Deal Intelligence: Transcript → MEDDPICC → CRM Review

What this chain helps you accomplish

Turn a real discovery-call transcript into a defensible deal-intelligence packet: an eight-field MEDDPICC evidence map, a prioritized register of evidenced risks and open questions, one actionable seller next step, and a CRM-ready call record with field-by-field change proposals.

Unlike a generic AI call summary, this chain keeps customer statements, seller-entered CRM facts, interpretations, assumptions and unknowns separate. Each suggested CRM update must cite the underlying transcript line or approved evidence entry and be individually reviewed before manual entry.

The prompts run in your approved AI application; Promptifi provides the instructions rather than recording calls, accessing the customer's CRM, or executing updates for them. A missing fact remains unknown, not a negative answer or a reason to invent a better-looking deal.

When to use it

• Immediately after a substantive discovery, qualification or deal-progress call for an active B2B opportunity, once you have an approved transcript or suitably detailed source-labeled notes.
• Before an internal deal review or manager inspection when MEDDPICC gaps and commitments need to be supported by specific buyer statements.
• When deciding the next seller-controlled action on an uncertain enterprise opportunity, particularly around the economic buyer, paper process, decision steps, competition or Champion validation.
• Before manually updating Salesforce, HubSpot or another CRM after a customer meeting, provided the actual field allowlist and current values are available for safe comparison.
• When a forecast or close-date assumption needs scrutiny without automatically changing the CRM forecast.

Do not use this to manufacture qualification from a thin transcript, infer economic authority from job title, assign a Champion based on friendliness, or auto-advance an opportunity. If the conversation is purely an internal seller meeting, distinguish those statements from buyer evidence; it cannot serve as customer-confirmed MEDDPICC qualification. Use your organization's recording consent, privacy and CRM write policies.

Before you begin

Required and optional tools

Required: an employer-approved text-chat AI assistant capable of reading the permitted transcript or pasted segments, such as ChatGPT, Claude, Gemini or Microsoft Copilot; the real transcript or reliable notes; and a way to review the output. No paid AI plan, browser research, transcription service or live CRM connector is required. CRM field-change proposals require a manually supplied field allowlist and current-value snapshot. A connector, if separately available and authorized, is optional; this chain never performs writes.

For long calls, check your assistant's text/file limits. If it cannot reliably process the full transcript, split it into numbered non-overlapping segments, review each for missing sections, and retain the source references. Do not label a partial call 'complete'.

Exact input placeholders

• {{ACCOUNT_NAME}} — Required. Actual customer organization from approved opportunity data. Example: Summit Fleet Systems (fictional). If uncertain, verify before using the CRM draft.
• {{OPPORTUNITY_NAME}} — Required for CRM association; optional for research-only analysis. Source: actual opportunity record. Example: Service Analytics Modernization (fictional); if missing, mark UNASSOCIATED.
• {{CALL_DATE_AND_TIMEZONE}} — Required for commitment date interpretation. Source: calendar/call metadata, e.g. 2026-11-05, America/Chicago. If unknown, do not convert relative dates or infer a precise date.
• {{ATTENDEES_AND_ROLES}} — Optional. Actual names and roles as known from invitation/transcript; distinguish confirmed from unconfirmed. Example: Jordan Lee, Director of Operations (fictional); otherwise speaker unknown.
• {{SELLER_ROLE_AND_OBJECTIVE}} — Required. Seller's actual role and purpose, e.g. AE clarifying evaluation and next steps; source: seller.
• {{CURRENT_OPPORTUNITY_STAGE}} — Optional, but required to assess stage-specific exposure. Source: current CRM snapshot. Example: Discovery (seller-supplied); if unknown, say NOT VERIFIED.
• {{TARGET_CLOSE_DATE_AND_BASIS}} — Optional. Date and whether buyer-confirmed, seller-entered or unknown; example: 2027-01-31, seller-entered planning date. Never call it a buyer commitment without source evidence.
• {{TRANSCRIPT_SCOPE}} — Required. COMPLETE or PARTIAL, and segment identifiers; source: actual call export. If unsure, mark UNKNOWN rather than declaring comprehensive coverage.
• {{TRANSCRIPT_WITH_LINE_NUMBERS}} — Required. Authorized actual call transcript, speaker tags and L001-style line IDs if available; example: 'L001 Jordan: We review exception reports weekly.' If not available, stop Step 1 rather than reconstructing statements.
• {{EXISTING_CRM_FACTS_WITH_SOURCES}} — Optional. Allowed dated CRM facts, field names, record IDs and as-of timestamps; example: 'Stage=Discovery, snapshot 2026-11-05'. If not accessible, state NOT PROVIDED and do not invent.
• {{KNOWN_BUYER_COMMITMENTS_WITH_SOURCES}} — Optional. Explicit, sourced buyer commitments from prior confirmed messages or the transcript, with owner/date/reference; otherwise none verified.
• {{CRM_FIELD_ALLOWLIST_AND_CURRENT_VALUES}} — Required for Step 4 field-change proposals; optional for earlier analysis. Supply exact permitted CRM object, record ID, field name/API name, existing value and as-of date, and any stage policy. Example: 'Opportunity 00X-example, Next_Step__c=Unspecified (verified 2026-11-05)'. Without this, Step 4 is narrative-only and HOLD; never infer a schema.
• {{AI_DATA_HANDLING_POLICY}} — Required. Organization's approved AI usage, transcript/recording consent, retention restrictions and redaction scope. Example: 'approved enterprise assistant; redact customer personal information; no external sharing'. If permission is unclear, use a sanitized note or do not upload.

