Key Takeaways
- CAN Community Health built a governance structure, an AI steering committee and staff surveys, before implementing any AI tool, not after
- The rollout produced measurable results: roughly $17,000 in annual savings across more than thirty centers in six states
- Adoption moved fastest where clinic-level champions vetted the tool themselves, not through an organization-wide mandate
Language access utilization data, meaning encounters by language, response time and patient satisfaction by language group, tells a CHC operations leader more about where care is breaking down than almost any other dataset already sitting in their systems. CAN Community Health, a nonprofit network operating more than thirty health centers across six states, built a governance structure around exactly this data before it ever built a dashboard and the numbers that came out of it (a 92 percent patient satisfaction score, roughly $17,000 in annual savings) show what happens when a CHC decides what to measure before it decides what to buy.
1. What "language access utilization data" actually means in practice
Every interpreted encounter generates four data points, whether or not anyone is looking at them. The gap is not the data itself. It is the absence of a structure that decides what to do with it before the rollout starts.
1.1 The four data points every encounter already generates
Every interpreted encounter produces which language was requested, which modality handled it (AI, phone, video or in-person), how long it took to connect and whether the patient came back. None of this requires new instrumentation. It requires someone deciding to look.
1.2 Why CAN Community Health started with governance, not a dashboard
At CAN Community Health, Vice President of Care Delivery Innovation Katy Wendel described this gap directly on the Care Culture Talks podcast. Her organization did not start with a dashboard. It started with a governance structure built to generate one: an AI steering committee, staff surveys measuring fear and friction before training and a deliberate decision to consolidate vendors rather than adding point solutions. "We want to have a holistic view of what we're bringing into our organization," Wendel said. That holistic view is what eventually produces measurable data.
2. The governance layer comes before the dashboard
A CHC that wants to manage language access with data needs a structure that decides what to measure before it decides what to buy.
2.1 What the steering committee actually evaluates
CAN Community Health's steering committee, in place for about a year and a half, includes clinical, technical and data staff specifically so language access decisions get evaluated against the same criteria as any other technology investment like ethical guidelines, HIPAA compliance and vendor consolidation to avoid duplicated spending.
2.2 Turning staff fear into training data
That structure matters because it changes what gets measured. Instead of tracking "did we buy an interpretation tool," the committee tracks adoption friction directly. Before training even began, CAN surveyed staff about their fears (job displacement, hallucination risk, loss of empathy) and built training around the actual responses rather than a generic rollout. That survey data became an input to the rollout plan, not an afterthought collected after the fact.
3. Champions move the adoption curve and the curve is measurable
Every CHC operations leader has watched a good tool fail to get used. The variable that predicts whether it gets used is rarely the tool itself.
3.1 The North Miami example
CAN Community Health's clearest example came from its North Miami clinic, which serves a large Haitian Creole speaking population. Haitian Creole carries real translation complexity and staff there were skeptical going in. Instead of mandating adoption, the organization brought in Haitian Creole speaking staff members as champions to vet the tool directly. "They went out and then said this works," Wendel explained. "This is getting our patients into care faster." That single data point, patients getting into care faster in the toughest language population in the network, did more to move adoption than any policy memo could have.
3.2 Tracking adoption by clinic, not just organization wide
The lesson generalizes past this one clinic. If a CHC wants adoption data that actually moves, it needs to track usage by clinic and by language, not just organization wide. A network average can mask a specific site where adoption stalled and a specific site where it took off and the site where it took off almost always has an identifiable champion behind it.
4. What the numbers looked like once CAN Community Health measured them
The results are worth stating plainly because they show what happens when governance produces real numbers instead of a general sense that things are going well.
4.1 Patient satisfaction and cost
CAN Community Health's patient satisfaction scores rose to roughly 92 percent, a figure Wendel describes as high for the industry, tracked through the organization's standard patient survey rather than a special AI specific instrument. Switching to flat rate interpretation billing saved the nonprofit organization approximately $17,000 annually, a number Wendel called small in absolute terms but significant for a nonprofit's budget discipline.
4.2 Staff-reported gains across roles
Staff across roles, not only clinicians but front desk, case management and behavioral health specialists, reported faster intake and clearer oversight of what was communicated to patients. None of those numbers required new software to generate. They required someone deciding in advance what would count as success and then measuring against it consistently. That is the entire discipline of language access utilization data: decide the metric before the rollout, not after.
5. Building a dashboard both a CFO and a compliance officer will read
Section 1557 requires meaningful access across every patient touchpoint, not only the clinical encounter itself: scheduling, intake, consent, discharge instructions and follow up. A dashboard built to satisfy a compliance officer needs to show coverage across those touchpoints. A dashboard built to satisfy a CFO needs to show cost per encounter, consumption and the shift away from per minute billing pressure.
5.1 The three fields that matter most
For a CHC building this out, three fields matter more than any others: modality used per encounter (so leadership can see the actual AI to human ratio rather than assume one), patient choice rate (how often patients select a human interpreter when both are offered, which is itself an equity signal worth tracking) and time to connection by language (which surfaces exactly where a language gap in staffing or technology is costing patients time). The CHCF report on AI and language access frames this same tension for California safety net providers: technology adoption without a measurement plan tends to produce anecdotes, not data.
5.2 No new hardware required
None of this requires new hardware or a separate reporting system. It requires CHC leadership treating language access the way they already treat patient volume or no-show rates: as a number reviewed on a cadence, not a policy stated once and left alone.
The bottom line
Language access data was never the hard part. Every interpreted encounter generates it by default. The hard part is deciding, before rollout, what will count as success and building the governance to review it consistently afterward. CAN Community Health's experience across thirty-plus centers shows what that discipline produces: measurable adoption, measurable savings and a satisfaction score most safety net providers would be glad to report. Explore how No Barrier's 295+ language access options fit into a data-driven language access program or see how a similar approach played out in No Barrier's federally qualified health center case study.
Sources
- Katy Wendel, Vice President of Care Delivery Innovation, CAN Community Health. Interview on Care Culture Talks.
- U.S. Department of Health and Human Services, Office for Civil Rights. Section 1557 guidance on federal financial assistance and language access.
- California Health Care Foundation. AI and Language Access, March 2026.
.svg){kind=small}