Practice 03 / Applied R&D

Research that ends in something you can ship.

We take an unproven idea, build the part that would fail first, and tell you plainly whether to continue. When the answer is yes, the same team builds the production version.

Back to all services
When it applies

When R&D is the right shape of work.

Research is not a substitute for delivery. It is what you do when the delivery plan depends on something nobody has tested yet.

  1. 01

    Nobody can say whether it is possible

    The idea is credible and untested. A programme cannot be planned around it until someone builds the part that would fail first.

  2. 02

    The business case depends on an assumption

    One unproven number carries the model. We build the narrow thing that tests it, so the investment decision rests on evidence rather than optimism.

  3. 03

    Two approaches both look reasonable

    Argument does not settle it. Building both to demonstration depth does, and it is usually cheaper than committing to the wrong one.

  4. 04

    A prototype needs to become a product

    Something works on a laptop and nobody will put it in front of customers. That gap is engineering work with a known shape, and we do it often.

Engagement types

Four kinds of engagement,
each with a defined end.

We will tell you which one your question needs, and we will say when the honest answer is that no research is required.

Feasibility study

A written position on whether the thing can be done, what it would take and where it would break. No build, or only enough to test the risky part.

Technical assessmentRisk registerArchitecture positionEffort modelRecommendation

Prototype

Something people can use, built to explore the shape of the product rather than to carry load. Made to be thrown away without regret.

Interaction prototypesWorking front endsSimulated back endsUser testing support

Proof of concept

A demonstration that a specific mechanism works end to end. Scope stops at the proof, deliberately, rather than drifting towards a product.

End-to-end demonstrationBounded scopeInstrumented resultsGo or no-go position

Prototype to production

Taking a proven idea and rebuilding what has to be rebuilt. Security, data handling, testing, deployment and support, with the original intent preserved.

Production architectureSecurity reviewTest coverageCI/CDHandover and support
Evidence

Proving AI survey capture before anyone built it.

The question was whether research study data could be collected by AI voice call, web and mobile, with photographic capture that stays comparable between rooms and handsets. We built the proof rather than the product.

Proof of concept

What the demonstration established

  • An automated voice agent runs a research survey conversationally over a phone call and returns the answers as structured data, with no coordinator on the line.
  • The same survey is delivered across voice, web and mobile, so channel choice sits with the participant rather than with the study.
  • Calibrated photographic capture on ordinary phone hardware, using a printed colour reference card in frame to correct for lighting and camera.
  • Scope stopped at the proof. Each path was taken to the point where it could be run end to end in front of an audience, and no further.
  • The proof of concept makes no diagnostic, clinical or regulatory claim, and no pilot metrics were produced or are asserted.

The capability built here, in AI voice interaction, cross-platform delivery and camera colour calibration, is now available to client work.

How it runs

Bounded, evidenced and easy to stop.

A research engagement should be the cheapest thing you buy from us, and the one most likely to change your plan.

  1. 01

    Write down the question

    One sentence, agreed before anything is built, stating what has to be true. Everything in scope exists to answer it. Everything else waits.

  2. 02

    Find the part that would fail

    We build the riskiest element first. A prototype that proves the easy half has told you nothing you did not already believe.

  3. 03

    Set the stopping rule

    Time and scope are fixed at the start, with the criteria for success written down. That is what keeps a proof from quietly turning into an unsupported product.

  4. 04

    Report honestly

    You get what worked, what did not, and what remains unknown. A negative result delivered early is a good outcome, and we will say so plainly.

  5. 05

    Cost the real build

    If it continues, the same engineers scope the production version. The second build starts from the code and the knowledge, not from a handover document.

Deliverables

What you receive,
and what it protects.

Every engagement ends with something you can run and something you can circulate.

What you receive

Small engagements, concrete outputs.

  • A working demonstration you can run in front of your own stakeholders
  • A written technical position covering approach, risks and limits
  • An architecture position and effort model for the production build
  • The source code and the environment it runs in, in your accounts

What it protects you from

The cost of finding out late.

  • Committing a programme budget to an assumption nobody tested
  • A prototype presented as production-ready because it demonstrated well
  • A research result that cannot be reproduced or explained afterwards
  • A second supplier rebuilding from scratch because the first left nothing usable
Capability

What we have already built.

Capability developed on our own products and proofs, held in-house and available to client work.

AI voice interfacesComputer visionCamera colour calibrationGenerative image & videoCross-platform mobileConsumer web productsOffline-first architectureMarketplace platformsSimulation & benchmarkingThird-party API integrationCloud prototypingRapid front-end builds

Test the assumption
before you fund it.

Tell us the one thing your plan depends on. We will scope the smallest piece of work that settles it.