Preparation instructions

1. Check that the transcript is permitted to be used with the chosen AI tool, that recording/transcription consent requirements have been met, and that sensitive or unnecessary personal information is removed according to policy.
2. Preserve transcript speaker labels and existing timecodes or line numbers. If needed, add local L001, L002, etc. without changing dialogue. Each claim needs a traceable source location.
3. Separate a copy of current CRM facts from the call transcript and identify who supplied them and when. A CRM number is not a customer quote, and an old CRM value is not automatically current.
4. Record exact CRM field names allowed for proposed changes and their actual current values when ready for Step 4. If you cannot get the list, run Steps 1–3 and produce a narrative-only Step 4 draft.
5. Keep any missing economic buyer, buying deadline, contracting requirement or competitor information UNKNOWN until actual evidence is available.

MEDDPICC uses eight elements: Metrics, Economic Buyer, Decision Criteria, Decision Process, Paper Process, Identify/Implicate Pain, Champion and Competition. For methodological context, MEDDICC's October 2026 FAQ: https://meddicc.com/faqs . The chain supplies its own working definitions and does not require a separate MEDDPICC prompt or paid training resource.

How to run the chain

1. Open a fresh conversation in an employer-approved AI tool. Paste the completed STARTING CONTEXT from Chain Steps, then paste or attach the authorized transcript. If content is too large, segment it with non-overlapping line ranges and confirm full coverage before analysis.
2. Run Prompt 1. Compare every meaningful buyer quote to the transcript, inspect source line references, participant attribution, and all eight MEDDPICC statuses. Correct any invented or ambiguous evidence; approve MEDDPICC_EVIDENCE_PACKET_v1 explicitly.
3. Run Prompt 2 in the same conversation using the latest approved evidence packet. Review the published 0–4/U rubric, the evidential basis of each score, and the separately identified unscored gaps. Approve DEAL_RISK_REGISTER_v1 or repair it before proceeding.
4. Run Prompt 3 with both approved artifacts. Confirm the next-best action is seller-controllable, justified by an approved risk/gap and does not present a tentative buyer commitment as real. Approve SELLER_ACTION_PLAN_v1.
5. Provide the current-value CRM field allowlist and any relevant stage or forecast policy, then run Prompt 4 using all three approved artifacts. If these are unavailable, accept narrative-only notes and leave the change set on HOLD; do not guess fields.
6. Review each proposed CRM change row independently: current value, new value, evidence ID/line, owner, date, amount, stage and object. Choose APPROVE / REJECT / EDIT for each row. There is no 'approve all' and no automatic write, even if the assistant has a CRM connector.
7. Copy only individually approved, reverified fields to your CRM manually, subject to your employer's permissions. Retain unknowns and rejected rows as notes rather than silent field replacements. Record the approved seller next action separately from any buyer promise.

Approval, versioning and context continuity

Use the exact approval pattern 'APPROVED — MEDDPICC_EVIDENCE_PACKET — v1' (similarly for DEAL_RISK_REGISTER and SELLER_ACTION_PLAN). For corrections request a complete v2 and approve that version; earlier versions are superseded. Step 4 field rows must be approved individually and remain UNREVIEWED until your manual decision.

If a new chat is needed, paste only the permitted Starting Context plus latest complete approved artifacts and the source-labeled transcript needed for the next step. Do not rely on access to another chat, CRM or recording. If the same issue produces repeated rejected field changes, or the assistant proposes a forbidden field, STOP Step 4 and return to Step 1 to correct evidence or field scope; do not patch indefinitely.

Completion

You are finished when you have an approved evidence packet, reviewed risks, a realistic next seller action and a CRM review packet with every field change either individually approved, rejected, edited or held. Completion does not imply that a CRM write or customer follow-up occurred.