Agentic AI for radiology operations: 6 takeaways from leaders who’ve deployed it

Imaging demand keeps climbing and some leaders expect their workload to double or triple. Reimbursements aren’t keeping pace and hiring your way out of the backlog isn’t an option.

In our recent webinar with The Imaging Wire, two imaging leaders shared how agentic AI is helping their teams keep up. Our CEO joined them to share lessons from deploying agentic AI across healthcare organizations:

• Jessica Bordeaux, CEO of Raleigh Radiology
• Ajay Deshmukh, CEO of KMH Cardiology Centres Inc.
• Rishi Nayyar, Co-Founder and CEO of PocketHealth

If you’re evaluating agentic AI for your organization, here’s what you need to know.

1. Why AI has become the answer

• Staffing gaps are everywhere. Shortages and turnover now span radiologists, technologists and entry-level roles.
• Budgets can’t stretch further. Reimbursements aren’t rising fast enough to fund higher wages for schedulers and techs.
• The gap between home and work is obvious. Leaders use AI in their personal lives, then arrive at work to antiquated systems, fax machines and notepads.

The result: AI projects that once took more than a year to win executive buy-in are now approved in weeks.

“Humans have always been the duct tape between processes… But we’re at somewhat of a breaking point.” Rishi Nayyar, PocketHealth

2. Why your EHR won’t save you

If your EHR already holds the data you need, why not use it to automate? The panel gave three reasons:

• It’s built to store data. EHRs capture information at a point in time. Agentic work is longitudinal, like waiting on a prior authorization before scheduling an exam.
• Your workflows are unique. Processes, including those around referrals, booking, protocoling, vary by payer, region, site and modality. Large EHR vendors rely on uniformity, so deep customization doesn’t fit their model.
• The incentives don’t line up. Ajay noted that incumbent vendors have little reason to build tools that replace modules they already sell.

His advice: Start with what your people need to get done each day, then find technology that supports it.

3. Hire AI like a staff member

Rishi’s advice: Onboard agentic AI the way you would a new hire.

• Give it a workstation (a virtual one works)
• Install your RIS and PACS
• Grant it Active Directory access
• Assign it complete, end-to-end workflows

Why go this far? Making everyone 10% more efficient barely eases staffing pressure, because the same people still carry the same context and responsibility. Automating full workflows end-to-end is what frees up real capacity.

Jessica added that this model makes results easier to see. You know what the agent touched and how consistently it performed, so you can move staff into managing it or improving processes.

4. How to evaluate vendors

Red flags to watch for

• Demo features that won’t be live on the day you sign
• No plan for where redistributed work will land after go-live

What to require

• HIPAA compliance (or your local privacy laws) and clarity on whether your data trains other models
• A clear answer on what happens to your data if the vendor is acquired
• Data sovereignty that fits your jurisdiction
• 24/7 support, because healthcare runs around the clock
• KPIs written into the contract from day one
• A real say in the product roadmap

The sweet spot: Ajay looks for vendors early enough to be free of legacy revenue incentives, yet mature enough to understand the risks.

5. Measuring ROI beyond soft savings

• Track productivity daily. KMH scores more than 120 administrators on the work they complete, weighted by business value and divided by hours worked. After AI deployments, those scores rose 20% to 50%.
• Catch claim issues early. At Raleigh Radiology, AI flags reports where the reason for the exam won’t support the claim, so the team can fix them before submission.
• Look for quick wins. KMH’s automated billing reminders for specialized cardiology codes delivered what Ajay called a “slam dunk” return.

6. Act now or implement under pressure

Patients, referring providers and payers will soon have AI agents of their own. When contacting your center takes one tap, Rishi warned, organizations without an agent on their end will face hold times measured in hours.

“If an AI or agentic solution isn’t going to come from you as an organization, it’s going to come at you from somewhere in the system.” Ajay Deshmukh, KMH Cardiology Centres Inc.

• The technology is ready. Voice models, LLMs and screen automation are production-ready today, which Rishi said wasn’t true a year ago.
• The starting points are clear. Back office tasks including revenue cycle, fax handling and inbound and outbound calls all offer self-evident ROI.
• Waiting raises the stakes. Rishi cautioned that implementing in late 2027 means doing it defensively, under far more pressure.
• People stay central. The panel sees more work ahead than staff can handle, with AI freeing teams to spend more time with patients.

Want to go deeper? Watch the full webinar or book a meeting with our team to explore agentic AI for your healthcare operations.

About PocketHealth

PocketHealth automates the operational complexity of healthcare: the manual, non-clinical work that surrounds every patient visit. For more than a decade, the company has automated this work from both the provider and patient sides, building a distinct understanding of how providers operate and how patients engage in their care. Its agentic operations platform extends that expertise across the enterprise, automating non-clinical workflows end-to-end while working inside existing systems and adapting to each site’s SOPs. The result for providers: improved capacity, financial performance and patient outcomes. PocketHealth is SOC 2 and HIPAA and PHIPA compliant, and trusted by 900+ health systems. Request a demo today.

×