At HealthIMPACT Fall Forum the focus is on making healthcare AI earn its place

By Rohit Mahajan
Co-Host, The Big Unlock Podcast

I was at the HealthIMPACT Fall Forum at Microsoft Times Square in New York City, where the panel discussions were exciting, interactive, and grounded in the realities of healthcare transformation.

One session in particular, “The Architects’ Scorecard: What Enterprise AI Actually Returned,” brings an important set of questions into focus. Which AI investments have earned their place across clinical, operational, and revenue workflows? Who owns the results? And when should an organization decide that a deployment has run its course?

These are the questions healthcare leaders need to be asking as AI becomes part of everyday operations.

As co-host of The Big Unlock podcast, I spend a great deal of time exploring how organizations translate technology into better care and stronger operational performance. The discussion here reinforces a principle I keep coming back to: stay grounded in the mission, understand the problem, and focus on the outcomes required.

AI investments need to earn their place

Healthcare organizations have many opportunities to apply AI. They also have limited budgets, stretched teams, and competing priorities.

A portfolio approach helps bring discipline to those choices. Every deployment should have a clear purpose, an accountable owner, and evidence that it is delivering value.

For a clinical workflow, that means examining quality, safety, and the experience of the people using it. For an operational workflow, it might mean faster access, fewer missed steps, or more capacity for staff to support patients. For a revenue workflow, it means understanding whether the improvement justifies the full cost of implementation and ongoing operation.

The question I would ask is straightforward: what is different for the patient, the clinician, or the organization because this capability is now in place?

That answer should guide the investment decision.

Ownership has to extend beyond the purchase

Another important part of the discussion is how organizations divide responsibility for AI results, from the person who approves the investment to the person who answers when an agent gets something wrong.

This becomes especially relevant as AI moves from generating information to taking actions within a workflow.

Technology teams, clinical leaders, operational teams, finance, and security each have responsibilities. Those responsibilities need to connect. Someone must own the intended outcome, someone must monitor performance, and staff must know how to escalate a problem.

For me, governance becomes practical when people can answer a few basic questions. What is the system allowed to do? Where does human judgment enter? Who can intervene? Who is responsible for correcting a failure?

Clear answers make it easier to adopt AI with confidence.

Retiring a deployment is part of responsible leadership

The willingness to stop an AI deployment deserves as much attention as the decision to launch one.

An organization may have invested considerable time, money, and reputation in a solution. That can make it difficult to acknowledge that the value has not materialized.

Yet continuing a deployment also has a cost. It consumes budget, requires support, and takes attention away from other opportunities. If staff are working around the system, or if the original problem remains unresolved, leaders need to examine why.

I believe organizations should establish review criteria early. They should know what success looks like, what signals require intervention, and what would justify retiring the capability.

Making those decisions based on evidence helps keep the mission at the center.

Closing the gap between patient needs and provider capacity

The broader HealthIMPACT focus on scaling care beyond hospital walls gives these questions added urgency.

Patients need timely access, clear communication, and continuity across their care journey. Providers need the capacity and support to meet those needs.

AI can help bridge that gap when it is designed around the work. That requires understanding the patient’s experience, the care team’s responsibilities, the information they need, and the handoffs where delays or missed actions occur.

This is where I see the opportunity for humans and AI to work together: giving people better information, supporting coordination, and creating more capacity for the decisions and relationships that require human expertise.

The discussions here are a useful reminder that successful transformation requires thoughtful choices about what to scale, what to improve, and what to stop.

I look forward to bringing these questions into future conversations on The Big Unlock podcast, and hearing how healthcare leaders are turning AI investments into measurable improvements for the people they serve.