PreviewContactGet the Book and Resources

Chapter 09

CRM, AI, and the Sales Technology Stack

A rep, running behind, asked an AI tool to draft outreach using just a name and company. The AI invented a plausible detail, I saw your recent European expansion, completely untrue. Sent unread, the prospect replied confused: the company had no European operations at all.

Fig: Unchecked vs. checked: the AI draft, sent two ways

Every beginner hits this wall eventually: an internet full of must-have sales tools listicles, each recommending a different stack, most written by the tools’ own marketing teams or affiliates earning a commission on your signup. Here is the one rule that cuts through all of it: buy tools to solve a bottleneck you have already personally felt, not a bottleneck a blog post told you that you might have.

Technology should make a sound sales system easier to operate. It cannot repair an unclear market, an irrelevant message, weak discovery, or an unsuitable product. AI can accelerate bounded tasks impressively, but speed without evidence, governance, and review simply increases the rate of error and reputational risk.

01

Start with the workflow, not the tool

Ask yourself: what, specifically, have I been doing manually, more than three times, that this tool would do instead? If you can't answer with a specific, repeated, already-felt task, you're buying based on hope, not evidence.

A CRM is your single source of truth, and most free tiers are genuinely usable, not just a crippled trial. Stay on the free tier until at least two concrete signs are true for you personally, like spending several hours a week on manual follow-up a paid feature would handle, or hitting the free tier's limits most months.

Commercial translation: buy tools to solve a bottleneck you've already personally felt, never one a listicle told you that you might have. The free tier of almost any CRM or sending tool is enough until a specific, felt bottleneck proves you've outgrown it.

02

Prioritize bounded AI tasks by risk

The dividing line isn't "is it AI," it's how bad and how hard to undo the failure mode is, and whether a human checks before consequences occur. Good early candidates: research summaries, meeting prep, draft follow-ups that get human review, and role-play practice. Higher-risk and best avoided: autonomous outreach with no human review, pricing decisions, and unreviewed customer commitments.

The single most defensible AI use is a first-pass email draft, personally read, fact-checked, and edited before sending. Never send an AI draft unreviewed; that single habit prevents almost every failure mode this chapter warns about.

Prioritize by risk

How bad, and how hard to undo?

Low riskHigh riskRole-play practiceno review neededMeeting prepquick scanDraft follow-up (reviewed)human reviewAutonomous outreachreview requiredPricing decisionsfull human review

Commercial translation: the riskier the consequence of being wrong, the more human review it needs. Private practice can run with none. Anything a customer will read needs a human checking every fact.

03

Use a control model for every AI workflow

Before any tool touches a real customer interaction, answer four questions: what data is it permitted to see, what data must never be entered, who is the named review owner checking the output before it has consequences, and what happens when the tool is uncertain or wrong.

Score any candidate use case on business value, task structure, error tolerance, data sensitivity, and review burden, then decide to pilot, hold, or reject. A fabricated, unchecked detail sent to a real person is common and entirely preventable with one habit: read before you send, every time.

Unchecked vs. checked

One fabricated detail, sent two ways

“...saw your recentEuropean expansion...”!Sent unreadHuman reviews →catches it, fixes itWe have no EuropeanoperationsThanks for reaching out

Commercial translation: define permitted data, prohibited data, a named review owner, and an escalation path for every AI workflow before the tool ever touches a real customer interaction.

The Signature Framework

Should I use AI for this task?

Interactive Decision Tree

Should I Use AI for This Task?

Will the output reach a real customer unreviewed?

Terms to Know

Mini glossary

System of record

The authoritative store for customer and opportunity data, your CRM.

Human-in-the-loop

A defined human review or approval step within an automated workflow.

Conversation intelligence

Tools that record, transcribe, summarize, or analyze sales conversations.

Applied Exercise

Score an AI use case

Fill in the book's scoring card for one AI tool or task you're actually considering.

Use case: [use case]. Business value: [business value]. Data sensitivity: [data sensitivity]. Decision: [decision].

The Complete Handbook

Everything in this site, plus the rest of the system.

This site gives you one real, working lesson per chapter. The book is the full 130-page reference: every framework, every worked example, and every fillable template in one place.

Full reference for all 10 chapters and appendices
Every template, script, and checklist in one place
Fillable worksheets: Fit Map, Negotiation Prep Sheet, Value Plan, and more
Instant delivery after checkout

Educational product. Not legal, financial, or professional advice.

Chapter 08Chapter 10