Before Behavioral Health Adds AI, Fix the Workflow

By Jelard Macalino, Chief Information Officer, Central City Health

AI is getting attention in healthcare. The promise is clear: faster documentation, better communication, smarter routing, better reporting, and less administrative burden on teams.

But in behavioral health, we need to be careful about where we start.

At Central City Health, a Detroit-based federally qualified health center that provides integrated medical, behavioral health, dental, and supportive services, my work spans technology strategy, enterprise systems, cybersecurity, data integration, and the technology that supports care delivery. The perspective here is my own, based on the work I do at Central City Health.

My software development background also shapes that perspective. Workflow, data, integration, and system design are always part of the conversation. Before I ask what AI can do, I usually ask whether the workflow and data beneath it are ready.

In this role, I have learned that a technology decision is rarely only about the technology. It affects data, workflows, staff, and ultimately the experience of care.

AI cannot fix a workflow that does not work. It cannot create a complete view of a patient if the systems underneath are disconnected. It cannot make care coordination easier if information is scattered across platforms that do not communicate with one another.

Ideally, an organization would have one EHR that supports its full care model. In reality, that is not always how healthcare technology grows.

Across the sector, systems tend to accumulate one decision at a time. An organization adopts a tool that meets the immediate need. Later, a new service line, reporting requirement, or regulation calls for something the original system does not handle well. Another tool is added. Over time, reasonable choices can create an environment that is difficult to unwind.

That matters in behavioral health because care is often connected to many parts of a person’s life. When the technology environment is fragmented, staff and patients can end up carrying the complexity. The question is not only whether a system works on its own. It is whether the systems together support the way care actually happens.

AI cannot repair fragmented workflows. It works best when the systems beneath it are ready.”

That is why I do not think behavioral health organizations need to rush into adding one more application to an already crowded environment.

Many teams are already dealing with app sprawl. Beyond core EHR systems, there may be separate tools for scheduling, communication, reporting, billing, compliance, referrals, analytics, and operations. Each new tool brings cost, implementation work, training, contract negotiation, and security review.

This is often missed in the AI conversation.

Much of the focus is on AI assistants, chatbots, and agents inside applications. Those tools can be useful. But they also introduce token or usage-based costs, governance requirements, monitoring needs, and questions about security and reliability. For organizations with limited budget and staff, those responsibilities matter.

A practical use of AI may be behind the scenes: helping organizations build, configure, or improve systems faster. For some, that may involve integrations, automations, reports, dashboards, and workflow tools. For others, it may mean working with consultants, low-code platforms, reporting tools, or vendor configuration teams.

In that model, AI is not always the thing on the screen. Sometimes it is the capability that helps an organization deliver the workflow it actually needs.

Building internal capability does not always mean hiring a full software development team or launching a large custom-platform project. It can start smaller. Choose one workflow that creates repeated friction, such as appointment visibility, referral tracking, duplicate data entry, or reporting. Then use the capability available, whether that is internal IT, vendor configuration, a consultant, low-code, reporting tools, or AI-assisted development.

For nonprofit healthcare organizations, this is especially important. Budgets are real. Capacity is limited. Buying one more product may solve a narrow problem while creating new complexity somewhere else.

That is why I see AI as part of an integration strategy, not as a replacement for EHR systems. The goal should be to build around the systems an organization already has: an integration hub, patient communication tools, reports, dashboards, or better appointment visibility across services.

Appointment visibility is a simple example. When a patient receives care across multiple services, staff may need to piece together appointments, follow-up needs, and communication from more than one system. A better workflow can move teams from scattered information toward clearer coordination and more reliable follow-through.

But this work has to be done responsibly. Behavioral health is highly regulated, and security and compliance cannot be an afterthought. Any AI-enabled workflow must protect sensitive information, respect privacy, and operate within clear governance. AI should support human judgment, not quietly replace it.

Innovation in healthcare does not always mean chasing a new tool. Sometimes it means connecting what already exists in a better way.

AI will not fix fragmented data, unclear workflows, or weak governance by itself. But when it is paired with integration work, sound security, and practical leadership, it can help behavioral health organizations reduce friction and build systems that better support the people doing the work.

The future of AI in behavioral health should not begin with the question, “What AI tool should we add next?”

It should begin with a better question:

“What workflow needs to work better for care to work better?”