91% Believe Epic. 64% Would Still Buy From a Startup.

Ninety-one percent of executives at Epic-based health systems say they have complete or significant confidence that Epic will execute on its AI roadmap and match the best external solutions on the market. Sixty-four percent of those same executives say they're likely or very likely to buy from a net-new startup. And non-profit community systems were nearly twice as open to startups as for-profits, 47% very likely versus 24% [1].

All of that is true at the same time. Founders who read only the first number raise a smaller round than they should. Founders who read only the second number spend eighteen months building toward a budget that was never allocated.

The numbers come from Redesign Health, which surveyed 112 senior leaders — CEOs, CIOs, CMIOs, vice presidents — at Epic-based systems across the US, fielded October 2025 through March 2026. Seventy-one percent call their organization "Epic-first." Eighty percent expect that bias to harden over the next three to five years. Forty-three percent said AI purchases land in a department or service-line budget first, before they ever migrate to central IT [2].

Read as a market map, the report says Epic is winning. Read as a template, it says something more useful to anyone with a product roadmap: almost everything that determines whether you get bought is knowable before you build. We keep meeting teams who ran that test after the Series A, when the answer costs a rewrite instead of a month. Healthcare is unforgiving about sequence.

Here's the diagnostic we'd run, in the order we'd run it.

Find the budget line before you find the feature

If 43% of AI dollars come out of a department or service line, the buyer you need may not be the CIO whose logo you've been chasing. It's the service-line VP who owns a P&L, a target, and a quarterly review with the CFO. Two calls will tell you whether that person exists for your product. Ask them which budget code this comes out of in the current fiscal year, and what gets cut to fund it. If the answer is "we'd have to create a line," you've just learned your first sale is a budgeting exercise. Budgeting exercises run three quarters longer than every model we've seen.

Get the ROI hurdle in numbers, from three buyers

Forty-nine percent said an external vendor has to show significantly higher ROI or better outcomes than Epic to win. Only 18% grade Epic and its competitors on the same scale. Sixty-three percent ranked easy integration as important, 59% faster time to value [3]. "Significantly higher" isn't a number until a buyer says it out loud. We've watched a founder discover in month fourteen that his buyer meant 3x with a twelve-month payback, on a product engineered to deliver a 20% improvement. Make three buyers put a figure and a timeframe on it before you scope the build. If none of them will, that's data too.

Decide which door you're walking through

Redesign's read is that startups still get bought when they solve a specific problem or prove strong ROI, or when Epic is lagging in a category [4]. Pick one. Say it out loud, then stress-test it with the only question that reliably gets an honest answer: if Epic shipped an 80% version of this in eighteen months, would you rip me out or keep me? Ask five CIOs. Nobody's contract is on the line yet, which is exactly why the answers are better than any win-loss review you'll run in year three.

It's worth knowing what encroachment actually looks like when it happens. Epic released a surgical block optimization module that walked directly into what Qventus and LeanTaaS were doing, and it changed how CIOs approached those contracts. Neither company got crushed. Both built on the position they already held inside health systems and kept adding capability [5]. Encroachment is survivable. Encroachment discovered late, with no relationship and no proprietary data, is not.

When the receptivity read comes back negative, repositioning is possible. It just isn't cheap. TytoCare spent years in episodic telehealth and is now moving into high-acuity chronic disease management, embedding FDA-cleared algorithms into CHF, COPD, and oncology monitoring pathways. That move arrived with a $25M-plus growth round led by Insight Partners and a new CEO, Adam Pellegrini, who came out of CVS Health, Fitbit, and Walgreens [6], alongside deeper integration work with health systems and payers [7]. That's the price of a second act: growth capital plus a leadership change. Most companies get neither funded.

The cheapest receptivity research available is hiring someone who has sat on the other side of the table. STAT reported a wave of executive departures at Epic this summer [8]. When people leave the dominant platform, the market's actual purchasing logic becomes hireable — as an operating exec, an advisor, a board member. A team with an ex-Epic implementation lead in the room learns in a week what a year of discovery calls won't surface.

Now the counter-argument that actually hurts. Stated intent is generous, and this kind of research can talk you out of a good company. The 64% who say they'd buy from a startup sit inside organizations that largely can't absorb what they buy: in Qventus's 2026 CIO research, only 4% of health system technology leaders said they'd scaled AI with measurable outcomes, 45% cited difficulty scaling pilots, and 74% named EHR vendor dependency as a top execution barrier [9]. A survey captures preference in a moment. It doesn't capture the CFO who kills the contract in month nine, after the pilot has already cleared clinical. And it's blind to the market that doesn't exist yet — the founders building ahead of demand are precisely the ones this exercise would discourage. We haven't resolved that. Anyone who tells you the data settles it is selling something.

Which is why we push teams toward a narrower question than "is this a good idea." What does this market require of me, priced and dated, before I commit the next $2M?

Before your next sprint planning, put four things on one line: the budget code, the name of the executive whose number improves, the dollar delta, and the month it shows up in someone's report. Three out of four isn't a market. It's a hypothesis with a burn rate.