Our buyer says they'll build it in-house. Do we fight the build or the delay?

Why do buyers keep saying they could just build it themselves?

A VP of Sales sent me a call recording on Tuesday with a one-line note: tell me where this went wrong. His company does $24M in revenue selling labor forecasting and shift scheduling software to third-party logistics providers. Warehouses. The kind of business where being wrong about Tuesday's headcount costs real money by Thursday.

The call is good for thirty-eight minutes. The ops director who brought them in is nodding. Then the prospect's director of engineering, quiet the entire time, unmutes and says: "This is nice. Honestly though, we could probably build this."

The rep did what most reps are trained to do. He had a number ready. Three engineers, nine months, roughly $600K fully loaded, and that's before anybody maintains it. He was polite about it. As far as anyone can tell, he was also correct.

The deal went dark. Five months later the VP checked in with the ops director, who had wanted this badly. Nothing had been built. Nothing was scheduled to be built. Engineering got pulled onto a warehouse management system migration in February and the forecasting project was never staffed at all. She was still building Tuesday's headcount plan in the same spreadsheet she'd been using the day of the call.

Here's what I told him. He didn't lose that deal to a build. He lost it to a sentence, and the reason the sentence beat him is that his entire message was designed to argue with it.

What are you actually losing to when a buyer says they'll build it?

In most stalled build-versus-buy deals, no build ever happens. The alternative that beat you gets a slide, a rough estimate, and a spot on a roadmap, and then it gets outranked by whatever breaks next quarter. Call it the Phantom Build: the internal project that wins the deal and never gets staffed. The buyer's problem stays unsolved, your quarter takes the hit, and nobody inside that company ever experiences it as a decision they made.

Start with why the estimate is unreliable. When you ask an engineering team whether they could build something, you aren't requesting a forecast. You're asking competent professionals to assess their own competence in front of their boss. There's exactly one answer to that question that costs nobody anything to give, and it's yes.

That yes is honest. Most of these teams genuinely could build it. The unreliable part is the second half nobody says out loud, which is "and we will, soon, ahead of the eleven other things on the board."

When your rep opens a cost argument, he's contesting the only part of that sentence that was true. Worse, he's contesting it with the person whose support he needed. Tell a director of engineering his team can't build your product and you've asked the most senior technical person in the room to defend himself. He will. He'll go home and scope it down. He'll come back with a version that does four of your twenty things, and those four turn out to be the ones his ops director actually needed.

The fix starts with treating the Phantom Build as a named rival with a frame of its own, the way How do you run a competitive positioning analysis? treats "do nothing." Right now it sits in most forecasts as a competitive loss, which is the one thing it definitely isn't.

Why is this harder in 2026 than it was three years ago?

The build objection used to be rare enough to handle deal by deal. It shows up in most cycles now, because the thing it references got dramatically cheaper. Google's 2025 DORA report, drawn from nearly 5,000 technology professionals, put AI adoption among software development professionals at 90%, up fourteen points in a single year. Every engineering team your buyer employs ships faster than it did at your last pricing review, and all of them know it.

“Most B2B software was never hard to build.”

... SaaStr, April 2026

That line is uncomfortable and mostly right. The moat in B2B software was rarely the code. It was knowing which problem to solve, in what order, and what goes wrong once real volume hits it. AI took an axe to the first part and left the second one standing.

You can watch the shift in the build vendors' own numbers. Retool's 2026 Build vs. Buy report found 35% of teams had already replaced at least one SaaS tool with something they built, and 78% expected to build more this year. Worth naming the obvious about that figure: the survey went to 817 Retool customers and builders, and Retool sells the thing people build with. Treat it as the loudest available version of the trend rather than its true size. The direction holds even when the magnitude flatters the source.

What this does to your message is specific. Any argument resting on the difficulty of building is depreciating in real time. Knowing what to build, and what it costs to be wrong for two years, is the part that got more valuable. Most B2B messaging is still written for the world where the code was the hard part. It's the same erosion behind How do we defend our market position when AI lowers switching costs?, pointed at your open deals instead of your installed base.

How do you tell a real build from a Phantom Build?

Three tests. None of them need a new tool and all three run on deals you already have. An afternoon, start to finish.

1. 1The Rebuttal Audit. Pull your last five deals where a build came up and read what your rep said next. Mark each answer as cost, capability, or consequence. Cost is the engineer-years math. Capability is the feature list. Consequence is what the problem does to the buyer while it stays unsolved. Most teams come back four or five out of five in the first two columns, which means the message is built to win an argument it should never be having.
2. 2The Named-Person Test. For every open deal where an internal build is the stated alternative, write two things on a line: the name of the engineer who'd be assigned, and the month they'd start. Your champion will usually tell you if you ask plainly. Every blank line is a Phantom Build, and Phantom Builds need a different play than real ones, because you're selling against a calendar rather than a competitor.
3. 3The Month Thirteen Line. Write one sentence describing your buyer's world thirteen months after they choose the build. Not the launch. The maintenance. Who owns it when the engineer who wrote it changes teams, what happens the first time a rule changes underneath it, who answers the phone at 6am. If your message doesn't carry that sentence, your buyer is comparing your product against their own optimism, and optimism has no support burden.

The Month Thirteen Line is the one that surprises people. Ask a founder what happens after a prospect's internal build ships and most have never written it down, which is strange, because it's the exact stretch several of their current customers already lived through before they called.

What do we see across B2B companies losing these deals?

