Why "What's Eating Up Your Time?" beats "Where Would You Use AI?"

Ask a therapist, psychologist, or psychiatrist "which processes or workflows would you like to use AI for?" and you'll usually get a pause. Not a concrete answer. At Delphyr, we've learned the hard way: terms like AI, workflow automation, and agentic support talk easily about technology and awkwardly about the actual work. So we changed the question. Instead of asking where AI fits, we ask: "which task takes up the most time for you or your colleagues right now?"

That's when the conversation opens up. People know the answer immediately. A patien file they have to dig through before a handoff. A note they're still finishing after hours. A letter they've basically written three times already that day (just in slightly different form). In other words: it's better to start with a discussion about the work that's actually taking too much time. Not the technology.

Vision matters. But it needs to land somewhere concrete.

That's not to say the bigger picture doesn't matter. It does, and at the board level, questions about how and why workflows should be supported, are usually front and center. Will staffing hold up under the current workload? Can we handle natural attrition? What will AI mean for quality of care? Is it safe to implement AI in our EHR?

All real questions. But a vision only becomes convincing once it lands on a task your team actually recognizes. So let's get concrete. Here's what AI support can look like in practice for mental healthcare teams using Delphyr day to day.

Three tasks that quietly eat your team's time

Preparing for a treatment coordination meeting. 

We’ve learned from our clients that this can easily take an hour: searching through records, gathering information, putting together a summary. With AI search and summarize functionalities directly embedded in the EHR, our clients expect that this preparation can drop to ten minutes. So: pulling together information that is already sitting in the record, no longer eats up human time.

Preparing for a multidisciplinary team meeting

Same pattern, different meeting. Most of the time doesn't go to the discussion itself, it goes to gathering what's relevant from a record, often pulling from multiple clinicians' notes.

Drafting a discharge letter 

A recurring document with a largely fixed structure, still written from scratch more often than not. AI can help draft a first version based on what's already documented. This saves real time, and the clinician stays in control, reviewing and refining before anything goes out.

Notice what these three have in common: none of them ask for new clinical insight. They're all time spent searching for, summarizing, and re-typing information that already exists.

Start small. Keep the bigger picture in view

When you start with asking yourself which task is taking too much time right now, and what's a realistic first step to relieve it, the bigger vision will follow. For us this means an integrated AI layer that eventually supports the full treatment process, from intake to discharge. 

In practice this means that for our clients we often start with summarizing and searching records, moving to recurring documents like treatment coordination preparation and discharge letters, and eventually expanding into broader support across the full treatment process. Not because the vision isn't there, but because the first step has to be concrete and recognizable to everyone using it day to day.