A private equity firm can now clone your product before it finishes reading your pitch deck.
A summary of Financial Times reporting describes how Bain & Company has built hundreds of software prototypes as part of what it calls “outside-in” diligence. The firm uses AI coding tools including Claude Code to recreate acquisition targets’ products in a matter of days. The logic is blunt: if a consulting team can rebuild the visible functionality of a SaaS platform before the term sheet is signed, that platform’s moat may be considerably shallower than the pitch suggested. What survives the clone test, on the other hand, is the part of the business actually worth paying for.
Cloning the interface says nothing about the business behind it
Bain’s practice started as a diligence exercise on a single AI-native healthcare company in October 2025. It’s since become standard procedure at the firm, which tells its own story about how confident buyers have become that most software interfaces are reproducible on demand.
Carl Austin, a managing director at consulting firm Slalom, offered the honest caveat: a clone proves a product’s interface wasn’t especially hard to build, nothing more. It says little about whether the underlying system will hold up at scale, stay secure, pass an audit, or survive contact with a team that has to maintain it for years. Judged only on those grounds, the exercise tells you more about the market than about the target.
That distinction is the one most software companies have spent the last two years discovering the hard way. The features that used to justify a premium multiple are, for a growing number of products, something a consulting team can reproduce over a long weekend.
Building got cheap. Understanding didn’t.
The same commoditization pressure is showing up on the cost side of the ledger, just from the opposite direction.
Uber burned through its entire 2026 AI coding budget in four months, and the company’s own president has publicly questioned whether the resulting surge in AI-generated code is translating into products customers actually notice. Gartner, separately, projects that AI coding costs will overtake the average developer’s salary by 2028. Read together, the two data points describe the same phenomenon Bain’s diligence teams are exploiting: building software has become fast and cheap enough that speed alone stopped being a defensible advantage sometime in the last 18 months.
Dovetail Software puts the underlying shift plainly. “The moat is your customer data, and how well your team turns it into better product decisions,” the company says. “Anyone can ship a feature. Anyone can clone a competitor’s feature. What you can’t clone is a deep understanding of your user, their pain points, and the ability to use it to solve real problems.”
Shipping fast used to prove something. It doesn’t anymore.
Two years ago, shipping a feature fast was itself a signal of competence. A team that could translate customer needs into working software quickly had cleared a real bar. Vibe coding has quietly removed that bar for almost everyone, which means the feature itself no longer proves the team behind it understood the customer at all.
Dovetail Software traces the actual capability shift to something narrower than the coding tools themselves. “You never used to be able to analyze qualitative data at scale,” the company says. “AI made that possible.” That’s the part of the last two years that’s genuinely new: parsing thousands of support tickets, sales calls, and survey responses for the patterns inside them was a human-hours problem before large language models existed, and now it isn’t.
But Dovetail Software is direct about where that capability quietly runs out. An LLM can give the perception that it’s handling large volumes of qualitative data without actually being a system that compounds. It doesn’t build understanding over time, and it doesn’t store what it’s learned so that understanding becomes more usable for whatever the team is trying to solve next. Ask the same general-purpose model to analyze this month’s customer feedback and it starts from the same blank slate it started from last month, no matter how many times it’s answered a similar question before.
Compounding understanding beats a capable model
That gap, between a model that can process customer data and a system that actually accumulates understanding from it, is where Dovetail Software is placing its bet on where value settles once building stops being scarce.
Dovetail’s Sun’s Out launch describes the platform as the always-on intelligence layer across the entire organization. It closes the gap between customer signals and the decisions built on them. The design premise is specific: every survey response, support ticket, and sales call that flows through the system doesn’t just answer today’s question. It sharpens the answer to next month’s question too, because the underlying structure of what customers are saying keeps accumulating instead of resetting with every new prompt.
That’s the exact property Bain’s clone tests can’t detect and a horizontal AI tool can’t replicate by getting a bigger context window. A cloned interface starts from zero. A compounding customer-intelligence system starts from everything the organization has already learned, and the gap between those two starting points only widens the longer both run.
Speed was never the real finish line
None of this makes building fast worthless. Shipping quickly is still how a team tests whether an idea has legs. What’s changed is what shipping quickly no longer proves on its own.
Dovetail Software frames the endgame in terms of divergence rather than a single winner-take-all outcome: one group of companies spending heavily on AI output that never quite translates into revenue, and another group that has made customer understanding a genuine, compounding advantage. The companies in the first group are the ones a private equity associate can clone in an afternoon. The companies in the second group are the ones where the clone works, technically, and still can’t explain why customers keep choosing the original.
This content is provided for informational purposes only and is not a substitute for professional advice. AFP editorial staff were not involved in the creation of this content.
Support AFP
Share
