Today we are opening the doors on InTouchLabs, the research arm of InTouchNow AI. This page, this site and this post are our first public release under the InTouchLabs name, and there is a great deal more behind them.
Why a lab, and why now
InTouchNow AI has answered patient calls for NHS general practice since 2019. We started with people, and we built voice AI only once we understood the operational job it had to do. Today AI answers more than 200,000 patient calls every month across 150+ GP practices, at every hour of the day, including weekends and bank holidays.
That scale creates something rare in healthcare AI: a live environment where research questions arrive daily, in the words of real patients and real practice teams. Which accents still trip speech recognition? What does a safe translated conversation look like? Which requests could become completed actions in the practice's own clinical system? How do you prove any of this before it touches patient care?
InTouchLabs exists to answer those questions properly, and to publish as we go.
What we will study
The lab runs four research programmes, each grounded in what practices need and each feeding directly into what InTouchNow AI delivers.
- Voice. Natural conversation on every line: turn taking, interruptions, regional accents, noisy lines and callers who need more time.
- Translation. Real time speech translation so patients can speak to their practice in the language they think in, with terminology that stays safe across languages.
- Computer use. Agents that operate practice software the way a trained member of staff does, completing workflows in clinical systems such as EMIS and SystmOne with a full audit trail.
- Evals. Clinical grade evaluation of healthcare AI: benchmarks, adversarial testing and regression suites that measure safety, accuracy and fairness before any model handles a patient call.
How we will work
Research here follows one loop: observe a real problem in practice, build with frontier models, prove the result against our evals framework and the standards NHS buyers expect, then deploy through InTouchNow AI with support after go live. Governance is not an afterthought: UK GDPR, clinical safety documentation to the DCB0129 and DCB0160 standards, and alignment with the Digital Technology Assessment Criteria apply to lab work just as they do to the delivery business.
One line will not change: AI does not triage, diagnose, advise or prescribe. The practice team stays in charge at every step.
What is coming
There is a lot going on in the lab right now. Over the coming months we will write up early results from the voice and translation programmes, share how we evaluate healthcare AI before it takes a patient call, and talk about the partners we are working with across UK healthcare. New writing will land on this blog first.
Work with us
Three doors, all open. If you are a GP practice, a Primary Care Network, an Integrated Care Board or an NHS team and want to talk about the research we do, write to us. If you build healthcare technology or run health services anywhere in the world and our work is relevant to yours, we would like to compare notes. And if you are simply interested in healthcare AI, as a researcher, clinician, engineer or student, engage with the lab: challenge our thinking, propose an evaluation, or follow the work as it is published.
You can reach the whole lab at hello@intouchlabs.ai. We read everything.
The InTouchLabs team