Here's a pattern I keep seeing across healthcare organizations.

A department proudly walks through their AI tooling stack. Four different vendors. Overlapping alerts. Conflicting recommendations on the same patient. Three of the tools scraping the same data from the same EHR, just presenting it slightly differently.

The clinicians stay polite. You can see it on their faces anyway.

They aren't ignoring one tool. They're ignoring all four.

This is the part of the healthcare AI story that nobody likes to tell. We talk about the shortage of AI in medicine. The real crisis in many organizations is the opposite. Too many tools. Too many vendors. Too many dashboards fighting for the same 30 seconds of a clinician's attention.

Healthcare doesn't need more AI tools. It needs fewer, better ones.

𝗧𝗛𝗘 𝗙𝗟𝗢𝗢𝗗 𝗜𝗦 𝗥𝗘𝗔𝗟

Every week there's a new healthcare AI startup. New pitch deck. New "revolutionary" platform. Every major conference has a vendor hall that looks the same as last year's, just with more booths.

If you work in a health system, you know the pattern. The sales emails don't stop. Every vendor promises the same three things: predict risk earlier, reduce clinician burden, improve outcomes.

And most of them can technically do a piece of what they promise. The demos work. The pilot numbers look good. The problem shows up later, when the tool has to live inside a real workflow next to the other 14 tools already there.

Nobody in the room ever asks the question that matters most: do we actually need another one?

𝗧𝗛𝗘 𝗥𝗘𝗔𝗟 𝗣𝗥𝗢𝗕𝗟𝗘𝗠 𝗜𝗦𝗡'𝗧 𝗖𝗔𝗣𝗔𝗕𝗜𝗟𝗜𝗧𝗬

I keep hearing leadership frame this as a technology problem. "We need better AI." "We need smarter models." "We need to catch up."

The problem is almost never capability. The problem is integration.

Most health systems already have:

• An EHR with some predictive features built in

• A risk stratification tool that came with their population health platform

• A clinical decision support module somebody procured three years ago

• A scheduling optimization tool that was supposed to fix no-shows

• A documentation assistant that one department piloted and never rolled out

None of these talk to each other. Each one has its own login, its own interface, its own alerts, its own definition of "high risk." A single patient might be flagged by three different systems using three different logic chains, with recommendations that contradict each other.

Adding a 15th tool to this environment doesn't solve anything. It makes the fragmentation worse.

𝗪𝗛𝗔𝗧 𝗦𝗨𝗖𝗖𝗘𝗦𝗦𝗙𝗨𝗟 𝗢𝗥𝗚𝗔𝗡𝗜𝗭𝗔𝗧𝗜𝗢𝗡𝗦 𝗗𝗢 𝗗𝗜𝗙𝗙𝗘𝗥𝗘𝗡𝗧𝗟𝗬

I've been paying close attention to which health systems actually get value out of their AI investments. They don't have more tools than everyone else. They have fewer.

The pattern I see in the ones that work:

• They pick 2 or 3 tools that fit specific, well-defined workflows. Not platforms that promise to solve everything. Tools that do one thing inside one workflow and do it cleanly

• They integrate those tools deeply into the existing clinical systems. The AI output shows up in the EHR, not in a separate portal. Clinicians don't log in to anything new

• They kill or consolidate the rest. Ruthlessly. If a tool has low adoption after 6 months, it doesn't get a second year. The opportunity cost of keeping a low-value tool alive is higher than most leaders realize

• They measure actual usage, not licensed seats. A vendor report saying 500 people have access means nothing. How many logged in yesterday? How many took action based on what the tool told them?

The organizations drowning in AI tools usually got there by saying yes too many times. The organizations getting value from AI got there by saying no.

𝗧𝗛𝗘 𝗩𝗘𝗡𝗗𝗢𝗥 𝗘𝗖𝗢𝗦𝗬𝗦𝗧𝗘𝗠 𝗜𝗦 𝗡𝗢𝗧 𝗬𝗢𝗨𝗥 𝗙𝗥𝗜𝗘𝗡𝗗 𝗛𝗘𝗥𝗘

Vendors are incentivized to expand their footprint. More features. More modules. More dashboards. More seats. That's how their business works.

Healthcare, unfortunately, needs the opposite. You don't need a vendor with 200 features. You need a tool that does one thing perfectly inside the workflow you actually have.

This ties directly back to Issue #3, the 5 vendor questions. If you go through those five questions seriously before any purchase, most of the tools in your current stack would not have made the cut. The ones that survive are the ones designed to fit a real workflow, not the ones designed to check the most boxes on a procurement spreadsheet.

A tool that does 1 thing exceptionally well beats a platform that does 10 things adequately. Every time.

𝗧𝗛𝗘 𝗙𝗜𝗫 𝗜'𝗩𝗘 𝗦𝗘𝗘𝗡 𝗪𝗢𝗥𝗞

The departments I've watched escape this pattern didn't add a fifth tool. They did an audit.

They listed every AI-adjacent tool in use. They interviewed clinicians about which ones they actually used versus which ones were technically deployed. They mapped the overlap.

The result: kill three of the four. Take the one that genuinely fits the workflow and invest in deep EHR integration for it. Redirect the license budget from the killed tools toward the remaining one.

Within a couple of months, clinicians in those departments go from ignoring everything to using one tool daily.

The budget savings are real. The adoption numbers are real. The part people don't expect is how much cognitive load comes off the clinicians. They stop getting contradictory recommendations. They stop having to mentally triage which tool to trust on which decision. They get their attention back.

𝗧𝗛𝗘 𝗨𝗡𝗖𝗢𝗠𝗙𝗢𝗥𝗧𝗔𝗕𝗟𝗘 𝗧𝗥𝗨𝗧𝗛

The reason most organizations don't do this is political, not technical.

