Your Newest Competitor Works Inside Your Customer: The Build vs. Buy Inversion of 2026

A renewal call I heard about earlier this year went sideways in a way that would have been almost unimaginable two years ago. The vendor was a mid-market workflow tool, about $60,000 a year, embedded in the customer's operations team for four years. Good usage. Friendly champion. The kind of account that gets marked "safe" in the renewal forecast.

The customer's answer was not that the price was too high, or that a competitor had come in cheaper. It was that an ops manager had spent two weekends with an AI coding assistant and rebuilt the 20 percent of the product her team actually used. It ran on their own data warehouse. It did the one thing the vendor never got around to shipping. Nobody in procurement had to approve anything, because there was nothing to buy.

That vendor did not lose to a competitor. It lost to a customer.

For Founders, CROs, Chief Marketing Officers, and Product Leaders at B2B software companies, this is the loss column nobody has a field for yet. Your CRM has picklist values for "lost to competitor" and "no decision." Almost none of them have "customer built it themselves," which means the fastest-growing threat to your renewal base is currently invisible in your own reporting.

The number that should reset your assumptions

Retool surveyed 817 builders in late 2025 and published the results in February 2026. Thirty-five percent had already replaced at least one SaaS tool with something they built themselves. Seventy-eight percent expected to build more of their own tools during 2026.

The second number is the one to watch. This isn't fringe behavior at engineering-heavy startups. Only about a third of the respondents were engineers. The rest came from operations, product, data, marketing and sales ops, IT, finance, and business analysis. Sixty-four percent were senior managers or above. These are the exact people who sit on your buying committees and sign your renewals.

The categories under pressure are equally specific. Workflow automation topped the list at 35 percent, internal admin tools at 33 percent, BI at 29 percent, CRMs and form builders at 25 percent, project management at 23 percent, and customer support at 21 percent. If your product lives anywhere near "we give your team a place to run a process," you are in the blast radius.

Gartner's version of the forecast is more conservative on timing and blunt on outcome: 35 percent of point-product SaaS tools get replaced by AI agents by 2030. Systems of record, platforms with real network effects, and genuinely complex enterprise software carry a lower risk, and will mostly absorb AI rather than get eaten by it. Point tools are the ones on the menu.

The public markets already priced in a version of this. In early February 2026, software stocks sold off hard after Anthropic released a set of Claude Cowork plugins that automate work across legal, sales, marketing, and data analysis. The S&P Software and Services Select Industry Index dropped more than 20 percent. Thomson Reuters fell close to 18 percent in a single session, its worst day on record. JPMorgan analysts called the reaction overblown and built on broken logic. Morgan Stanley was less relaxed. Whichever camp turns out to be right about the multiples, the underlying behavioral change is measurable and it is happening in your accounts right now.

What actually changed, in economic terms

The old build-versus-buy math protected you. Building an internal tool meant a real engineering team, a real roadmap slot, six to twelve months, and a six-figure budget. Buying was faster and cheaper for anything that was not core to the business, which is why the SaaS category compounded for two decades.

AI-assisted development collapsed the "cheap enough to buy" side of that comparison. What used to take ten months and more than $100,000 now routinely lands in about three months for roughly $30,000, and the prototype that starts the conversation shows up in a weekend. Gartner projected back in 2024 that 75 percent of enterprise software engineers would use AI code assistants by 2028, up from under 10 percent in early 2023. That curve moved faster than the forecast.

Two concrete examples from the Retool research make the pattern legible. ClickUp evaluated a wave of AI vendors for its GTM operations, found none with the right integrations, and built six internal tools connected to Salesforce, Zendesk, and Snowflake instead. Their GTM AI product manager put it plainly: they realized they could build the tools themselves and save on multiple subscriptions. The result cut $200,000 a year in automation software. At Harmonic, a $20,000-per-year third-party tool had support so slow that the head of automation decided rebuilding it internally was faster than waiting for a reply. That single frustration turned into a cultural default. The company now runs 33 internal apps, and the first question when someone wants new software is why they cannot just build it.

The Harmonic story is really about a support ticket. The technology just lowered the cost of acting on an irritation the customer already had, and had been carrying for a while.

Why this shows up in your numbers before it shows up in your losses

Here's the uncomfortable part. Most of this activity is happening where you can't see it and, frequently, where your customer's own IT department can't see it either.

Sixty percent of the builders surveyed had shipped something outside IT oversight in the previous year, and 25 percent said they do it regularly. Their reasons were not ideological: 31 percent said they could build faster than IT could deliver, 25 percent said the existing SaaS did not meet the need, and 18 percent said the IT process was too slow. Half of them are running production software built with AI, and among those, 49 percent say the tool saves them six or more hours a week.

So the sequence in an at-risk account looks like this. A department quietly stops expanding. Seat counts flatten, then drop by three or four. Usage narrows to a couple of core screens. Support tickets from that team go quiet. Ten months later, someone tells your CSM the renewal is being "reviewed." By then the replacement has been running in production for two quarters and has its own internal fan club.

None of those early signals fire an alert in a standard health score, because each one individually looks like normal seasonality. The combination is the tell: usage narrowing to a thin slice of your product while the customer's overall headcount stays flat is the signature of someone rebuilding the parts they care about.

