What 15 Million Patient Interactions a Year Reveal About AI in the Front Office

Jared and Frederik Mueller go back around 15 years, to playing intramural football together at college. Today Frederik is co-founder and CEO of Third Way Health, which runs the front office for US medical practices with a combination of dedicated team members and a technology platform: calls, outbound patient engagement, referrals, prior authorisations, document management, care gaps and patient self-service.

He started in healthcare in London at the Advisory Board Company, then worked across the UK, Europe, the Middle East and Canada. His view is that front offices are far more alike than they think.

The biggest differences between the UK and the US, he says, are payment and data flow: the US spends far more time dealing with insurers and value-based care arrangements, and its technology is more varied but also more fragmented, with dozens of EHRs that struggle to share data.

Third Way now handles around 15 million patient interactions a year across roughly 2,000 providers in 16 specialties. The biggest lesson from that volume, he says, is almost philosophical: a constant tension between providers, who want to treat each patient in the way they think best, and administrators, who need some standardisation to run at scale.

Self-booking is his example. The technology has existed for more than a decade, but the rules and 'shadow rules' around a provider's diary hold it back, and in his experience the best-run groups resolve that tension through trust.

On AI, his starting point is that if you automate a broken process, you haven't gained anything. He estimates each provider generates around half a change a week, from sickness cover to new preferences, so technology goes stale unless someone governs the process behind it.

He cautions against judging AI by how many calls it resolves, because a resolved call can still be a poor experience, and against assuming that automating a share of calls removes the same share of staff.

He argues that patients should choose how they interact, that AI done well can end the single queue where simple and urgent calls wait together, and that not all automation is equal: ambient scribes and eligibility checks are a different problem from scheduling.

He expects voice AI to be commoditised and says what it really needs is context and workflows. With front-office turnover he puts at 40 to 50 percent a year, he believes technology alone rarely fixes a practice's staffing problem.

He and Jared also unpick the claim that 'our PMS can do that', the difference between an EHR, a practice management system and a CRM, and why, like most people with Excel, practices use a small fraction of what their systems can do.

His test for any technology is whether you can really operate it, what happens when you rely on it and it breaks, and whether to do it yourself or work with a partner who focuses on that one problem every day.

He closes with what the UK and the US could learn from each other: the UK's simpler single-payer administration, and the US appetite for care innovation.