Before adding AI to GTM, choose the decision you want to improve

AI can generate another sales email, summarise another call or produce another dashboard. The more useful question is whether it helps the commercial team make a better decision or complete an important task more reliably.

Starting with a tool can create activity without a clear business outcome. Starting with a workflow makes the opportunity, the risk and the measure of success easier to define.

Name the commercial problem

Choose a specific problem rather than a broad ambition such as “make sales more productive”. Examples include inconsistent handovers after discovery, slow preparation for account reviews or customer signals that are difficult to consolidate.

Describe who experiences the problem, how the work is currently done and what improvement would matter. If the workflow itself is unclear, map it before adding automation. AI will not resolve conflicting ownership or an undefined process simply by working faster.

Establish a baseline

Observe the current task before running a pilot. How much effort does it require? What errors or omissions occur? How often does someone need to redo the work?

Use measures that reflect the task. For a call-summary workflow, assess completeness and accuracy as well as preparation time. For account research, assess the relevance and verifiability of the information rather than the volume of text produced.

Time saved is not automatically revenue created. Identify what the team will do with the released capacity and check whether that change actually happens.

Set boundaries for data and decisions

Determine what information the tool needs and whether that information may be used with the chosen provider. Review access controls, contractual commitments, retention and applicable privacy requirements before including customer or employee data.

Keep sensitive information out of a pilot unless the organisation has approved its use and the necessary safeguards are in place. Use synthetic or appropriately sanitised examples when testing does not require live personal data.

Separate assistance from authority. A tool may suggest a next action or draft a customer message, but the person accountable for the decision should know what evidence supports it and what needs verification.

Design the human review

“Human in the loop” is not a complete operating instruction. Define who reviews the output, what they must check and when the output must be rejected or escalated.

Illustrative example: AI drafts a post-discovery handover from an approved call transcript. The account owner checks the customer's objectives, commitments, dates and unresolved questions before sharing it with delivery. Missing information stays marked as missing; it is not filled with a plausible guess.

The review should be practical enough to perform consistently. If checking the output takes longer than doing the task properly, revisit the use case or the tool configuration.

Pilot within a clear boundary

Choose a bounded workflow, a small group and an agreed review period. Record the baseline, approved inputs, required review and conditions that would stop the pilot.

Evaluate both benefits and failure modes. Track rework, unsupported statements, adoption and the effort needed to maintain the workflow. Gather feedback from the people receiving the output, not only the people generating it.

Avoid expanding because a demonstration looks impressive. Expand when the workflow delivers a repeatable improvement with acceptable risk and someone is accountable for its ongoing operation.

Make adoption part of the commercial plan

A successful pilot still needs ownership, enablement and feedback. Decide who maintains instructions, who handles exceptions and how changes to the tool or business process will be reviewed.

AI becomes valuable when it strengthens the commercial operating system rather than creating a separate layer of activity. Start with one decision or task worth improving, establish the evidence and scale only what works.

Explore the Revenue Engine Methodology to connect AI initiatives with the wider commercial system.