Build vs Buy Just Became Your Biggest Competitor: What McKinsey's 2026 Data Means for Outbound

The competitor you are losing to does not have a website

If you run outbound in 2026, your win-loss report is lying to you. Not deliberately — it is just missing a category. Somewhere between "went with a competitor" and "no decision," there is a growing bucket of deals where the prospect looked at your product, went quiet, and then quietly had two engineers build a worse version of it in a fortnight.

Build vs buy has always existed. What changed this year is the speed and the volume. And we now have hard numbers on it.

McKinsey's State of AI 2026 survey, published in late August and based on 1,719 executives across 97 nations surveyed between 4 May and 8 June 2026, found that nearly a third of respondents said their organizations opted for in-house AI builds rather than purchasing at least one software product or feature, using agentic coding tools instead. Among the organizations McKinsey classifies as AI high performers, that figure rises to nearly half, against 31% for everyone else. In the technology sector specifically, reporting on the survey put the number at 41%.

That is not a fringe behaviour. That is a third of your total addressable market having already killed at least one purchase decision that would previously have been yours.

This post is about what that does to B2B outbound — which deals die, which loss reasons you are misclassifying, which accounts are now worth more than they look, and how to run signal-based outreach in a market where "we'll just build it" is a credible answer for the first time.

What the 2026 build vs buy data actually says

Three data points, from three different sources, all pointing the same direction.

McKinsey: a third have already skipped a purchase

The headline of the State of AI 2026 report was the ROI gap — only 6% of organizations qualify as AI high performers, attributing at least 5% of EBIT to AI, a figure flat since 2025, and only 37% can attribute any earnings impact at all. That is the number the press ran with.

The buried finding is more interesting for GTM teams. Agentic deployment is scaling fast at the top end: 40% of respondents at organizations above $1B in revenue reported scaling agents in one or more functions, up from 27% the year before. Smaller organizations stayed flat at 22%. And software coding agents are the standout growth area — the specific use case where agentic AI has an obvious, bounded, measurable job.

The economic consequence McKinsey names directly: capital that used to flow to SaaS licence fees is now flowing to inference tokens. That is a shift in where the budget goes, not just how much. It also comes with its own drag — one in five respondents said AI operating costs have actively constrained their use of the technology.

Michael Chui, a senior fellow at McKinsey's QuantumBlack and a co-author, told The Register that the lag between investment and return is consistent with technology history: "History doesn't repeat itself, but it rhymes."

Retool: 35% have already ripped something out

Retool's 2026 Build vs. Buy Shift report, based on a survey of 817 builders and customers, found that 35% of teams have already replaced at least one SaaS tool with a custom build, and 78% expect to build more custom internal tools in 2026.

Two further findings from that report matter enormously for anyone selling into these accounts:

• 60% of respondents have built software outside IT oversight in the past year, and 25% do so frequently. The build decision often does not go through procurement, which means it never shows up in any intent data you buy.
• 51% have shipped production software built with AI that their team actually uses, and about half of those report saving six or more hours a week.

Retool has an obvious interest in this narrative — it sells the platform. Read the numbers with that in mind. But the direction is corroborated by McKinsey, which does not sell an app builder, and by the market itself: in February 2026, an Anthropic model release triggered a selloff in global software stocks on precisely this thesis. Public markets have already priced in some version of "buyers will build."

The buyer was already halfway out the door

Layer this on top of what we already knew about how B2B buying works now. Gartner found that 67% of B2B buyers prefer a rep-free experience, and roughly half of software buyers now start their research with an AI chatbot rather than a search engine or a vendor site.

So the sequence for a modern buyer with a problem looks like this: ask an AI assistant what the options are, get a shortlist, and — this is the new step — ask the same assistant how hard it would be to just build the thing. In 2023 that last question got a discouraging answer. In 2026 it gets a scaffolded repo.

