Activity is not progress
Last of three. A process can be dramatically faster while the business barely moves. The smaller number was far less impressive and much more useful.
How we work · AI-assisted delivery
AI-assisted engagement & delivery
AI carries the work between the conversations. A named nVisionIT person owns every commitment we make to you.
Our engagement model has not changed: a conversation about the outcome you want, a Sprint 0 that puts something real in front of you, delivery in two-week milestones, and a relationship that continues past go-live. What is changing is the engine underneath it. We have built three systems of our own to take on the work that used to consume the weeks between those conversations, with the aim that Sprint 0 ends with a running prototype in weeks. The first is already at work: our delivery orchestrator has written user stories and code for sales proofs of concept and a client project, though no engagement has yet used all three together. We are bringing them into client delivery in stages, and we will tell you which of them is working on your engagement. AI assists and a person decides, and our quality checks do not change.
Not sure where to start? Take the ten-question AI readiness check →
Why we changed the engine and kept the promise
The slow part of a project was never the building
It was the months spent discovering what the requirement actually was, and the change requests that arrived once the gap surfaced. We put AI to work on exactly that part, and left the human judgement where it belongs.
Weeks
From first conversation to a working prototype you can use
Sprint 0 ends when you have used it and approved it
Every artefact
Reviewed and approved by a named engineer before it reaches you
Nothing AI produces crosses the line unreviewed
One record
Your objectives, decisions and test evidence in one governed place
Inspectable for the life of the engagement
What it is designed to fix
- Requirements that look complete until month three → gaps surfaced in the first conversation
- A discovery phase that produces documents → a working prototype at Sprint 0
- Change requests as the default commercial model → milestone acceptance, then invoice
- AI used quietly, with no one accountable → a named human owner on every deliverable
The engine
Three systems we built and own, coming into delivery in stages
You are entitled to know what is doing the work on your engagement, and to ask what it was allowed to touch.
Intelligence Hub
Our governed knowledge and decision engine. It checks your stated objectives against best practice and against our own delivery record, and every answer it gives is designed to cite its source and to say when it came from AI. The Hub informs a decision. A person makes it.
The Hive
Our AI delivery orchestrator. It turns confirmed objectives into user stories, helps flag the requirements the conversation missed, and writes code and tests against them inside the sprint discipline. It is already in use on sales proofs of concept and client work, and it is how we add delivery capacity without adding headcount or the coordination cost that usually comes with it.
nVisionIT.Framework
Our in-house application base. Every web-based build starts from it, with tenancy, identity, audit and error handling already in place, so a proof of concept reaches you quickly and the one you sign off can grow into production on the same base.
The four phases
What happens, and what you sign
The shape of the engagement is the same one we have always run. Each phase below shows what the engine does, and the point at which a person, yours or ours, puts their name to something.
- Initial discovery, made rigorous
A conversation about the outcome you want. Behind it, the Hub is designed to check every stated outcome against best practice and our governed delivery record, and the Hive helps turn the picture into user stories and flag what the conversation missed. We repeat the loop until the requirements are stable. You sign the confirmed objectives before anything is built. - Sprint 0, ending in something that runs
We build a working proof of concept from the signed requirements, on the Framework where a web application is needed, alongside the costed roadmap, prioritised backlog, target architecture and success metrics. Our engineers review and shape it before you ever see it. You approve the prototype, and that approval is the exit from Sprint 0. - Iterative delivery in two-week sprints
Plan, build, demo, deploy, retro. Where the Hive’s orchestrator is used, it writes tests with the code; elsewhere our developers add the tests the work needs. Either way, our QA reviews each milestone before it is demonstrated. You accept each milestone demo, and only accepted milestones are invoiced. Pivot at any milestone boundary. - Ongoing value review
Go-live is not a handover. The Hub is designed to track outcomes against the objectives you signed at the start, so what we learn feeds the governed record for the next engagement. You sign user acceptance and go-live before anything reaches production.
Where a person signs
The non-negotiables
No AI output crosses a commitment boundary without a named human act. A skipped gate is logged and investigated as an incident.
You sign the requirements
Before a line of the solution is built, you approve the confirmed objectives in writing.
You approve the prototype
Sprint 0 does not end, and delivery does not start, until you have used the working proof of concept and said yes.
You accept each milestone
Acceptance precedes invoicing, every sprint. Rejections and the rework they cause are visible in the next sprint record.
We review before you see it
A standing internal rule: every AI-generated artefact is reviewed and approved by a named engineer or consultant before it leaves nVisionIT.
You sign go-live
User acceptance testing is yours to run and yours to sign, before anything reaches production.
One named owner, always
Every deliverable carries the name of the person who reviewed it, approved it and is accountable for it. You can ask who, at any point.
How we test
Three layers, three different owners
Self-testing delivery is a claim worth being specific about, so here is what it actually means on your engagement.
LAYER 1
Tests written with the code
AI generates tests alongside every build and they run in the pipeline on every commit. Our engineers review them. A generated test only counts once it executes and passes.
LAYER 2
Human QA before the demo
Our QA reviews each milestone before it is shown to you. This is a person, every sprint, and it happens whether or not the pipeline is green.
LAYER 3
Your acceptance testing
You test against the objectives you signed, before go-live. Test suites, review records and UAT results are filed on your engagement record.
The guardrails
What governs the AI on your engagement
Our Responsible AI Use Policy applies to every stage above. The short version is on this page; the policy itself sets out the controls, and we will share it with you.
Zero Exposure Rule for client dataApproved tools only, from a governed registrySensitive content stays on our tenantAgentic tools only where you agree in writingNamed human reviewer on every deliverableClient AI Disclosure on requestPOPIA & Mauritius DPA alignedWCAG 2.2 AA checked
From our Insights
What we have learned doing this work
More intelligence does not automatically create more value
Second of three. We compared what we had produced against what anyone had actually asked for. A large share of it could not be matched to a question, and that changed how we choose what to work on.
The AI worked. Our assumptions did not.
First of three. We believed more output would create more value. Measuring outcomes rather than activity changed what we believed, and the biggest lesson turned out not to be about AI.
See it on your own problem
Bring us an outcome you want and we will run the first conversation properly: your objectives checked against best practice and our delivery record, with the gaps surfaced while you are still in the room.