Blogs

Insights from Haris Shuaib

Can you tell us about Newton’s Tree and your role?

I’m Haris Shuaib, founder and chief executive officer (CEO) of Newton’s Tree. Newton’s Tree is the operational layer hospitals use to deploy, govern, and continuously monitor clinical artificial intelligence (AI).

Hospitals are no longer just buying AI tools—they are being held accountable for how those systems behave in production.

How have your priorities changed as AI has evolved, and what are your biggest areas of focus today?

The shift is from evaluating AI to operating it.

A few years ago, the focus was on whether AI worked. Today, the question is whether organizations can run multiple AI systems safely in production. Most health systems can pilot one tool. Very few can manage ten.

At the same time, AI has expanded beyond imaging into areas like clinical documentation, revenue cycle, and decision support. That increases both the opportunity and the operational burden.

Our focus now is helping organizations move from isolated pilots to a repeatable, system-level capability—where each new deployment is faster, safer, and easier than the last.

Where do organizations typically struggle most when moving from AI pilots to real-world deployment?

The biggest challenge is operating leverage. Most health systems have systems that become more painful as they scale AI, not easier. The more AI projects they take on, the more they need to manage, audit, evaluate, and monitor, across more vendors and tools.

We help shift that from negative to positive operating leverage—so the more AI they deploy, the easier it becomes to manage. The goal is to shift that dynamic—so the second, third, and tenth deployment are materially easier than the first. That’s what creates real operating leverage.

How has AI governance evolved with more clinical and agentic AI use cases?

Most governance functions were designed to approve or block projects. That model breaks down with AI.

You can’t manually review every deployment or rely on static approvals when systems are continuously changing. Governance has to move from a gatekeeping function to an operational system.

The best organizations now treat governance as a set of guardrails—combining policy, monitoring, and workflows that allow AI to operate safely in real time.

It’s not just about preventing the wrong things from happening. It’s about enabling the right things to happen at scale.

What capabilities are most important for organizations early in their AI governance journey?

The most impactful capability is a sandbox environment. Hospitals need a safe, secure space to test AI tools before pilot or implementation. This also enables head-to-head evaluation of vendors, which is increasingly important given how many strong solutions and case studies exist. A sandbox makes it possible to see “proof in the pudding” before committing.

What are the biggest barriers in AI procurement, and how can organizations improve efficiency?

The biggest driver of inefficiency is not clearly defining the value case upfront. Teams need to be very clear about the local problem they are solving, rather than starting with a solution. That clarity defines success criteria, informs request for proposals (RFPs), and makes it much easier to identify the right vendor.

Without that, procurement becomes vague, and six months later no one can clearly explain the return on investment (ROI). Strong procurement starts with a clearly defined, locally grounded value case.

How does Newton’s Tree support organizations through the AI lifecycle?

We structure the AI lifecycle into: select, test, deploy, and monitor. We start with “select” – helping identify the right use cases and vendors. Our in-house clinical AI team works closely with health systems to understand local context and shape value cases.

We then support RFP development and market engagement. In many cases, we act as a white-glove partner through the entire process. Some large systems don’t need this support, but many rural or under-resourced systems benefit significantly from having AI expertise embedded.

Once systems are live, we provide continuous monitoring and governance. The key is that this is not a one-off process. It’s a repeatable system that can be reused across multiple AI deployments.

How are health systems measuring ROI from AI?

From our perspective, ROI becomes clearest at the integration layer. We act as the “highway” between third-party AI applications and core health IT systems like picture archiving and communication systems (PACS) and electronic health records (EHRs). Without that layer, every AI integration can take months of IT work and significant internal resources.

With Newton’s Tree in place, integrations are already standardized. What might normally take six months and $50,000 to $100,000 in internal effort can be reduced to a few weeks. That speed is where ROI really compounds, because it increases organizational appetite for AI adoption.

How do you support adoption and change management?

Technology is only one part of the solution. We think about people, platforms, and policies. That includes training, education, and defining the right processes and governance structures.

The goal is to move organizations from visibility to reliance—where AI is not just available, but actively used and trusted in decision-making.

How do you address the gap between perceived and actual risk in AI systems?

This comes up frequently. When we help hospitals implement a new application, we support full risk assessment as part of the select and test process. That includes IT security, regulatory considerations, and clinical governance. A key issue is that risk is often treated as a single concept, when in reality it exists across multiple layers.

There are risks related to input data, model behavior, and how users interact with the system. For example, something can be technically secure but still clinically unsafe, or clinically valid but used incorrectly in practice.

We help organizations assess risk at the level of the specific use case and then monitor how that risk evolves over time.

How did Newton’s Tree get involved with CHAI?

We have strong relationships across CHAI who we’ve worked with or known through the ecosystem.

I also co-founded an open-source community (MONAI Deploy) with colleagues at NVIDIA and several U.S. health systems focused on AI deployment infrastructure in hospitals. That experience showed me the value of building coalitions and shared standards.

CHAI represents a similar opportunity—bringing together leading organizations to define best practices for safe and effective AI adoption.

What role does transparency play in AI adoption and governance?

Transparency is essential in a space this complex and fast-moving. The biggest tool we have for safe AI adoption is visibility into how systems work. Model cards and public registries help standardize that information so stakeholders can make informed decisions.

Importantly, this isn’t about calling people out, it’s about ensuring the right people have the right information in a usable format. You don’t want to overwhelm clinical or operational teams with technical detail, but you do want enough transparency to distinguish between solutions and understand tradeoffs.

What is your long-term vision for Newton’s Tree?

We see Newton’s Tree as analogous to systems like the EHR or PACS. When healthcare became digital, there was a need for centralized systems of record and coordination. The same will be true for AI. Instead of managing dozens of disconnected AI tools and vendors, health systems will need a central orchestration layer. That’s what Newton’s Tree is building.

Any final thoughts for health systems working on AI governance today?

The best health systems are proactive. Waiting for regulation alone is not a strategy. The organizations that succeed are those that engage early, collaborate, and actively shape how AI is deployed.

You improve outcomes by engaging with peers and communities like CHAI, and by staying focused on maximizing benefit while minimizing risk.