Which categories are actually at risk

Not everything is equally exposed. Retool's report found every SaaS category under some replacement pressure, with workflow automation and internal admin tools leading, followed by CRM, BI, project management, and customer support. But the useful cut is not by category — it is by what makes a product hard to rebuild.

The uncomfortable read: if your product's core value is a nicer interface on data the customer already owns, you are now in a knife fight with a $20-a-month coding subscription. If your value is data they cannot get, risk they do not want to carry, or a network they cannot join alone, build vs buy barely touches you — but your messaging still has to say so, out loud, early.

What build vs buy does to your outbound, concretely

1. You are misclassifying your losses

Most CRMs offer "lost to competitor," "no budget," "no decision," and "timing." None of those is "they built it." So build losses get filed as no-decision, which is the loss reason sales leaders systematically ignore because it looks like weak qualification rather than a competitive defeat.

Fix this first, because it is free. Add a build-in-house loss reason, backfill the last two quarters by asking the reps directly, and look at the shape of it. If it is above 10% of closed-lost, your positioning problem is larger than your pipeline problem.

2. Cold outbound now competes with a free prototype

The average cold email reply rate sits around 3.4% in 2026 and has been falling for years. A generic cold pitch was already a weak instrument. Against a buyer who has a half-working internal version of your product sitting in a staging environment, it is close to useless — because your email is arguing about features, and their objection is about sunk cost and ownership.

This is the part where volume-based outbound gets structurally worse, not just marginally worse. More sequences into a market that has a free alternative does not produce more meetings. It produces more unsubscribes and a worse sender reputation.

3. Renewals are now a competitive front

Historically your renewal risk was a competitor with a lower price. Now a meaningful share of churn is a customer replacing you with something they built — and Retool's finding that 60% of builds happen outside IT oversight means the first sign of this is often not a procurement conversation. It is a champion going quiet, a usage curve flattening, and a new internal tool appearing in someone's LinkedIn post about what they shipped this quarter.

4. Your ICP just gained a variable

Engineering density inside the account is now a scoring input. An 80-person company with 45 engineers behaves completely differently from an 80-person company with 6. The first one builds. The second one buys, and buys faster than it used to, because it has watched its peers waste six months on a build.

If your lead scoring does not include something like engineers-as-a-share-of-headcount, or recent hiring for internal tools and platform roles, you are treating two very different buyers as the same account.

The window: build decisions are made in public

Here is the good news, and it is the reason signal-based outbound matters more in a build-vs-buy market rather than less.

Build decisions are not silent. They are among the most publicly discussed decisions in B2B, because the people making them are engineers and operators who post about their work, ask for advice before starting, and complain loudly when it goes wrong. The decision leaks — you just have to be listening at the right moment rather than mailing the account on a 90-day cadence.

Signals that a build decision is forming

These are the moments where a conversation is still possible, before the repo exists and before anyone's reputation is attached to the build:

• Someone at a target account posting "has anyone built X internally rather than buying?" on LinkedIn, Reddit, or X
• A job posting for an internal tools, platform, or automation engineer with your category named in the responsibilities
• Engineering leaders at the account publicly discussing agentic coding rollouts, AI development budgets, or vibe coding governance
• A procurement freeze or "consolidation review" announced publicly — often the trigger event for a build
• Threads asking for cost breakdowns of tools in your category, which is usually the CFO conversation surfacing in public

Signals that a build has already failed

The second window, and often the better one, opens six to nine months later. Internal builds have a characteristic failure curve: they ship, they work, and then they rot because nobody owns maintenance and the person who built it changed teams. McKinsey's own data hints at the mechanism — the gains from AI stay diffuse and hard to measure unless workflows are genuinely redesigned around them.

Watch for:

• Posts about maintenance burden, "we built this and now nobody maintains it," or internal tool sprawl
• The original builder changing roles or leaving the company (a job-change signal you should already be tracking)
• Public complaints about the exact operational problem your product solves — rate limits, deliverability, compliance, data freshness
• Rising AI operating costs being discussed openly, which is the moment the build's TCO stops looking free

Both windows are the same shape: a specific person, at a specific account, saying something specific and time-bound. That is exactly what warm, signal-based outbound is built for, and exactly what a list-and-blast sequence cannot see. Platforms like Updately exist to catch these moments — monitoring the places where build conversations actually happen, scoring the account against your ICP, and putting a message in front of the right person while the decision is still open rather than three months after it closed.

