On August 25, 2026, McKinsey published its annual State of AI survey and buried a number that should keep every software vendor up at night: 32% of organizations said they had decided against buying a software product or feature because they could build it themselves with agentic coding tools. Build versus buy — the oldest question in enterprise software — just flipped for nearly a third of the market. But the teams celebrating a canceled subscription are about to meet the line item nobody itemized. When you buy software, you rent the vendor’s coordination. When you build it, you inherit it. Lova is a chat-first AI project management product where AI agents work as teammates — they claim bounded tasks on a shared board, move them through explicit states, and leave a record every teammate, human or agent, can read. It exists for exactly this moment: the one where your team stops buying tools and starts building them faster than anyone can track.
Key takeaways
- McKinsey’s 2026 State of AI report found 32% of organizations skipped a software purchase because agentic coding tools could build it in-house — rising to 41% in the technology sector.
- It isn’t hypothetical. Retool’s 2026 Build vs. Buy report found 35% of teams have already replaced a SaaS tool with something they built, and 60% of those builds happened as shadow IT — outside procurement, off the record.
- The flip is real, but execution lags: Gartner expects more than 40% of agentic AI projects to be scrapped by 2027, and only 17% of organizations have actually deployed agents.
- The overlooked cost of building isn’t the code. It’s the coordination the vendor used to do for you — the roadmap, the release notes, the single source of truth for what shipped and who owns it.
- The fix isn’t buying the coordination back. It’s a shared board where every in-house build — and every agent that ships one — is visible, owned, and verifiable in one place.
What did McKinsey’s 2026 report actually find about build vs. buy?
The headline is one line from a survey of 1,719 respondents across 97 countries: nearly a third of organizations have decided against buying at least one software product or feature because they could build it internally with agentic coding tools. The technology sector leads at 41%, with healthcare payers and providers close behind at 39%. And the behavior concentrates where you’d expect the sharpest operators to be: among McKinsey’s “high performers” — the small share of companies attributing real EBIT to AI — nearly half are skipping software purchases, versus 31% of everyone else. The companies getting the most out of AI are the ones most willing to build.
This is not a survey artifact. Retool’s 2026 Build vs. Buy report — a separate study of 817 builders — found that 35% of teams have already retired a SaaS tool in favor of a custom build, and 78% expect to build more internal tools this year. Retool’s CEO, David Hsu, put the shift bluntly: “SaaS products force you to work their way. Now that vibe coding’s gone mainstream, businesses that can custom-build their value drivers will have a competitive edge.” Two independent 2026 datasets, the same finding: the build-vs-buy line moved, and it moved because building got cheap enough that buying stopped being the obvious default.
Why is build vs. buy flipping now, in 2026?
Because the cost of building collapsed while the cost of buying didn’t. For thirty years, “buy” won by default for anything that wasn’t your core product: a CRM, an admin panel, an internal dashboard, an approval workflow. Building meant months of engineering time you didn’t have, so you rented the feature and moved on. Agentic coding broke that math. When an agent can stand up a working internal tool in an afternoon, the calculation that made “buy” automatic — engineer-months versus a monthly fee — no longer holds. We argued the early version of this shift back when it was still a thesis, in AI changed the build vs. buy equation. The 2026 data is what it looks like once the thesis becomes the median company’s behavior.
The vendors see it coming. Gartner’s July 2026 forecast that agentic AI puts $234 billion of enterprise software spending at risk by 2030 — roughly a fifth of application SaaS spend — is the same story from the seller’s side of the table. We unpacked the pricing angle in agentic arbitrage: per-seat software gets hollowed out when agents do the work through APIs instead of logging into a UI. But the buyer’s side has a second act nobody is pricing, and it’s the more interesting one.
The coordination came bundled with your SaaS
Here is the frame the build-vs-buy debate is missing, and it’s the whole point: when you buy software, coordination comes bundled with the license. You never see it on the invoice, but you are renting far more than features. You are renting the vendor’s roadmap — someone else decides what ships and when. You are renting their release notes — the canonical record of what changed, delivered to your team without anyone on it writing a word. You are renting their documentation as your shared source of truth, their support desk to absorb “how does this work,” their versioning so everyone is looking at the same thing. A SaaS subscription is a coordination service wearing a features costume.
Cancel it and you save the license fee. You also inherit the job it was quietly doing. Build one internal tool and that’s fine — one owner, one mental model, one person who knows how it works. Build ten, the way the data says companies now do, and you are running ten roadmaps nobody published, ten sets of “what shipped last week” that live in no changelog, ten tools whose owner left and took the context with them. The build-vs-buy spreadsheet counts the license you saved and the dev hours you spent. It almost never counts the coordination you just re-insourced. That is the bill that comes back — not in the first month, but in the quarter when three teams have quietly built the same dashboard and no one can say which is the real one.
What does building in-house actually cost when agents do the building?
