UCL BARTLETT · AUGUST 2026

Does voice-first capture improve the consistency of survey data?

Alix builds repairs intelligence tools for social housing. Walk and Talk lets a surveyor dictate a property survey as they walk it, so the report is drafted before they get back in the van. We think it produces better evidence than typing one up afterwards. We cannot credibly prove that about our own product, which is why we are looking for a research partner.

Watch the 60 second demo

Get the app on TestFlight

THE STUDY

What we would want to find out, and how

This is a sketch, not a protocol. The design is exactly the part we would want a research partner to pull apart, and if the answer is that it does not stand up, that is the most useful thing anyone could tell us.

DESIGN

A within-subject crossover. The same operatives survey matched properties, in randomised order, once with voice-first capture and once with typed reporting. Properties stratified by archetype and defect type.

GROUND TRUTH

An independent expert survey on a subsample, so the study measures accuracy against a reference standard rather than only agreement between the two conditions.

MEASURES

Completeness against the hazard prompt set; agreement between surveyors on the same defect; photographic evidence captured per report; time from visit to submitted report; downstream rework and work order accuracy.

CONFOUNDS

Order and learning effects, novelty, surveyor experience, property heterogeneity. Handling these properly is the hard part, and the reason we are asking rather than doing it ourselves.

WHY IT IS OPEN

Speech interfaces for field data capture are barely studied in the built environment, and the sector now has a regulatory reason to care about the quality of evidence rather than only the speed of collecting it.

WHY WE CARE

If it holds, the case to a landlord stops being about time saved and becomes about the quality of the evidence. If it does not hold, we would rather know now than find out from a customer.

WHERE WE ACTUALLY ARE

Pilot stage, which is the opportunity

Walk and Talk is in trials with a small number of landlords, and there is no paired voice-and-typed corpus yet. So this would not be a study analysing data after the fact. It would be designed into a pilot before that pilot starts, which is a rare thing to be offered.

What Alix offers

Surveyors using the tool on real properties, in live trials

Historic typed reports from the same teams, as the comparison arm

The chance to shape what gets captured, before it is captured

Engineering time to instrument the app for the study

Introductions to landlord partners for site access and information governance

What we would ask

An academic sponsor willing to lead a funding application

Help designing the paired condition, so order, novelty and property mix do not confound it

A ground truth and inter-rater protocol

Co-authorship on what comes out of it

Independent evaluation we can put in front of landlords

WAYS IN

Four routes, from a coffee to a funded programme

STUDENT PROJECT · NEXT INTAKE

One MSc dissertation on this study. No funding required, and the fastest way to test whether the fit is real.

ADVISORY · IMMEDIATE

A named academic on our advisory board, or the study design written up as a paper together.

IMPACT ACCELERATION · WEEKS, NOT MONTHS

ESRC or EPSRC funding through UCL Innovation and Enterprise to run the study once as a pilot.

KTP · ROUND 3 CLOSES 14 OCTOBER 2026

Innovate UK Knowledge Transfer Partnership, 12 to 36 months, with UCL as lead applicant and Alix as the business partner. SMEs are funded at up to 67 per cent of costs.

Tell me whether the design holds up

If the answer is that it does not, that is the most useful thing anyone could say to us. If the other open questions interest you more, say so: triage accuracy against surveyor HHSRS ratings, how false negatives are distributed, time to remediation, and what predicts mould recurrence are all sitting there too.

Email Bee