Everyone Is Selling You a Solution. Almost No One Is Selling You a Decision.

There is a number that should stop every health system leader who is about to approve another technology investment.

Roughly 80% of healthcare AI projects fail to scale beyond the pilot phase. Seventy percent never reach production at all. The median time from pilot approval to quiet shutdown is fourteen months. (healthtechdigital.com)

The enterprise picture is no better. MIT's research, widely reported, found that 95% of enterprise AI pilots fail to deliver measurable ROI — and that the average sunk cost of an abandoned AI initiative reached $7.2M. (healthcareitnews.com)

Read those numbers slowly, because the instinct is to read them as a technology problem. They are not. The technology in most of these projects worked. The models performed. The dashboards rendered. The pilots produced exactly what they were designed to produce.

What failed was everything around the technology — the decision about whether it was the right investment, the comparison against the alternatives competing for the same capital, the assumptions about how value would actually be captured, the governance that should have caught the problem before fourteen months and $7.2M had evaporated.

The market is saturated with solutions. What it is starving for is a way to decide between them. And until that changes, the failure rates will not.

The market is not empty — it is misaligned

If you are a health system leader trying to make better strategic and capital decisions, the market offers you no shortage of things to buy. The problem is that none of them are built to solve the decision problem. They are built to solve adjacent problems — and then positioned as if solving the adjacent problem solves the decision problem too.

There are five categories of what is actually available. Each is good at what it was built for. Each leaves the same gap.

1. Analytics and BI platforms. Tableau, Power BI, and their healthcare-specific equivalents are excellent at making data visible. They turn raw information into charts, trends, and dashboards that a human can read. But a dashboard tells you what happened. It does not tell you what to do about it, which competing initiative deserves the capital, or what you are giving up if you fund one priority over another. Most health systems already own these tools. They are not the bottleneck. The bottleneck sits one layer above them.
   

2. Data platforms. Health Catalyst, Innovaccer, Arcadia, and others are strong at aggregating data, building pipelines, and integrating across EHR and claims sources. Several health systems have committed multi-million-dollar contracts to them. But these platforms manage data — they do not produce decisions. There is no prioritization layer, no CFO-grade ROI defensibility, no mechanism for comparing one strategic initiative against another. The output is a data product, not a capital allocation decision. And a familiar pattern follows: high investment, low realized value, because the platform delivered exactly what it promised — clean, integrated data — and the organization still could not decide what to do with it.
   

3. Consulting firms. Deloitte, McKinsey, Chartis, Optum, and the rest bring genuine expertise, executive trust, and political cover. When a health system needs help with strategic prioritization or capital allocation framework design, a consulting engagement is the default move. But the consulting model has a structural flaw for this specific problem: it is episodic. The framework lives in the engagement. When the consultants leave, the framework leaves with them. The next decision cycle starts from zero — a new engagement, a new team, often a new methodology, and another $500K to $3M. Nothing compounds. The same prioritization problem gets solved three times across three engagements, and the organization pays full price each time.
   

4. Execution and workflow tools. Project management platforms, governance trackers, and PMO tooling are good at managing work once it has been approved. They do nothing about the quality of the decision that approved it. An initiative that should never have been funded does not become a better investment because it is tracked well.
   

5. The status quo. This is the most common alternative of all, and the most underrated. Excel models, PowerPoint decks, governance councils, weekly intake meetings, spreadsheet-based scoring, executive committee politics. It is embedded, familiar, and carries no procurement cost. It is also not scalable, not defensible to CFO scrutiny — the scoring is whatever the spreadsheet says — and it disappears the moment the analyst who built it leaves. The status quo persists not because it works, but because it distributes accountability so widely that no single person owns the dysfunction.

Five categories. Each solves a real problem. None of them solves the decision problem — because none of them was built to. The gap between analytics and execution, between data and a defensible capital decision, is structurally unoccupied.