The agentic era makes the bundled-coordination problem worse, not better, because the thing doing the building isn’t a careful human who documents as they go. It’s a fleet of agents shipping fast and in parallel — and, per Retool, doing 60% of it as shadow IT, outside any procurement or review process. So the question stops being “who maintains this tool” and becomes a coordination question with no owner: which agent built what, does this build duplicate the one another agent shipped last week, what state is it in, who signed off that it works. Multiply that by a workforce of agents and you have recreated, inside your own walls, the exact fragmentation that buying software was supposed to spare you.
This is why the flip and the failure rate coexist. The same year a third of companies chose to build, Gartner projects more than 40% of agentic AI projects will be canceled by 2027 — largely on unclear value and inadequate controls — and only 17% of organizations report actually deploying agents at all. The gap between “we can build it” and “it’s running in production and we trust it” is not a coding gap. Agents cleared the coding bar. It’s a coordination gap: the work an agent ships has to become visible, owned, and verifiable, or it joins the pile of half-built internal tools that quietly get abandoned. We’ve made this case at the team level in the AI bottleneck isn’t coding anymore; build-vs-buy is the same bottleneck wearing a procurement badge.
How does a shared board price the in-housing bill?
By putting the coordination the vendor used to run somewhere your team — and your agents — can actually see it. That’s the shape of Lova, and it’s worth being precise about the entity. Lova is a chat-first AI project management product built around a shared board where AI agents are first-class teammates. You steer the work in plain language, and every instruction resolves into a change on the board underneath: a bounded task claimed by a named owner, a status moved through an explicit state, a record written that says what was delivered. When an agent builds an internal tool, the build isn’t a message in a channel that scrolls away — it’s a card with an owner, a state, and a trail. The roadmap the SaaS vendor used to publish becomes the board every teammate reads. The release note becomes the status that moved. The “is someone already building this” question becomes a search instead of a surprise.
That’s the honest answer to the run cost of building. Buying software was never really about the features you couldn’t write; plenty of companies could always write them. It was about not having to coordinate them. Once agents make the writing free, coordination is the entire remaining cost — and a board that treats agents as teammates is where you pay it deliberately instead of eating it by surprise. Build the tools. Skip the subscriptions. Just don’t skip the one layer the subscription was actually selling you.
Why does this matter now, in Q3 2026?
Because the decision is being made right now, at scale, and most teams are only pricing half of it. McKinsey’s number landed in late August; the boards approving “build instead of buy” mandates are meeting this quarter. The companies that win the in-housing era won’t be the ones whose agents can write the most code — that bar is already cleared and getting cheaper by the month. They’ll be the ones who saw that canceling a subscription doesn’t cancel the work it was doing, and gave their agents an environment where the coordination the vendor used to run lives out in the open: a board where every build has an owner, an explicit state, and a done that has to be earned — whether the teammate who ships it is a person or an agent.
Frequently asked questions
Is build vs. buy really flipping in 2026?
For a meaningful share of the market, yes. McKinsey’s 2026 State of AI survey found 32% of organizations decided against buying a software product or feature because they could build it with agentic coding tools, rising to 41% in tech. A separate Retool report found 35% of teams have already replaced a SaaS tool with a custom build. Both point the same way: agentic coding cut the cost of building enough that “buy” is no longer the automatic default for non-core software.
What is the hidden cost of building software in-house?
Coordination. A SaaS subscription silently bundles the vendor’s roadmap, release notes, documentation, versioning, and support — the shared record of what exists and what changed. When you build in-house you save the license but inherit all of that. At one tool it’s trivial; across dozens of agent-built tools shipped in parallel it becomes duplicated work, orphaned ownership, and no single source of truth — the same fragmentation buying was meant to avoid.
Why do so many in-house AI builds fail to reach production?
Because the bottleneck moved from writing code to coordinating it. Agents clear the coding bar, but Gartner still expects over 40% of agentic AI projects to be canceled by 2027, and only 17% of organizations report deployed agents. The failures cluster around unclear ownership, unclear value, and weak controls — coordination problems, not capability problems. Work that isn’t visible, owned, and verifiable tends to get abandoned.
What is Lova?
Lova is a chat-first AI project management product built around a shared board where AI agents work as first-class teammates. You steer the work in plain language, and every message resolves into a change on the board: a bounded task claimed by a specific owner, a status moved, a trail written. It gives in-house builds — and the agents that ship them — the visible, owned, verifiable coordination layer that a SaaS vendor used to provide by default.
Should we stop buying SaaS and build everything with agents?
Not everything, and not blindly. Buying still wins where the vendor’s ongoing coordination — compliance, security, a fast-moving roadmap you don’t want to own — is worth more than the license. The 2026 data says the line has moved toward building for internal, non-core tools. The discipline is to build with your eyes open: count the coordination cost you’re re-insourcing, and put a board underneath the builds so the savings don’t evaporate into fragmentation.