Three rules. No exceptions.
Every studio has values. Most are vague. These three are the rules we actually use to decide what we'll do, what we won't do, and how we'll do it — every day, on every engagement.
Ship products, not slides.
Strategy decks, journey maps, opportunity matrices — they're easy to produce and easy to applaud. None of them generate revenue. None of them are honest until someone has to build the thing they describe.
Every VIZULABS engagement ends with a deployed system in production. A dashboard your operators actually open. An API your other services already consume. A pipeline that's already moving data. That's the bar.
What this means in practice
- · No artefact-only phases. Discovery is ≤ 1 week.
- · Production parity from day one — no mocks.
- · Decisions made by the person writing the code.
- · "We'll figure it out in dev" is a smell. Spec it.
Lesson learned
A three-month project we took on a verbal scope. Halfway through, the founder remembered a third user role nobody had mentioned. Two weeks of rework.
Senior craft, every commit.
Most agencies pyramid: senior on the pitch, mid through the middle, junior in the weeds, seniors firefighting at the end. Result: uneven craft, polite reviews, a system you don't fully understand by the time you inherit it.
We don't pyramid. The person who scopes the project is the same person who designs the schema, draws the components, writes the tests, and pushes to production. One senior brain holds the whole context. Velocity is high because there are no handoffs.
What this means in practice
- · Project teams cap at three. Usually one or two.
- · No "delivery managers" between you and work.
- · You meet the team on call one. They stay.
- · If we lack a skill, we'll recommend who to hire.
Lesson learned
A five-person team with one nominal "delivery manager." Two months in, decisions started routing through them. Velocity halved, blame got soft. Never again.
AI as leverage, not theatre.
The interesting question about AI isn't "can we add a chat to this?" — it's what work does this make possible that wasn't possible before? Agentic orchestration. Embeddings search across messy docs. Code-gen that closes 6-hour tasks in 20 minutes.
We refuse AI where it would replace judgement, taste, or accountability. A designer's eye is not a diffusion prompt. A CFO's read on a quarter is not an LLM summary. We use AI where it removes friction; we get out of its way where it would remove value.
What this means in practice
- · Every AI feature gets an evaluation suite.
- · Cost guards on every model call.
- · Provider abstraction by default.
- · No PII in prompts unless contractually agreed.
Lesson learned
A client asked us to "use AI everywhere." We added it to a calendar UI. It was slower, more confusing, and 11× more expensive. Rolled it back. Now we ask: leverage or theatre?
Five quieter rules.
Not as loud as the three above, but they're how those three actually hold up under deadline pressure.