In February 2026, the Fraser Institute published a study from the University of Toronto's Rotman School of Management, titled How Implementing System-Wide Solutions Can Amplify the Impact of Artificial Intelligence on Health Care. Its author, Avi Goldfarb, holds the Rotman Chair in Artificial Intelligence and Healthcare, and argues that AI can produce substantial gains in healthcare productivity, but only when it changes how care is organised rather than being inserted into workflows designed long before AI existed. Most AI tools sold to Canadian clinics today fall into the second category, which the study calls a point solution: an AI application dropped into a single step of an existing process without altering anything else around it. A point solution can make one task faster, but the workflow surrounding that task stays exactly as it was, which limits how much benefit a clinic actually experiences relative to what the technology could deliver.
Why a Faster Single Step Does Not Fix a Slow Clinic Workflow
The paper's central example is the AI scribe that listens to a physician's conversation and drafts a clinical note, a tool that genuinely saves time on that single task. Its argument concerns everything the point solution leaves untouched. A true system solution would not stop at the note. It would also record the encounter, structure the note into the correct clinical format, update the patient's medical record, arrange any necessary follow-up, coordinate with the pharmacy, and flag a concerning lab result to the right person, with every step functioning as one connected process rather than a series of manual handoffs between disconnected tools.
The study illustrates this with a well documented episode from economic history. When the electric motor became commercially available in the 1880s, factory owners largely replaced their central steam engine with a central electric motor and left the rest of the factory unchanged: the same layout, the same machines, the same worker routines built around a single central power source. Meaningful productivity gains did not appear for decades, not because electricity had been oversold, but because nobody had yet redesigned the factory floor around what the new power source actually made possible. Healthcare is repeating the pattern, with AI layered onto workflows built before AI existed, producing the same modest, incremental gains that electricity did in its first decades of use.
A separate global analysis published by McKinsey in late 2025, titled "Healthcare AI: From Point Solutions to Modular Architecture," reached a closely related conclusion: the rapid, uncoordinated adoption of point solutions has created a fragmented technology environment with its own operational friction, and the organisations best positioned to benefit from AI will be the ones that move deliberately toward connected, modular systems rather than accumulating more disconnected tools.
What a System Solution Actually Requires
A system solution, in this framing, requires a genuine change in how work is organised, not a faster version of the same process. That means connecting the data generated at every point of patient contact rather than leaving it siloed inside whichever tool happens to handle one step, and it means clinics trusting that connecting these pieces will simplify their operations rather than adding another system to learn. This is where a great deal of well intentioned AI adoption goes wrong in practice. A clinic that buys a scribe from one vendor, a booking tool from a second, and a patient communication platform from a third has not built a system, however capable each tool is on its own. It has assembled three point solutions, each with its own login and its own place where the same patient's information has to be entered again by hand, which is exactly the fragmentation this study and the McKinsey analysis both identify as the real obstacle to AI improving healthcare productivity.
Building a Connected System, Not a Collection of Point Solutions
Avoiding exactly this outcome shaped how Elevation Labs built JOUD Health: a unified clinic technology platform designed as one connected operating layer for a clinic, not a set of features assembled under a shared brand.
The AI voice agent that answers a clinic's calls, the digital intake and check-in workflow that gathers patient information before arrival, and the automated confirmation and reminder system that reduces missed appointments all draw on the same data layer and the same connection to the clinic's EMR. A booking taken by the voice agent, an intake form completed at home, and a confirmation sent the night before all update one shared patient record, not three separate ones that later have to be reconciled by hand. A patient who calls to reschedule does not create a gap between what the phone system has recorded and what the front desk can see.
That connectivity produces effects a single purpose tool cannot replicate. Because call handling, intake, and the medical record already sit inside one system, staff also get a live view of which patients are waiting, ordered by clinical urgency rather than arrival time, with conflicts such as a same day booking with two physicians flagged automatically before they become a billing problem. Nobody set out to build that as a standalone feature to sell. It exists because the underlying pieces were already connected, which is close to exactly what this study argues point solutions cannot produce.
What This Means for a Clinic Evaluating AI
It is worth being precise about the limits of this comparison. The paper is addressed to policymakers, and the redesign it proposes, changes to scope of practice, payment models, and regulation across an entire healthcare system, is far larger than anything one clinic or vendor can accomplish alone. No platform is going to resolve the structural questions it asks of Canadian health policy.
What does translate to an individual clinic is the test his argument implies for evaluating any AI tool before buying it. Ask what happens to the information a tool produces once its task is done: does it flow automatically into the systems the clinic already depends on, or does it sit inside its own interface until a staff member carries it somewhere else by hand. A tool that cannot answer that well is a point solution, however capably it performs its one task, and the clinics that benefit most from AI over the next several years will be the ones asking that question before they sign a contract rather than after.
Canada's shortage of family physicians, examined in The Family Physician Shortage in Canada: What It Means for Independent Clinics, will not be solved by software alone. But recovering the administrative capacity lost to disconnected systems is one of the more practical levers available to Canadian clinics right now, and realising that gain depends on choosing tools built from the outset to operate as one system rather than a growing collection of point solutions that each solve their own small problem in isolation.
Elevation Labs builds clinical-grade operational infrastructure for Canadian primary care. Book a demo to learn more.