The budget context makes it worse. Zylo's 2026 SaaS Management Index puts average annual SaaS spend at $55.7 million per organization, up 8 percent year over year, with median spend of $9,455 per employee and portfolios holding steady at 305 applications. Spend is still rising while app counts have stopped growing, which means the money is concentrating. Vertice's 2026 index puts renewal inflation at 12.2 percent, roughly five times general inflation, and found that 78 percent of IT leaders hit unexpected charges from consumption or AI-based pricing in the past year, with 61 percent cutting projects to absorb the overrun. Every one of those cut projects is a candidate for someone to rebuild internally with a coding agent and a grudge.

The third option in every evaluation

New deals feel the same shift, just earlier in the cycle. For years a competitive evaluation had two real outcomes: pick a vendor, or do nothing. There is now a third, and it often arrives as a working demo built by someone on the buying team before your first call.

That prototype does something specific to your deal, and it isn't usually to kill it outright. It resets the price anchor. Once a director has seen a rough version of your core screen running on their own data, your $80,000 quote gets compared against a weekend of internal effort rather than against your competitor's $95,000 quote. The prototype is worse than your product in every way that matters over three years and better in the one way that matters in the room: it exists, it's theirs, and it cost nothing to make.

The prototype also usually can't survive contact with production, which is where you win these. Roughly one in five builders who tried to ship AI-built software got stopped by hallucinated code or wrong data structures, and a similar share ran out of time or lost the budget. The demo that impressed everyone in the conference room still needs authentication, audit logs, error handling, and someone to own it at 2am.

Two other numbers explain why this argument lands unevenly. Gartner expects up to 40 percent of enterprise applications to include task-specific AI agents during 2026, up from less than 5 percent in 2025, so buyers are being told constantly that agents can do this work. Meanwhile 35 percent of organizations have established no AI productivity metrics at all. Buyers are under pressure to build and have no way of proving whether what they built paid off. That vacuum is an opening for any vendor willing to show up with real measurement instead of a defensive posture about being replaced.

The honest case for why you still win

None of this means buyers are about to build their way out of enterprise software. The same survey that produced the alarming numbers also produced the reasons builds stall, and they are worth memorizing because they are your renewal argument.

Among organizations that could not push AI automation further, 33 percent named unclear ROI, 30 percent named budget, and 26 percent named maintenance burden. On the technical side, 42 percent cited a lack of engineering bandwidth, 41 percent cited security and compliance, and 39 percent cited integration problems between systems. Only 8 percent of builders use AI-generated code without changes, and 44 percent test thoroughly before anything ships. The people doing this work are not naive about it.

The industry benchmarks on custom software make the long game clearer. Ongoing maintenance typically runs 15 to 25 percent of the original build cost every year, and over a system's life, maintenance accounts for somewhere between 50 and 80 percent of total cost of ownership. Complex enterprise builds sit at the high end because of integrations and regulatory change. PMI's 2026 Pulse of the Profession found that roughly a third of complex projects fail, about twice the overall project failure rate.

There is also a live precedent for AI-driven replacement overshooting. Gartner predicted in February 2026 that half the companies that cut customer service staff because of AI will be rehiring by 2027. Replacing something is easy. Living with the replacement is the expensive part, and it arrives on a delay.

So the customer who rebuilt your workflow in a weekend now owns a system with no vendor SLA, no compliance attestation, no roadmap, and a single point of human failure who might take another job in eight months. That is a real argument. It only works if you make it before the build happens, in a tone that sounds like a peer rather than a vendor protecting revenue.

What to change this quarter

Start by making the threat visible. Add "customer built internally" as a closed-lost and churn reason in your CRM, and backfill the last four quarters from your win-loss notes. Most teams I have seen do this find the category was already there, filed under "budget" or "no decision." You can't argue about a number you don't have.

Then reprice the argument around the parts that are genuinely hard to rebuild. Gartner's own framing is a useful filter: what survives is a defensible data layer, network effects, and depth in a specific domain. Generic workflow UI is not defensible anymore. The proprietary benchmark set, the integration you maintain against six vendors' changing APIs, the compliance certification your customer would otherwise have to earn themselves, the model trained on data no single customer possesses, those are the pieces a weekend build cannot reach. If your pitch deck still leads with screens and features, it is arguing on the ground where you are now weakest.

Get honest about total cost of ownership in your sales conversations, using the customer's numbers rather than yours. A build that saves six hours a week and costs a senior engineer 15 to 25 percent of the original build cost annually to maintain is a legitimate decision for some teams and a bad one for others. Sellers who help buyers do that math credibly, including the cases where building is the right call, become the people buyers call before they build, which is the only position from which you can influence the outcome.

Become the thing they build on. The Retool data shows a builder population that wants control over data, security, and permissions, and that mostly uses AI to write pieces of code rather than whole applications: 72 percent generate snippets they integrate themselves, and only 31 percent prompt their way to a complete app. That is a population that will happily build against a good API, an embeddable component, or an agent-ready interface. Vendors who ship those give the internal builder a way to satisfy the itch inside the subscription instead of outside it.

Finally, fix the thing that starts most of these projects. Harmonic's 33 internal apps trace back to a slow support queue. When 25 percent of shadow builds happen because the existing SaaS did not do what the team needed, your roadmap backlog and your support response time have quietly become churn risk. The feature request you have deferred three times is now a build spec sitting in someone's AI assistant.

The part worth remembering

The competitive set for B2B software just expanded to include everyone who uses it. That is a genuinely new condition, and the vendors who handle it well will not be the ones who argue hardest that buyers should not build.

They will be the ones who understand exactly which 20 percent of their product a motivated customer can recreate in a weekend, and who make sure the other 80 percent is where the value lives. If you cannot name that 20 percent in your own product by the end of this quarter, someone in one of your accounts already can.