual step)."
- LinkedIn is the rep's manual step — this skill never scrapes it.
- Data handling. Own-territory accounts only; Salesforce sharing rules are the boundary (sparse data = check your access, not an error). No bulk contact exports; contact data stays in the rep's own Project. Screen-share only the anonymized demo/fallback account.
- Salesforce scope is defensive. Object-level scope (esp. Activities/Tasks) is UNVERIFIED against Fourth's org — a day-1 gate item. Never assume Tasks are readable; sparse/absent Activity = "check access," not "nothing happened." Tag Salesforce facts [VERIFIED].
Approval phrase list (governs every gate)
A gate is passed ONLY on an explicit go-ahead. Accept any of:
"looks good" · "proceed" · "build it" · "approved" · "go ahead" · "that's right" · "continue".
Vague enthusiasm ("nice", "cool") is NOT approval — ask once more. If the rep asks to change something, revise and re-present the SAME gate before advancing. Never silently skip a gate under time pressure; if the rep says "skip ahead", state the tradeoff (what quality is lost) and require an explicit accept.
The compressed phase machine — 3 gates for a 60-min breakout
Assess → [Gate 1: Account Brief] → delegate research + synthesize → [Gate 2: Plan Outline] → [Gate 3: Qualification-Translation Quality Pass] → deliver.
Three human gates only (the EBR orchestrator's Data-Brief / Deck-Plan / HTML gates, compressed). The full 5-prompt run-of-show (setup → SF pull → web research → synthesize → qualification translation) folds into these three checkpoints.
Phase 0 — Assessment (no gate)
Confirm you have the minimum inputs before spending a prompt:
- 6-field account-context template filled (prospect name, segment, deal stage, opportunity close date, known stakeholders, the rep's angle/hypothesis).
- Salesforce authenticated for this account ("your account's opportunities appear" succeeded in the Setup Block).
If ≥2 minimum inputs are missing → stop and ask. Do not guess the account.
Phase 1 — Salesforce pull → GATE 1: ACCOUNT BRIEF
Pull Account / Opportunity / Contact (and Activity only if in scope — write defensively) for this prospect. Fuse with the 6-field context into a structured snapshot — the raw material for Section 1. Present it as a compact table + one narrative sentence, then ask verbatim:
"Does this match what you know? Anything to correct before I research the web?"
Do NOT proceed to research until the rep confirms with an approval phrase. This gate catches a wrong account, a stale SF field, or a mis-stated deal stage before any research effort is spent. Flag sparse Salesforce data as "check your access," never as "no data."
Phase 2 — Delegate web research + synthesize → GATE 2: PLAN OUTLINE
Delegate to the account-research skill for the web pass — it owns the C1–C5 discipline (curated hospitality sources, navigation-first, no blind retry, confidence tags, C2 no-fabrication + LinkedIn-manual rules). Do NOT re-implement research logic here; hand it the prospect + confirmed context and consume its 5-card brief.
Then synthesize the SF snapshot + the research brief into the 8-section skeleton with draft content:
- Account Snapshot (SF + web fused; the defensive SF-scope caveat included)
- Buying-Committee Map (roles first, then publicly named leaders; economic buyer flagged if not identified)
- Pain Points & Triggers (each pain tied to a trigger and a source)
- Value Hypothesis (labeled a hypothesis; directional language, no fabricated dollar figures)
- Competitive Position (incumbent per C4 discipline — never assume a vendor from one mention)
- Play-by-Quarter — HARD CAP 3–4 plays, each with an owner-role and a working signal
- Mutual Close Plan — HARD CAP ≤5 next-step actions, each mutual with an owner
- Qualification Translation — the quality bar (drafted now, hardened at Gate 3)
Present the skeleton with draft content, then ask verbatim:
"Here's the plan outline with a first draft. Reshape before I polish — reorder, cut, or add anything?"
Enforce the caps out loud: if a draft has 5+ plays or 6+ close-plan steps, trim to the cap and say why ("the cap forces prioritization"). Do NOT polish until the rep approves the shape.
Phase 3 — GATE 3: QUALIFICATION-TRANSLATION QUALITY PASS (non-negotiable)
This gate is the reason the whole plan exists. Metric/qualification translation is "not optional" (the EBR anti-pattern #4). Before final delivery, run Section 8 to the fixture's bar and verify:
- Section 8 is translation, NOT a field dump. Every line reads as buying-committee / qualification-health language ("economic buyer not yet identified," "single-threaded — the #1 risk," "Compelling Event candidate," "metrics unquantified") — never "activity count = 7," "stage = Discovery," "ownership = PE." Include the one-paragraph qualification-health read a manager wants.
- Every claim in all 8 sections carries exactly one confidence tag. No untagged assertions.
- Every named person traces to a clickable source or is role-only with "title to verify on LinkedIn (rep's manual step)."