Somebody championed each of those 14 tools. Somebody signed the contract. Somebody presented the pilot results at a leadership meeting. Killing a tool means admitting that a past decision didn't work, and healthcare organizations are not great at that.

But the cost of keeping low-value tools alive is compounding. Every month you don't consolidate, clinicians get more burned out on AI. By the time you actually want to deploy something useful, nobody trusts it. The boy who cried AI.

Fewer, better tools isn't a cost-saving play. It's a credibility play. The tools you keep have to earn their place in the workflow every quarter. The ones that don't, go.

- Guryash

P.S. If your organization has been through a tool consolidation exercise, I'd love to hear what worked and what didn't. What was the hardest tool to kill? What surprised you about the aftermath?

Want more? Follow me on LinkedIn where I share daily insights on healthcare AI implementation: linkedin.com/in/guryashsingh

Use this 10-question audit on every AI tool currently running in your organization. Answer each question honestly for each tool. The results tell you what to keep, what to consolidate, and what to kill.

𝟭. 𝗨𝗦𝗔𝗚𝗘: What is the actual daily active usage among the intended users?

• High (60%+ of intended users engaging daily) → Keep investigating
• Moderate (20-60%) → At risk, diagnose why
• Low (under 20%) → Strong candidate to kill or consolidate

𝟮. 𝗜𝗡𝗧𝗘𝗚𝗥𝗔𝗧𝗜𝗢𝗡: Does the tool live inside the EHR or require a separate login/portal?

• Fully embedded in EHR or existing clinical system → Keep
• Partial integration (SSO but separate interface) → Investigate whether deeper integration is possible
• Standalone portal → Very high risk of sustained non-adoption

𝟯. 𝗪𝗢𝗥𝗞𝗙𝗟𝗢𝗪 𝗙𝗜𝗧: Does using this tool add steps to an existing workflow or replace steps?

• Replaces steps (saves time) → Keep
• Neutral → Needs strong clinical outcome evidence to justify
• Adds steps → Kill unless the outcome value is extremely high

𝟰. 𝗢𝗩𝗘𝗥𝗟𝗔𝗣: Do any other tools in your stack perform the same or similar function?

• No overlap → Keep evaluating
• Partial overlap → Consolidation candidate
• Full overlap → Kill one of them, immediately

𝟱. 𝗖𝗟𝗜𝗡𝗜𝗖𝗔𝗟 𝗧𝗥𝗨𝗦𝗧: If you asked 10 frontline clinicians whether they trust the output of this tool, how many would say yes?

• 7+ → Keep
• 4-6 → At risk, diagnose trust gap
• 3 or fewer → The tool is functionally dead, even if technically deployed

𝟲. 𝗢𝗨𝗧𝗖𝗢𝗠𝗘 𝗘𝗩𝗜𝗗𝗘𝗡𝗖𝗘: Can you point to a measurable clinical or operational outcome this tool has changed in the last 12 months?

• Yes, measurable and documented → Keep
• Yes but anecdotal → Fix measurement before the next renewal
• No → Strong kill candidate at next contract cycle

𝟳. 𝗖𝗢𝗦𝗧: What is the total annual cost including license, integration, training, and internal support time?

• Compare this number against your question 6 answer. If the cost is high and the outcome evidence is weak, the math is telling you something

𝟴. 𝗢𝗪𝗡𝗘𝗥: Who inside the organization is accountable for this tool's adoption and results?

• Named owner with time allocated → Keep
• Named owner but no bandwidth → At risk
• No clear owner → Very high kill candidate. Orphaned tools almost never recover

𝟵. 𝗠𝗔𝗜𝗡𝗧𝗘𝗡𝗔𝗡𝗖𝗘: What is the ongoing engineering and clinical effort required to keep this tool running?

• Low and stable → Keep
• High but value justifies it → Keep, monitor
• High and value unclear → Consolidate or kill

𝟭𝟬. 𝗙𝗨𝗧𝗨𝗥𝗘 𝗙𝗜𝗧: In the next 24 months, does your IT roadmap naturally sunset or consolidate this tool?

• Yes, it will be replaced by an existing platform → Kill now rather than later to avoid wasted investment
• No, it's a long-term piece of the stack → Keep, continue monitoring
• Unclear → Flag for roadmap review

𝗦𝗖𝗢𝗥𝗜𝗡𝗚 𝗔𝗡𝗗 𝗗𝗘𝗖𝗜𝗦𝗜𝗢𝗡 𝗙𝗥𝗔𝗠𝗘𝗪𝗢𝗥𝗞

For each tool, count how many of the 10 questions point toward "keep" versus "kill/consolidate."

• 8-10 keep signals: Invest deeper. This tool is working. Consider expanding its footprint
• 5-7 keep signals: Conditional keep. Identify the weak dimensions and fix them in the next quarter
• 3-4 keep signals: Consolidation candidate. Look for a tool in your stack that can absorb this one's function
• 0-2 keep signals: Kill. Start the contract unwinding process and redirect the budget

𝗛𝗢𝗪 𝗧𝗢 𝗥𝗨𝗡 𝗧𝗛𝗘 𝗔𝗨𝗗𝗜𝗧

• Step 1: List every AI or AI-adjacent tool in your organization. Don't forget the ones buried inside platform contracts
• Step 2: Assign each tool to an owner who'll answer the 10 questions honestly
• Step 3: Pull real usage data, not vendor-reported numbers
• Step 4: Interview 5 frontline clinicians per tool. Their answers matter more than the dashboards
• Step 5: Make the keep/consolidate/kill decisions as a leadership team, not as individual champions of individual tools

This audit is exclusive to the email edition of HealthTech Singh. Use it to clean up your AI stack before you add anything new.