That gap is where the 80% failure rate is born.

The tools are in the shed. The decision layer isn't.

Matt Frankel, Adaptive's General Manager for Health Systems — who has worked across health systems, payers, and the consulting side — frames the market the way a practitioner sees it: health systems already have tools in their tool shed. The problem is that each one was built for a different purpose, and none of them connects.

"Health systems have data platforms built to store and surface information," he puts it. "Consultants are engaged to get to a recommendation, not to give the organization infrastructure on a go-forward basis. And we've got slide decks — we've all lived in that world. We don't need more slide decks. Slide decks are something to build, to present and persuade, but not to make decisions."

What ties those tools together today is usually a person — a smart strategy officer working hard to pull the siloed pieces into something coherent. That works until it doesn't. "It's not scalable," Matt notes. "It suffers when you've got mass turnover in the healthcare industry and a strategy officer leaves and someone new comes in." The institutional knowledge walks out the door. The next cycle relearns lessons the organization already paid for.

Chris Donovan, Adaptive's CEO, describes the same gap from the strategy-formation side.

When there is no data-driven structure underneath the process, "the loudest voice in the room, or an opinion-based strategy, gets leveraged."

His example is concrete: the head of a cardiology department comes in with a costly strategic investment they're confident will pay off over time. How do you compare that against the dozen other opportunities competing for the same capital — in a data-driven, non-opinion-driven way? Most health systems have no honest answer. So the comparison never happens, and the decision gets made on conviction and political weight rather than shared logic.

The old model, in Matt's words, comes down to one question: how did that decision actually get made? "Nine times out of ten, it's either an organizational political decision, or there was a champion." The new model has to produce a different answer — a true comparison of ROI across options, built on a consistent level of data. Less political, less gut-feel, more data-driven. That is the whole shift in a sentence.

Why the failure rate is a decision-layer problem

It is worth being precise about the mechanism, because the conventional reading gets it wrong.

When an AI project fails to scale, the post-mortem usually blames implementation — change management, data quality, integration complexity, adoption resistance. Those are real. But they are downstream symptoms. The upstream cause, in case after case, is a decision that was never structured well enough to succeed.

Was this the right initiative to fund, compared to the alternatives? The organization rarely knows, because the alternatives were never compared on common terms.

What did value capture actually require? The operational pre-conditions — workflow changes, staffing models, governance decisions — were assumed rather than examined. The financial projection treated them as solved when they were not.

What were the leading indicators that would tell us early if this was working? They were never defined, so the project ran for fourteen months before anyone could say with confidence that it had failed.

Deloitte's research on measuring digital transformation ROI in healthcare names this gap directly: organizations do not lack data. They lack a consistent, structured framework for evaluating and comparing investments. (deloitte.com) The investments fail not because the data was missing, but because the decision logic above the data was missing.

This is the through-line connecting the failure statistics to everything we have been writing about. A business case is not a decision. The room cannot compare what was never built to be comparable. The trade-offs that were never made explicit got made in the hallway. And the result, aggregated across an entire industry, is an 80% scaling failure rate and a $7.2M average write-off per abandoned initiative.

The market has responded to this by selling more solutions. More AI. More dashboards. More platforms. More point tools. But the problem was never a shortage of solutions. It was the absence of the layer that decides between them.

What we built instead

Adaptive Product's Strategy Intelligence Platform is not another entry in any of those five categories. It is the layer that sits above them.

It does not replace your analytics tools — it uses their outputs. It does not displace your data platform — it sits above it and finally turns that data investment into capital allocation discipline. It does not compete with your consulting relationships — it makes the decision logic those engagements produce reusable, so you stop paying full price to solve the same problem every cycle.

What SIP does is occupy the empty layer: it makes strategic and capital decisions comparable, defensible, and reusable across cycles.

Concretely, that means three things the other categories structurally cannot provide.

