Canvas hosted a webinar for customers and our community on what AI is changing about how healthcare software gets built and run. Here’s the vision, and why the architecture underneath it matters.
Most conversations about AI in healthcare fold into a single question: can the model do the task? The way we’re building at Canvas reveals why that’s the wrong frame.
There are two distinct shifts happening, and they’re easy to conflate.
The first is already here: AI has changed how healthcare organizations build on and adopt Canvas. What used to require an engineering team and a development cycle now happens in a single conversation. A team can identify a painfully manual workflow at 9 a.m. and have a working version running in their development environment later that day. Since Canvas introduced Studio at its BUILD summit in May, customers have built and deployed over 5,000 plugins. More than half of all Canvas customers have already shipped something with Studio.
The second shift is earlier, and bigger. AI agents are beginning to join the care team itself, doing real clinical and operational work inside the EMR rather than just helping people build it. This is where the models are only now becoming reliable enough to trust, and where Canvas is being deliberately careful.
A hierarchy of needs for AI in the EMR
Canvas frames its roadmap using an adaptation of Maslow’s hierarchy, a model borrowed from Sierra CEO Bret Taylor.
At the base: agents that help you adopt Canvas. Getting live, seeing patients, moving operations in. Navigator and Studio already do this = broad adoption.
In the middle: agents that help you improve what you’ve already adopted. Proactively recommending plugins, flagging errors in production, suggesting configuration changes. This is the shift from AI you prompt to AI that suggests. It’s surfacing now through Platform, Canvas’s newly launched home for the tools that manage Canvas.
At the top: agents on the care team, doing the work. This is the frontier, and it’s where the two agents Canvas previewed, an HIE reconciliation agent and Hyperscribe V2, live.
Why the architecture is the differentiator
Here’s the part a prospect evaluating EMRs should sit with.
Any modern platform can offer three kinds of agents:
- First-party agents the vendor builds and you buy
- Second-party agents you build yourself
- Third-party agents you purchase through a catalog.
Canvas supports all three, but so does the competition. That isn’t the differentiator; the difference is what happens when you run more than one agent at once.
In a typical EMR, agents reach the system through API integrations. One agent integrates with the EMR’s patient state, reading and writing. Add a second agent doing the same thing, and you’ve created the problem of a duplicated state: what’s the source of truth, and how do two external integrations stay coordinated with the EMR and with each other? Every agent you add compounds the complexity.
In Canvas, agents run inside the EMR runtime, on the same SDK and tools. There’s no duplicate state, because it’s actually a shared state, with coordination fully built in. And governance, safety, and reliability follow from that architecture rather than being tacked on as an afterthought.
The practical implication: deploying a single-point solution agent isn’t where Canvas is winning. Plenty of tools do that just fine. The true advantage expands as you coordinate multiple agents into a system, which Canvas believes is where every healthcare organization is heading. The more agents you run, the more the architecture matters.
Built on a certified foundation, priced to outcomes
Underlying all of this is an ONC-certified EMR that’s been delivering outpatient care for over a decade, with the SDK that lets each customer effectively build their own EMR on top. Canvas isn’t a foundation-model provider; it uses leading models, primarily Anthropic’s, as a layer on that platform.
First-party agents are priced to outcomes, not seats. Hyperscribe V2 is priced per visit, because its job is a ready-to-sign, billable note. The HIE agent is priced per chart reconciliation, because that’s the work it does.
The throughline: reliability above all
Canvas is prioritizing reliability and quality over speed. In a market full of claims about what AI can do, restraint itself is a position of conviction and confidence, and nowhere is it more integral to earn trust than in building clinical software.
The call to action is simple: go try it. If you’re curious what it looks like when building into your EMR requires a conversation instead of a roadmap, Canvas is the singular place to start.
