Sometimes the smartest AI project is the one you decide not to build.
0 Volt starts every engagement with a small, fixed-scope Signal Check: we run your real data through the models to find out whether the result you're hoping for is really in there to begin with. That way the big decision comes after the evidence, rather than ahead of it.
Ours begin by checking whether the data can carry the idea at all.
Large AI programmes typically ask leaders to commit budget and headcount before anyone has confirmed the thing is even possible. We prefer to work the other way around: before a line of production code is written, we put your real data through the models and find out whether the result you're after is in the data at all.
When the go or no-go finally arrives, it rests on evidence you can see rather than a promise you have to take on faith. If the signal is there, we move on to building it; if it isn't, you've learned that for a fraction of what a full programme would have cost you.
Three very different answers, all from one client.
A go, a no-go, and an idea bigger than the one they came in with. That is exactly what starting small is supposed to surface.
Reading take-offs straight off engineering drawings
A short feasibility test showed enough promise to justify building it, and it now reads bearings, heights and dimensions directly off engineering drawings. For each feature we trial a couple of competing approaches (MLP against text, raster against vector) and regression-test them against each other, settling on whichever is the simplest thing that holds up.
Turning incoming emails into job bookings
The question was whether we could read incoming emails and turn them into bookings without a person in the loop, so we tried it. The honest answer was no: their standard operating procedure simply wasn't detailed enough to trust a model with, and we said so rather than paper over it. The "no" still earned its keep, because it pointed us at a better idea: sync the mailbox to the ERP, let the system learn each job by watching it happen, and have it ask a human only when someone does something the SOP never anticipated. We haven't built that yet; it's working its way through their budget process now.
One model doing the work of three tools
They'd also asked us to stitch a handful of standalone tools into a single workflow, but working through it we realised three of those tools were really the same model being computed three times over. Because building from scratch costs so much less than it used to, we handed back a question worth sitting with rather than a quote: whether one purpose-built thing wired into the ERP they already own would beat holding the old tools together with tape.
Early engagements with one client; happy to walk you through specifics under NDA.
A well-evidenced “no” is just as valuable as a yes.
Once momentum builds, every AI idea starts to look like a project that simply has to happen, but not every one of them deserves the budget, and the expensive way to discover which is which is twelve months and a fully funded programme down the track.
A short, fixed-scope sprint can invalidate a shaky assumption long before it hardens into a roadmap, which protects the budget, the team, and your own credibility for the bets that genuinely are worth making. We measure ourselves on the quality of the decision you walk away with, not on whether the big build happens to go ahead.
Four low-risk ways to start before you commit.
Each one is fixed in scope and price, and each ends with something tangible in your hands rather than another slide deck.
Discovery Workshop
For teams who know AI matters but aren't yet sure where it fits. We sit down with your workflows, map how the work really moves, and come away with the handful of opportunities that look worth putting to the test.
Signal Check
Before anyone commits to building anything, we run your real data through the models to see whether the signal you're hoping for is there to be found. It's an analysis exercise, a mix of regressions and structured testing through the LLMs, and at the end of it you're free to decide on the build with evidence in hand instead of ahead of it.
Prototype a Slice of Value
Once the Signal Check confirms there's something real to build, we build a small but working slice of it. Real software your team can put their hands on and react to, well before the larger commitment is on the table.
End-State Visualisation
A demo or interactive mockup of the finished idea, built to help leaders picture the destination and secure internal buy-in.
The ground we've covered for ourselves.
Led by someone who has actually shipped the things he's advising on.
0 Volt is led by an engineering leader whose background runs through continuous delivery and distributed systems, alongside the slower work of bringing teams up with them. That work has spanned financial services, healthcare, mining, and enterprise software, and it tends to keep the AI conversation honest about what a business can carry. When an engagement calls for a particular specialist, a small network of them comes in, and not before.
No significant commitment is asked for until the evidence has properly earned it.
Fixed-scope work that's designed to fail cheaply, so the only things you go on to build are the ones worth building.
A straight recommendation either way, backed by working software whenever the engagement calls for it.
Wondering whether an AI
idea actually holds up?
Tell us what you're weighing up, and we'll give you an honest read on whether it's worth pursuing and what the smallest sensible first step would look like. If we think it won't work, you'll hear that from us just as plainly.