Commercial Due Diligence with AI-Moderated Expert Interviews

Commercial due diligence is an interview problem before it's an analysis problem.

A typical CDD engagement for a mid-market PE deal calls for 20–40 expert interviews over a 2–3 week window. Strategy consultancies running CDD for a corporate acquirer face the same math. And in both cases, the interview targets are more ambitious than the execution plan — because the execution plan is always "a senior associate runs calls in between everything else they're doing."

That's the real commercial DD problem. Not sourcing. Not analysis. Execution capacity on the interviews themselves.

This post is a methodology for running commercial DD at interview scale — the six interview types that matter, what each one is supposed to capture, and the concrete week-by-week timeline that makes a 25-interview engagement actually land on schedule. AI-moderated interviews show up in the timeline because they're what makes the arithmetic work, but this isn't an AI pitch. It's a primary-research playbook that happens to include AI as a capacity multiplier.

The six interview types that matter

Most commercial DD interview lists look like a jumble of "people who might know something." The engagements that land well sort interviews into six types, each with a specific job. Miss a type and the CDD deck will have a hole the partner will notice.

1. Current customers

Why you need them: Pricing, expansion behavior, perceived alternatives, churn signals. Current customers are the closest thing you get to a ground-truth read on the target's revenue story.

Typical count per engagement: 6–10 calls. You want representation across customer segments, tenures, and sizes.

Data points to capture:

• Realized price (including discounts, credits, contract terms)
• Year-over-year spend change
• Expansion intent (more seats, more modules, cross-sell)
• Who they evaluated before buying the target's product
• What would make them switch, and how likely they are to

The signal that changes the thesis: Customers who describe the target's product as "the one we happened to pick" instead of "the one we picked because X, Y, Z." Commodity positioning kills multiples.

2. Former customers

Why you need them: The churn story you hear from management is almost always wrong. Former customers give you the real reasons, the real replacement cycle, and the real bill of materials for winning them back.

Typical count per engagement: 3–6 calls.

Data points to capture:

• Why they left (specific, not "service issues")
• Who they switched to, and for what specific reasons
• Whether the switch was driven by cost, capability, or relationship
• Whether they would consider coming back, and under what conditions

The signal that changes the thesis: A cluster of former customers who all left for the same competitor in the same quarter. That's a product gap or a sales-team breakdown, and it shapes the post-close value-creation plan.

3. Distributors and channel partners

Why you need them: In any deal with indirect revenue, channel partners see inventory, reorder patterns, and competitive shelf dynamics that management can't fabricate.

Typical count per engagement: 4–8 calls, depending on channel complexity.

Data points to capture:

• Inventory turnover by SKU
• Sell-through vs sell-in (the single most important number in a consumer CDD)
• Competitor shelf share and terms
• Any pending changes in the channel relationship (co-op funds, slotting, exclusives)

The signal that changes the thesis: Channel partners who are consolidating around a competitor's brand or migrating to a private-label alternative. That's usually the first leading indicator of a margin squeeze nobody in management has felt yet.

4. Competitors' customers

Why you need them: To validate win/loss stories the target's sales team is telling, and to find feature gaps and pricing pressure the target hasn't admitted in management meetings.

Typical count per engagement: 4–8 calls.

Data points to capture:

• Why they picked the competitor over the target (specific, evidenced)
• Where the target showed up in their evaluation
• Whether they considered the target at all, and why not if so
• What the competitor charges relative to the target

The signal that changes the thesis: The target never appears in the consideration set. That means the target's win rate number is selection-biased — they're only winning where they get evaluated, and they're not getting evaluated often.

5. Former executives and sales leaders

Why you need them: The org health story, the strategy execution story, the competitive landscape story, and the "what's really happening inside" story. Management will tell you the target has a great sales team. Former sales leaders will tell you whether that's true.

Typical count per engagement: 3–5 calls.

Data points to capture:

• Sales org quota attainment patterns
• Named competitive threats management didn't flag
• Internal politics that shape execution
• What they would change about the business if they came back

The signal that changes the thesis: Three former execs independently naming the same competitive threat or internal dysfunction. That's a signal the target's management is either blind to or covering something up — either is a reason to re-price.

6. Industry SMEs, analysts, and regulators

Why you need them: Market sizing, TAM, regulatory risks, demand-side trends, adjacent market shifts that will affect the target over the hold period.

Typical count per engagement: 2–4 calls.

Data points to capture:

• Market size and growth rates by segment, with methodology
• Regulatory changes pending or under discussion
• Adjacent market dynamics that will reshape demand
• How the target's positioning reads from outside the company

The signal that changes the thesis: An SME who tells you the target's addressable market is a fraction of what management is claiming, or that a pending regulation will reshape the demand curve during the hold period. Either can move IRR by hundreds of basis points.

What to capture in every commercial DD interview

Across all six interview types, the same three things should show up in every completed transcript. If they don't, the case team is leaving value on the table.

The 5 numbers every case team needs:

1. Realized or estimated price (the number, not "fair value")
2. Year-over-year volume or spend change (percentage, with direction)
3. Churn or retention rate (percentage, specific to the cohort)
4. Net Promoter proxy ("would you recommend...")
5. Share-of-wallet (competitive share within the buyer's category spend)

The 3 quotes every case team wants for the deck:

1. A market-assessment quote (where the category is going)
2. A competitive-positioning quote (how the target compares to named alternatives)
3. A product-differentiation quote (what the target actually does well, in the customer's own words)

The 1 signal that changes the investment thesis: Positive (hidden moat, unexpected pricing power, expansion optionality) or negative (customer concentration, regulatory overhang, sales-team dysfunction). Every interview should be evaluated for whether it surfaced one.