How to rewrite your outbound for a build-vs-buy market

Reframe from features to total cost of ownership

Feature comparison is the losing frame. An agentic coding tool can match your feature list on a whiteboard. What it cannot do is absorb the ongoing cost of running the thing.

The argument that lands is arithmetic, and you should be willing to do it out loud in a first email: two engineers at fully-loaded cost, for the build, plus a permanent maintenance tax, plus the inference bill, plus the opportunity cost of what those engineers were not building. Against your annual contract. If your number does not win that comparison, you have a pricing problem, not a messaging problem — and it is better to know.

Lead with the parts that are boring to build

Nobody vibe-codes SOC 2 evidence collection. Nobody enjoys building rate-limit backoff, retry logic, deduplication, audit logs, or role-based access. These are the least glamorous parts of your product and, in 2026, the most defensible. Move them from the bottom of your feature page to the first line of your outreach.

Sell to the person who owns the outcome, not the person who owns the build

The engineer who wants to build it is not your buyer and will rarely become one — building is more interesting than procuring. Your buyer is the person accountable for the outcome the tool delivers, who is carrying the risk of a build that slips two quarters. Those are different people with different incentives, and outbound aimed at the builder is outbound aimed at your loudest detractor.

Qualify build risk on the first call

Add one question to discovery: "If you don't buy anything, what happens?" In 2024 the honest answer was "nothing, we live with it." In 2026 a growing share of answers are "we'd probably build something." That answer should change how you run the entire deal — faster, with TCO framing, and with a champion who is senior enough to be measured on outcomes rather than shipping.

Do not fight the build, scope it

The strongest position in a build-vs-buy deal is rarely "don't build." It is "build the part that is differentiated to you, buy the part that is undifferentiated infrastructure." Buyers who have been burned by a stalled internal build respond far better to a vendor who helps them draw that line than to one who insists the line does not exist.

What to change this week

If you take five things from the 2026 build vs buy data, take these:

1. Add a build-in-house loss reason to your CRM today and backfill two quarters. You cannot fix a competitor you are not counting.
2. Audit your product against the risk table above. Be honest about which column you are in, and rewrite your top-of-funnel messaging to lead with your low-risk attributes.
3. Add engineering density to your ICP scoring. Engineer share of headcount and recent internal-tools hiring are now first-class inputs, not nice-to-haves.
4. Start monitoring for build-intent language, not just buying-intent language. "Has anyone built this internally" is a higher-value signal than a pricing page visit, and almost nobody is tracking it.
5. Rebuild your first-touch around TCO and operational burden, not features. The feature argument is the one you now lose by default.
6. Set a nine-month follow-up on every build loss. The failure curve is real, and the second window is usually easier to win than the first.

The broader read on McKinsey's 2026 survey is that most enterprises are spending heavily on AI without moving earnings — 94% of them, by the report's framing. That gap is the same reason internal builds keep stalling: tools get adopted, workflows do not get redesigned, and the value stays theoretical. Which means the build threat is real but not uniformly credible. Some of your prospects will build successfully. Most will build something that works for six months and then becomes a liability nobody wants to own.

Your job in outbound is no longer to be the best vendor on the shortlist. It is to be present at the two moments when the build question is genuinely open — before it starts, and after it breaks — with a message that is about cost of ownership rather than feature parity. That is a timing problem and a listening problem, not a volume problem, and no amount of extra sequence steps will solve it.

Sources: McKinsey State of AI 2026 · Tech Times coverage of the McKinsey survey · Retool 2026 Build vs. Buy Shift report · Reuters on the software selloff · Gartner on rep-free buying · The Register interview with Michael Chui