How to Qualify an Opportunity With AI

Assemble an evidence ledger

Collect meeting notes, emails, public account information, product constraints, prior interactions, and agreed next steps. Remove information the team is not authorized to process. Give each item a source label and date. Then classify every extracted statement as direct fact, stakeholder statement, inference, or unknown.

This approach mirrors basic risk-management discipline: establish context, define intended use, document assumptions and limitations, and identify human oversight. The NIST AI Risk Management Framework describes mapping context, goals, risks, and knowledge limits rather than treating model output as self-validating (AI RMF overview, AI RMF Core). The framework is voluntary and is not a sales method, but its evidence categories are useful here.

Ask AI to identify contradictions and blank fields, not to score the account automatically. A blank authority field means authority is unknown, not absent. A positive comment means a person expressed interest, not that a funded decision exists.

Sources: [1] [2]

Structure the qualification decision

Build the brief around five questions: What problem is evidenced? Who experiences it? Who can authorize change? What timing or trigger exists? What next step would create stronger evidence? Add fit constraints such as data sensitivity, geography, deployment requirements, or product boundaries.

For each answer, include the source ID and confidence rationale. Use a confidence label only as shorthand for evidence quality, never as a probability of closing. Ask AI to generate the smallest set of questions that could change the next-step decision.

Worked qualification brief

Suppose a synthetic enablement lead says, M04, “We have 40 people finishing training in November, and leadership wants evidence beyond attendance.” A public programme page, P02, confirms the cohort exists. No source identifies budget ownership. Product notes, C01, show that only synthetic material can be used in an initial exercise.

The brief records:

The weak version says, “Strong opportunity: decision-maker has urgent need and a 40-seat deal.” That invents authority, urgency, and commercial scope. The repair removes the forecast language and replaces it with questions whose answers would change the decision.

• Evidenced need: the lead wants post-training evidence beyond attendance, based on M04.
• Timing: a November cohort, supported by M04 and P02.
• Authority: unknown. The lead owns programme delivery, but budget authority is unverified.
• Fit constraint: a first exercise must use synthetic material, from C01.
• Next action: confirm the decision process and test willingness to review a bounded synthetic pilot design.

Use AI for challenge, not optimism

After drafting, run a red-team prompt: “For every positive conclusion, show the strongest missing fact or contrary interpretation.” Then ask whether the recommended next step is proportionate. A discovery call may be appropriate where authority is unknown; a proposal may be premature.

Keep lost or paused opportunities in the same evidence model. Comparing what was assumed with what later happened can improve question design, but a small internal sample does not establish a predictive scoring model.

Verification checklist

The output is a qualification brief, not an automated buying-intent score or revenue prediction.

• Every material statement has a dated source or is labelled assumption or unknown.
• Stakeholder language is not silently upgraded into an organizational commitment.
• Need, authority, timing, and fit remain separate.
• Contradictions and disconfirming signals are visible.
• The next step gathers evidence needed for a real decision.
• Sensitive data is handled under the organization’s approved process.
• A human account owner reviews the brief before it informs external communication.