How our pitch changed when our buyers got smarter

When I first started selling Colabra for M&A due diligence, I led with "AI-powered." I thought that was the hook. It wasn't. It was wallpaper. Every software company was saying the same thing at the same time, and buyers had already started tuning it out.

This post is about how our sales pitch actually evolved, based on what buyers told us and what stopped working.

2024: the feature era

In our first year selling into M&A, our pitch was mostly feature descriptions. Automated extraction. Red flag detection. AI-powered insights. Risk register. Data room intelligence.

It wasn't wrong. It just wasn't useful to the buyer. A VP of Corp Dev doesn't care that your tool has automated extraction. They care that their team is drowning in a data room with a six-week clock on it and the first-year associate is going to miss something.

The pitches that converted were the ones where I stopped describing what Colabra does and started describing what the buyer's week looks like today.

August 2025: the conversation that rewrote the pitch

In August 2025, I had a call with a managing director at a private equity fund. He was blunt about how his firm thinks about diligence spend.

He walked me through the two-phase psychology of how PE firms actually decide to spend money on diligence. Phase one is "is this worth pursuing." Low cost tolerance. Fast decisions. The goal is rapid elimination of bad deals before you commit serious resources. He wanted a tool that could "figure it out in one hour instead of one week" for under a thousand dollars a look.

Phase two is "protect the decision we already made." This is after the letter of intent is signed. Now the firm is committed, and the spend tolerance is completely different. Quality of earnings alone runs $250,000 and up. Legal diligence is another large number. Total diligence cost is typically 0.5 to 2 percent of deal value.

I had been pitching Colabra as one product. He heard two products, with different psychology, different price points, and different buyers.

That conversation reshaped how I sell. It's the reason our pitch today leads with "we replace the first week of associate work on a data room." That's the phase one framing. Phase two is how we talk to lawyers and transaction advisors, where the job is protecting a decision, not making one.

The jargon we stopped using

Somewhere along the way I realized that most of the words I was using were Colabra words, not M&A words. But before I get to the swaps, I want to name something more fundamental that I learned from the same set of conversations.

AI is rarely the first concern in an enterprise sales conversation. Security shows up first. Authority shows up first. Implementation pain shows up first. Budget shows up first. AI only becomes relevant when it sharpens an existing priority the buyer already has.

The questions I started asking in discovery changed because of this. Who owns the workflow today? What happens when it fails? Who has authority to buy something new? What gets evaluated by security, and how long does that take? What budget would this come from, and who controls that budget? Why does this matter now, and not six months from now?

Once I started leading with those questions instead of leading with capability, prospects engaged differently. The conversation moved from "let me tell you what our AI does" to "let me understand what's broken in your world today." Every other positioning shift I made flowed from that reframing.

Here are the specific word-level swaps I landed on:

• "Risk register" became "issues list." M&A professionals call it an issues list.
• "Automated extraction" became "reads every document." One describes a mechanism. The other describes an outcome a human can picture.
• "AI-powered" stopped appearing in our outbound emails entirely. It signals nothing now.

A smaller one that surprised me: we used to start cold emails with "Hi [Name], I'm Aoi." I cut it. The sender line already says Aoi. The first six words of an email are too valuable to waste on introductions the reader can see.

Persona-native language

By early 2026, we had three different versions of our pitch for three different audiences. Not three different products. Three different vocabularies.

To PE deal leads, we talk about IC-ready outputs and how much faster a partner can get to conviction. To law firm M&A partners, we talk about associate leverage, write-off reduction, and fixed-fee margin. To corporate development teams, we talk about workstream coordination and integration risk.

I recently rewrote a slide for a law firm audience. The original said "modern deal team" and "standardizing legal excellence." Both were slightly off. "Deal team" is PE and corp dev language. "Standardizing" is threatening to partners who pride themselves on judgment. I changed them to "modern M&A practice" and "repeatable quality." Same meaning. Native words.

The biggest lesson in that rewrite was that words I thought were assets had become liabilities. We mentioned "MCP" (Model Context Protocol) on a slide because it's genuinely important to how our product integrates. A law firm executive has no idea what MCP is and doesn't care. We also had "automated red-flag detection" in the copy. Partners worry about liability when something is framed as fully automated. They want to be in the loop. So we softened it to "AI-assisted." Same capability. Different risk posture in the reader's head.

What changed underneath the pitch

The bigger truth is that buyers got more sophisticated faster than I expected. A prospect I talked to in March 2026 was actively evaluating Copilot, Harvey, and Claude alongside us. Two years earlier, that same role wouldn't have heard of any of those tools. Now he was running a parallel evaluation on three of them.

Another prospect asked me, unprompted, how we differ from Blueflame and Hebbia. That question doesn't get asked in a market that isn't mature. It gets asked when buyers have already done their homework.

This is the shift that matters. You can't sell "AI" anymore because your buyer has already used ten AI products this quarter. You have to sell a specific, concrete job, in their vocabulary, against a named set of alternatives they've already evaluated.

The pitch that works in 2026 assumes the buyer is smarter than the salesperson. That's a different selling motion than the one that worked in 2024. It requires fewer claims and more specificity.

The through-line

If I had to compress the last two years of pitch evolution into one sentence, it would be this: we stopped selling what our product does and started selling what our customer's Tuesday looks like.

That reframing isn't specific to AI. It's specific to selling anything in a market where the buyer has real alternatives and real sophistication. But it's especially true in AI, because the buyer has heard the feature claims already. They want to know whether this one is the one that actually works for them.

The pitch that earns a meeting now is the pitch that shows you've been inside their week. The features are the proof, not the lead.