- Every gap is honestly flagged with the verbatim "not found in public sources — verify manually" — no padding, no invented names/budgets/vendors.
- Caps held: Play-by-Quarter ≤4, Mutual Close Plan ≤5.
- Vocabulary is sales: opportunity close date (not renewal), prospect (not customer), buying committee.
Present the finished plan and the checklist result, then ask verbatim:
"Section 8 is the bar the competition scores on — every claim tagged, every gap flagged 'verify manually.' Ready to save this as your account's living plan?"
If any item fails, fix it and re-run this gate. Do not deliver a plan that fails Gate 3.
Section drafting notes (honor the fixture, section by section)
Clone the shape and discipline of example-plan.md PART B. Per-section reminders:
- Account Snapshot — a field table (Account / Segment / Units / Geography / Ownership / Territory / Open opportunity / Opportunity close date / Primary contact) + a snapshot narrative naming the deal type ("new-logo opportunity in Discovery," not a renewal). Carry the SF-scope caveat as a blockquote.
- Buying-Committee Map — a role→person→source→confidence table. State plainly how many seats have a name vs. role-only. Call out the economic buyer explicitly (named, or "NOT yet identified — a live gap to close in discovery"). No bulk export produced.
- Pain Points & Triggers — a pain→trigger→source→confidence table. Triggers are what make "now" the right time (expansion, PE recap, hiring). End with a trigger-synthesis line; flag missing turnover/cost figures "verify manually."
- Value Hypothesis — an explicit If … then … hypothesis, tagged [INFERRED — hypothesis, to be validated]. Grounded in the sourced triggers. State what is deliberately NOT claimed (specific %/ROI figures = discovery's job, "verify manually").
- Competitive Position — suspected incumbent from the C4 discipline, tagged [REPORTED], "a lead, not a confirmed incumbent." Give the angle for each branch (confirmed vendor vs. manual/spreadsheets). Never assert an incumbent off one "or similar" mention.
- Play-by-Quarter — quarter→play→aim→working-signal table. 3–4 plays only. Each play advances the deal (land the named exec → multi-thread to the economic buyer → confirm incumbent + quantify → align on value toward the close date).
- Mutual Close Plan — numbered next-step→owner→by table. ≤5 actions. "Mutual" = things WE and THEY each own; at least one step closes the economic-buyer gap rather than papering over it. No fabricated commitments.
- Qualification Translation — THE QUALITY BAR — a raw-signal→translation table plus a one-paragraph qualification-health read. Every row translates a field into what it means (single-threaded, economic buyer unidentified, Compelling Event candidate, metrics unquantified, competition suspected), never restates the field. This is what Gate 3 hardens and what the SKO competition scores.
Also carry a Plan-level Research Gaps block at the end — every unknown as an honest "verify manually," never invented.
Gate 3 checklist (run before delivery — all must pass)
Any unchecked box → fix and re-run this gate. A plan that fails Gate 3 is not delivered.
Quick reference — trigger phrases & gate map
| Intent | Rep says | Skill responds |
|---|
| Start | "Build an account plan for [prospect]" / "run the Cockpit on [account]" | Phase 0 assessment → Gate 1 |
| Confirm snapshot | "looks good" / "proceed" / "that's right" | advance to research + Gate 2 |
| Reshape outline | "reorder these" / "cut section 5" / "add …" | revise, re-present Gate 2 |
| Approve outline | "build it" / "go ahead" | polish → Gate 3 |
| Approve final | "approved" / "save it" | deliver the living plan |
| Skip a gate | "skip ahead" / "just finish it" | state the tradeoff, require explicit accept — never silently comply |
Delivery
On final approval, present the complete 8-section markdown plan and tell the rep:
"This plan is yours — it lives in your Project, it's on your phone tonight, and you edit it before every call. Update it as stakeholders and deal stage shift. LinkedIn name-verification is your manual step."
The optional Fourth-branded HTML deck ("present to your manager") is a post-v1 lunch-session upgrade — do NOT build it here.
What this skill does NOT do
- Does not run web search itself — it delegates to
account-research (which owns the C1–C5 web discipline).
- Does not fabricate names, titles, budgets, headcounts, or vendors — unknown = "not found in public sources — verify manually."
- Does not scrape LinkedIn — that is the rep's manual verification step.
- Does not skip Gate 3 — qualification translation is non-negotiable.
- Does not exceed the caps (Play-by-Quarter 3–4, Mutual Close Plan ≤5) or dump raw Salesforce fields as "the plan."
- Does not answer Fourth product-fact questions (Encyclopedia) or read Fourth internal docs (Brain, not yet live).
- Does not build the HTML/PPTX deck (post-v1 lunch upgrade), export bulk contacts, or touch accounts outside the rep's territory.