If your interview script doesn't reliably produce these outputs, rewrite the script before you run the engagement. AI-moderated or human-moderated, the quality of the script determines the quality of the deck.

(For a broader view of how primary research fits into consulting workflows, see our playbook for consulting Insights Centers.)

The execution math problem

Here's the arithmetic every CDD lead has run and every CDD lead has pretended not to notice:

• Interview target: 25 calls
• Engagement window: 14 business days
• Senior associate capacity: 3–5 interviews per week, including synthesis
• Associates on the engagement: 1, sometimes 2

The ceiling is 10 interviews in two weeks with one associate, 15–20 with two. The interview target is 25. The math doesn't work.

The traditional workarounds all have costs:

• Cut the target to what the team can run. You ship a CDD deck with 10 calls behind it instead of 25, and the partner asks why you didn't talk to more channel partners.
• Escalate to expert networks. You add $25K–$75K in per-hour credits to the engagement, and the quality varies by expert the network hands you.
• Add another associate. Best case, staffing pushes the engagement start by a week.

AI-moderated expert interviews change the arithmetic because the capacity unit shifts from human hours to parallel execution. A single Insights Associate can coordinate 15+ AI-moderated interviews in the time they used to run 3. The target of 25 becomes feasible on the original team without any escalation. See how AI voice agents are reshaping commercial due diligence for the market context.

A week-by-week timeline for 25 interviews in 2 weeks

Here's what the execution actually looks like when you have AI-moderated interviews in the workflow.

Week 1, Days 1–2: Scoping and script design.

• Lock the interview target list with the case team (all 6 types represented)
• Draft one script template per interview type (customer, former customer, channel, competitor customer, former exec, SME)
• Define the 5 numbers + 3 quotes + 1 signal deliverable for each type
• Pre-sourcing of the expert contact list

Week 1, Days 3–5: First wave of interviews.

• Run 8–12 interviews across customer, former customer, and channel types
• Associate does NOT run these personally — AI runs them in parallel while the associate stays on script refinement and synthesis
• Transcripts land within hours; associate reads each and tags the 5 numbers + 3 quotes + 1 signal

Week 2, Days 1–3: Second wave + gap-fill.

• Run 10–13 interviews across competitor customer, former exec, and SME types
• Review first-wave findings and identify gaps — which of the 5 numbers didn't land, which signals need confirmation
• Schedule 2–3 gap-fill interviews with targeted experts

Week 2, Days 4–5: Synthesis and deck drop.

• Associate synthesizes across all 25 transcripts into the deck's commercial section
• Case team reviews the signals against the investment thesis
• Partner gets the CDD output with 25 interviews behind it instead of 10

The math works because the interview-running step — which used to be serial and capacity-constrained — becomes parallel and schedule-independent. The Insights Associate's work shifts from running calls to reading transcripts and writing findings.

What NOT to run via AI

The six interview types above are all AI-compatible. Three adjacent interview types are not:

Management team interviews. These are relationship-building calls, and the Partner's presence signals commitment. AI doesn't belong in a room where the client is evaluating whether the CDD team understands their business.

Sensitive regulatory or legal interviews. Where nuance matters — regulatory posture, pending litigation, compliance exposure — you need a human who can read the room, push where it matters, and back off where it doesn't.

Interviews the firm wants to own as a future relationship. If an expert could become a repeat source across future engagements, a human should build that relationship.

The rule: automate the capacity-bound interviews, protect the judgment-bound ones.

What changes in the CDD workflow

Three things shift when AI-moderated interviews enter the mix:

1. Associate time moves from running interviews to synthesis. The highest-leverage work in CDD was always the analysis layer. Now that's where associate hours live.
2. The CDD deck becomes more quantitatively defensible. Running 25 calls instead of 10 means the 5 numbers have real confidence intervals around them. The deck can say "73% of current customers we spoke to reported price increases above 8%" instead of "customers generally reported price increases." The partner stops asking how you know.
3. The partner can pressure-test hypotheses in real time. When the partner asks "did you test whether customer concentration is a risk," the answer is "yes — here are the 6 accounts that make up 40% of revenue, and here's what each told us about their contract cycle." That's the deck the client wants.

Getting started

If you're about to run a commercial DD engagement and want to try this without restructuring your firm's process:

1. Pick your next CDD. Not a current one — give yourself a week of setup.
2. Lock the target interview list at 25 (or wherever your scope calls for). Don't pre-cut for capacity. We're trying to hit the real number.
3. Draft the six script templates for the six interview types. Keep them tight — 8–12 questions each.
4. Run the first wave with AI-moderated interviews and compare output quality to your normal human-run interviews. Real-world direct comparison.
5. Synthesize and ship the CDD deck. Debrief afterward with the case team: did the larger interview sample change the findings?

The point isn't to prove a tool works. The point is to hit interview targets that used to be aspirational and see what that does to the quality of your CDD deliverables.

---

InsightAgent runs AI-moderated expert interviews for strategy consultancies, PE deal teams, and direct investors. See how it fits into consulting commercial DD and PE diligence workflows.