It makes initiatives comparable. Every priority that enters the decision process is evaluated on the same logic — the same ROI framework, the same assumption structure, the same definition of value. The room is no longer comparing narratives. It is comparing priorities on shared terms. The better-packaged initiative no longer beats the better initiative simply because it presented well.

It makes the logic defensible. SIP's scoring is deterministic, not black-box AI. The math is auditable. The assumptions are explicit and challengeable. When the CFO asks what is driving a number, the answer is in the system — traceable back to its inputs. A recommendation that comes out of SIP is built to survive board scrutiny, not just to win the room.

It makes the decision durable. The logic does not leave when the meeting ends or the consultant's engagement closes. It is preserved, governed, and reusable. When a decision needs to be revisited six months later, the organization returns to the original logic rather than rebuilding it from scratch. Decision quality compounds across cycles instead of resetting to zero each time.

That is the difference between buying a solution and building a decision capability. A solution is a thing you implement and hope scales. A decision capability is the structure that determines whether anything you implement was the right thing in the first place — and whether you will be able to tell, early, if it is not.

In a market where 80% of projects fail after implementation, the most valuable thing a health system can own is not another solution. It is the layer that decides which solutions are worth implementing at all.

A word on AI — because everyone is asking

In a market flooded with AI tools, the obvious question is whether AI alone closes this gap. It does not — and the reason is precise.

Large language models are genuinely transformative at deriving data points from unstructured information. But, as Chris puts it, they are not reliable at driving to accurate conclusions consistently. "A lot of data points," he says, "but any different seed or different assumption will lead to different conclusions." That inconsistency is disqualifying for capital allocation, where the same inputs have to produce the same defensible answer every time.

What makes the technology usable for decisions is an expert system layered on top of it — a governance framework, encoded institutional knowledge, and deterministic logic that keeps the reasoning consistent, persistent, and repeatable. That combination is the engine: the analytical power of modern AI, disciplined by an expert decision framework, so the output is something a CFO can actually defend. AI is the enabler. It is not the product. The product is the consistency.

This is also why the consulting-spend math changes. When the decision layer is a persistent utility rather than a series of episodic engagements, a health system can reallocate strategic spend away from repeatedly planning to derive value and toward actually deriving it. As Chris frames it, that is the critical shift — moving resources from planning on deriving value to capturing it.

The honest version of the argument

None of this means the analytics platforms, data platforms, or consulting firms are wrong to exist. They solve real problems, and most health systems will continue to need all three. SIP is not an argument against any of them. It is an argument that there is a layer none of them occupy — and that the absence of that layer is what the failure statistics are actually measuring.

The question for a health system leader is not "which solution should I buy next?" The market has plenty of answers to that question, and an 80% failure rate to show for them.

The better question is: "before I buy anything else, do I have a defensible way to decide whether it is the right investment — and a way to know, early, if it is not?"

If the answer is no, that is the gap to close first. Every dollar spent before closing it is a dollar exposed to the same failure rate everyone else is living.

We sat down with Matt and Chris to walk through all of this — the market gap, why solutions keep failing after implementation, and what the decision layer looks like in practice. Watch the conversation here →

https://www.youtube.com/watch?v=2V36KjKZpK0

And if you want to see what defensible decision logic looks like in real output form, download the Sample Decision Artifacts → Click here.

Sources

Health Tech Digital — The AI Implementation Gap: Why 80% of Healthcare AI Projects Fail to Scalehttps://www.healthtechdigital.com/the-ai-implementation-gap-why-80-of-healthcare-ai-projects-fail-to-scale/

Healthcare IT News — MIT: 95% of enterprise AI pilots fail to deliver measurable ROIhttps://www.healthcareitnews.com/news/mit-report-95-genai-pilots-companies-failing

Deloitte — Measuring the ROI of Digital Transformation in Health Carehttps://www.deloitte.com/us/en/Industries/life-sciences-health-care/articles/measuring-digital-transformation-roi.html