The build objection lands hardest where the product solves a problem the buyer's team already touches every week. Nobody says "we could build a payroll system." Plenty of people say "we could build a forecasting dashboard," because they've built four already and three of them are still running somewhere.

The estimate is the weak joint, and there's now real evidence about how weak. METR ran a randomized controlled trial in 2025 with sixteen experienced open-source developers across 246 real tasks in their own repositories. With AI tools available, they took 19% longer. Afterward, those same developers estimated AI had sped them up by 20%. METR is careful about this and I will be too: sixteen developers working in mature codebases they know intimately is a small, specific sample, and the authors explicitly decline to generalize it to all developers. Read it as a finding about calibration.

Calibration is the whole game here. The team that told your buyer "we could do this in a quarter" believes it. They're not lying to you or to their CEO. Estimating unbuilt software is one of the few professional skills where confidence and accuracy barely correlate, and the tools of 2026 have made the confidence half louder.

The pattern in the messaging is far more consistent than the pattern in the deals. Here's what each argument actually does once it's in the room:

Nothing in the bottom two rows requires a better product than the top three. Same software, same price, same rep on the call. What changes is which person in the room the message is speaking to and what it asks that person to do. Which is also why these deals so rarely get recorded as losses. They get recorded as Why do our demos go great but deals still stall?, filed under no decision and blamed on budget.

What changed when one company stopped arguing about cost?

A $31M company selling equipment maintenance software to food and beverage manufacturers. Plant floor work. Eighteen reps, deals in the low six figures, and a build objection in roughly half of qualified pipeline, almost always from a plant systems group that had already built two internal tools that mostly worked.

Their Rebuttal Audit came back five out of five on cost. Every rep had the same slide with the same fully loaded engineer number on it. Somebody built that slide two years earlier and it had been collecting internal praise ever since for how cleanly it handled the objection.

The Named-Person Test changed the room. Across eleven open deals where a build was the stated alternative, they could produce an engineer's name and a start month on two. Two out of eleven. Nine Phantom Builds sitting in a forecast, every one of them being worked as a competitive threat.

The Month Thirteen Line didn't exist anywhere. Not on the site, not in the deck, not in a single email. They had the raw material for it sitting in their own support queue: four customers who had built exactly this internally before buying, who could describe in detail what happened the first time a regulatory rule changed and the person who wrote the tool had moved to another team.

“We had the month thirteen story from four customers and we'd never told it to a single prospect. We were too busy proving they couldn't build it.”

... The company's VP of Marketing, after her Rebuttal Audit

They changed the argument and left the product alone. The cost slide came out of the deck. What replaced it was the number a plant manager already feels: unplanned downtime hours per line per month, what those hours cost that specific manufacturer, and what happened to that number for four named customers who tried the internal route first. The engineering conversation moved off whether the team could build it and onto what the team would rather be building instead.

Two quarters later the build objection still came up at about the same rate. What changed was what happened next. Deals that raised a build and then went dark for sixty days or more dropped from nine of the prior twenty to three of the next nineteen. Two accounts that had gone quiet on an internal project the year before came back on their own. They also lost a few of those deals faster, which their VP counted as a win. No new capability shipped in that window and nobody is claiming a revenue number.

What should you do this week?

Start with the audit, because it takes an hour and it usually settles the internal argument about whether there's a problem at all.

1. 1Run the Rebuttal Audit on five recorded calls and count the columns. Bring the count to your next pipeline review instead of an opinion. A five-out-of-five in the cost column ends that debate faster than any deck you could build about it.
2. 2Get a name and a month on every open deal where a build is the alternative. Ask your champion directly, in those words. The blanks reorganize your forecast on the spot and they tell you which deals need a calendar play rather than a competitive one.
3. 3Write your Month Thirteen Line this week and source it from a customer who lived it. Call the two accounts who built something internally before they bought from you and ask what year two was like. That conversation is the most valuable sales asset you'll produce this quarter and it costs you thirty minutes.

What all three tests keep pointing at is a document most companies don't have. A Magnetic Messaging Framework is the written narrative every rep, page, and proposal pulls from, and in a build-versus-buy market each of its four anchors has a specific job. Category design decides whether the buyer files you next to a vendor or next to a line in their own engineering backlog, and those are two different comparisons with two different winners. Villain framing puts you and the buyer's engineers on the same side of the table by naming what you're both up against, which is the problem rather than each other. The old-way / new-way contrast carries the shift that makes your whole case: building got cheap, and being wrong about what to build got expensive. The promised-land outcome is the sentence your champion repeats when her VP of Engineering asks why they can't just do this themselves, in a meeting you'll never attend.

Here's why that matters more in this fight than almost anywhere else. The build conversation happens without you. It happens in a Slack thread on a Thursday between two people deciding how to spend next quarter, and nothing you said on a call gets into that thread. Only the sentence your champion can repeat from memory gets in the room. If you haven't written that sentence, she'll improvise one, and hers will be about features, which is the same gap behind How do you equip a champion to sell you to the buying committee?.

PitchKitchen builds Magnetic Messaging Frameworks for founder-led B2B companies in the $5M-$75M range. Founded by Greg Rosner, founder of PitchKitchen and author of Story Craft for Disruptors, PitchKitchen fixes broken marketing messages and underperforming websites for CEOs whose sales are stalling because their message isn't doing the work. If you want a fast read on whether your homepage argues consequence or capability, the Brand Signal Score, PitchKitchen's free homepage messaging diagnostic at pitchkitchen.com/brand-signal-score, will score it in a couple of minutes.