How AI Is Shaping Build vs. Buy Decisions with Ford Kerr
This is part of our weekly executive insights series where Elion CEO Bobby Guelich speaks with healthcare leaders about their tech priorities and learnings. For more, become a member and sign up for our email here.
Role: Head of AI & Automation
Organization: Soar Autism Center
Can you give a quick overview of your role at Soar?
Soar provides autism therapy for kids aged 2-6 in an in-person, center-based model, offering ABA therapy, speech-language pathology (SLP), and occupational therapy (OT) under a coordinated care model. Basically, the parents drop their kids off, and we handle everything—including authorizations and multi-disciplinary coordination.
The simplicity of that for the parent translates into a lot of operational complexity for us. My role started with a broad mandate to connect Soar to the latest technology to improve operations. That evolved into building our data architecture to centralize information and enable automation, AI-driven decision-making, and workflow optimization. My team is now responsible for:
Data infrastructure, so that we have a single source of truth for all operational and clinical data.
Custom-built automation and tooling, especially for complex scheduling.
Strategic vendor selection and build vs. buy decisions.
It sounds like the coordination part of things is where there’s a lot of complexity; can you explain that a little more?
Scheduling is one of our biggest automation challenges because it’s an exponentially complex problem. Assigning the people to each child throughout the day involves factors like session length and workload balance, consistency and rapport with a set of technicians, specific skills and credentials, language needs, and parental requests. Doing that across a center with 40-50 kids, five days a week, with a roster of full time and part time folks gets complicated quickly!
Our algorithm processes over 30 million scheduling permutations per center per week, and that’s just for ABA therapy—SLP, OT, parental meetings, and breaks add more layers of complexity. The same automation-first mindset applies to lesson planning, documentation, and workflow optimization.
What have you learned about applying AI and automation to these challenges?
The biggest lesson is that building in-house is getting easier and has major advantages. We’re able to tailor the tech to the process, not the other way around, and eliminate unnecessary features and complexity. Building in-house allows us to create faster, more precise workflows for what our team actually does, reducing unnecessary steps.
Formula 1 pit stops are a great analogy. They went from 2 minutes to sub-2 seconds by building tools for the process, not the other way around. That’s the approach we take; custom-built automation makes everything faster, smoother, and more cost-effective than buying generic SaaS.
What are the practical implications of this build-first approach? I imagine there are specific capabilities you need in order to be able to create that kind of customization.
A few things have made this possible. For one, LLMs are removing technical barriers. For example, OpenAI’s o3-mini-high has been able to relatively easily refactor our code to run more efficiently. Dump it in and say “Make it go better” and it works. Organic growth also makes this easier since we can maintain consistency across the organization; for example, if we bought another center with a different cloud platform or different processes, that would be significantly more complex.
Where do you still buy instead of build?
The decision is still cost-benefit driven, but our willingness to pay is lower because building is easier than ever. We’re seeing that in other organizations, too. Klarna, a major payments company, recently moved away from Workday & Salesforce to build in-house AI-powered solutions.
How has your team composition evolved to support this?
Early-stage we really focused on deep engineering expertise. We needed top DevOps and full-stack engineers to build infrastructure. In the future I imagine we’ll shift toward product and automation-focused roles. As AI simplifies execution, product management and process design will become more critical.
Do you think this approach scales to larger organizations like hospital systems?
It’s tougher in large enterprises because of process variation. Every hospital operates differently even within the same health system, making standardization difficult.
There’s also organizational inertia. The status quo bias slows down adoption of new processes. That said, some are catching on. The opportunity is there—it just depends on whether an organization is willing to rethink how they work.