MMarketing Against The Grain
← All frameworks
Sales

Pattern-Matched Sales Deck

Classify each prospect and select the selling points proven for that type

Difficulty
Advanced
Time to result
~months to results
Steps
7
Confidence
92%

Insert a reasoning step between call capture and deck generation. Feed an LLM the prospect’s current conversation, known company characteristics, CRM history, and a set of predefined customer-type rules. The model first determines which type or pattern best describes the company. It then selects the positioning, proof points, and selling arguments that have historically worked for similar buyers and converts the resulting brief into a custom deck. This is distinct from simple transcript summarization: the output reflects both what the prospect said and what the organization has learned across previous deals. The method requires explicit segmentation logic, maintained sales playbooks, evidence-backed mappings, and a human check when classification confidence is low.

Origin

Kieran Flanagan proposed adding Claude between call notes and Gamma to recognize company patterns and select the selling points that had closed similar customers. Extracted from Marketing Against The Grain.

Core principles

  • 01Classify before personalizing
  • 02Use historical wins to choose selling points
  • 03Combine current-call context with account history
  • 04Encode sales knowledge as reusable decision rules
  • 05Keep human judgment over uncertain classifications

How to run it

  1. 1

    Define prospect types

    Identify recurring company categories that differ in needs, buying behavior, or response to particular messages.

    Pro tip Base categories on observable business characteristics rather than vague personas.

    Watch out Overlapping or subjective categories will produce unstable classifications.

  2. 2

    Map winning messages

    For each type, document the selling points, evidence, objections, and deal patterns that have worked historically.

    Pro tip Use closed-won and closed-lost evidence rather than representative intuition alone.

    Watch out Historical correlation does not guarantee that a message caused the win.

  3. 3

    Assemble account context

    Combine the latest transcript with CRM records, prior calls, account attributes, and relevant internal guidance.

    Pro tip Clearly label the source and date of each piece of context.

    Watch out Stale CRM information can send the model toward the wrong playbook.

  4. 4

    Classify the company

    Ask the reasoning model to identify the best-fitting company type and explain the signals supporting its choice.

    Pro tip Require an uncertainty score and allow an unknown classification.

    Watch out Forcing every prospect into a type creates false confidence.

  5. 5

    Select the sales playbook

    Use the classification to retrieve the matching selling points, proof, and objection-handling guidance.

    Pro tip Keep the playbook as controlled input rather than asking the model to invent sales policy.

    Watch out Do not use sensitive outcomes or protected attributes as classification signals.

  6. 6

    Generate the custom deck

    Build the presentation from the account context and selected playbook, emphasizing the prospect’s stated priorities.

    Pro tip Separate account-specific evidence from general claims visually.

    Watch out Do not imply that a general customer pattern proves this prospect will behave the same way.

  7. 7

    Validate the match

    Have the account owner review both the classification and the resulting deck before use.

    Pro tip Capture corrections to improve the classification rules over time.

    Watch out A visually convincing deck can conceal a poor underlying pattern match.

In the wild

Enterprise security buyer

A transcript and CRM record show a regulated enterprise with a lengthy procurement process and repeated security questions. The classifier maps it to the regulated-enterprise playbook, emphasizing audit controls, implementation governance, and relevant customer proof rather than the self-service features used for smaller buyers.

The proposal focuses on the buying criteria most likely to matter to that company type.

Fast-growing software company

A prospect is classified as a scaling SaaS business based on headcount growth, fragmented workflows, and prior calls. The deck prioritizes time-to-value, integrations, and examples from similar growth-stage customers.

The salesperson presents a more relevant argument than a generic feature tour.

Common mistakes

Inventing segments after the call

Company types should be defined and tested across historical evidence, not improvised to justify a preferred pitch.

Using superficial personalization

Changing logos and company names does not constitute pattern matching if the underlying sales argument remains generic.

Automating uncertain classifications

Low-confidence or novel accounts should fall back to human analysis instead of receiving a forced playbook.

Is it for you?

Best for

Sales organizations with enough reliable CRM history to identify recurring buyer types and winning messages.

Not ideal for

Early-stage teams with little historical evidence or businesses where every deal follows a genuinely unique buying process.

From the transcript

if you know certain characteristics of a company, you could actually have uh Claude pre-programmed with, well, tell me if it's a company like this.

Kieran Flanagan · 08:30

you can actually do some sort of pattern recognition in that Claude step

Kieran Flanagan · 08:30

you can construct the deck to completely pattern match to sell exactly to that customer in the way that you've know that customers like that…

Kieran Flanagan · 09:00

From the episode

This AI Workflow Turns Every Sales Call Into a Custom Deck