AI Sales Academy

How to write prompts that work

Prompting 11 min
Business problem

Most reps type a single sentence into ChatGPT: 'Write an email to an IT director asking for a meeting'. The model returns a bloated, robotic block of corporate prose that no busy executive will ever read. Instead of blaming the technology, reps must recognize that unconstrained models simply sample the most predictable (and tired) corporate phrases from the web.

Proven method

Adopt enterprise B2B prompt engineering: The RCPC Framework (Role, Context, Payload, Constraints).

1. Role (Persona): Define precise executive seniority and cognitive bias (e.g. 'Act as a ruthless automotive procurement director with 15 years evaluating vendor contracts'). Anchoring a persona alters neural token probability and sharpens vocabulary.

2. Business Context: Outline situational background: your product offering, target ICP, diagnosed commercial pain, and the strategic objective of this touchpoint.

3. Payload (Raw Data Input): Supply verified factual inputs inside a designated [DATA] block. Paste call transcripts, proposal excerpts, or financial filings. The model must compute over your data, not extrapolate from ungrounded memory.

4. Constraints & Negative Guidance: Enforce strict operational boundaries: word count limits (e.g. 'under 70 words'), blacklisted buzzwords ('forbidden: synergy, comprehensive, game-changer'), and output formatting (table, bullets, JSON).

Market example
BAD
'Write a follow-up after my meeting.' Output: 3 paragraphs of empty pleasantries, zero business specifics, vague call-to-action.
GOOD
'Act as a Senior Enterprise AE. Context: Yesterday I ran a 30-minute discovery demo with the CFO of a logistics fleet (120 trucks). Prospect fears telematics migration will exceed 2 months. Call Payload: [QUOTE: Our peak freight surge hits in November; system must be fully live by Oct 31]. Task: Draft an executive follow-up (max 80 words) that: (1) Anchors directly to the Oct 31 deadline, (2) Introduces our 30-day fast-track implementation SLA, (3) Asks for 15 minutes Thursday at 10 AM to sign off on the project plan. Zero small talk.'
Ready-to-use prompt
Role: Senior B2B Sales Enablement Strategist.

Context:
- Offering: [WHAT YOU SELL, E.G. AUTOMATED INVOICE RECONCILIATION PLATFORM]
- Recipient: [TITLE & SECTOR, E.G. VP CONTROLLER AT REGIONAL HOSPITALITY CHAIN]
- Objective: [DESIRED ACTION, E.G. SECURE 15-MINUTE TECHNICAL DIAGNOSTIC CALL]

Factual Payload (Raw Data):
[PASTE 2-3 VERIFIED FACTS: PROPERTY COUNT, ACTIVE ERP, RECENT LEADERSHIP CHANGE]

Execution Constraints:
1. Length: maximum 75 words.
2. Tone: peer-to-peer executive dialogue; zero sales jargon.
3. Blacklisted terms: innovative, comprehensive, industry leader, synergy, thrilled.
4. Call to Action: exactly 1 low-friction closing inquiry.

Generate 2 contrasting strategic angles (e.g. operational labor savings vs regulatory compliance risk).
Model rekomendowany: Claude 3.5 Sonnet / GPT-4o Copy and paste into ChatGPT, Claude or Perplexity
ALTERNATYWNE SCENARIUSZE

Variations & alternative scenarios

Reactivation Variant: Prompt to re-engage accounts that stalled after receiving pricing.

Margin Defense Variant: Prompt deconstructing price pushback into total cost of ownership (TCO) arguments.

Practical exercise

Select a one-sentence prompt you used recently. Refactor it into the 4-part RCPC framework. Benchmark the output reduction in required editing time.

Zapisano w przeglądarce

Steps to take

0/4