Skip to main content
ProtoComet
ProtoComet

  • Get AI into production on data you can trust.

    Turn disconnected data into reliable insights and practical AI capabilities. From data platform architecture to AI readiness and implementation, we help you build the foundations for better decisions and meaningful business outcomes.

Where this actually breaks

Most businesses hit at least one of these: three on the way to AI, and one in the data underneath it.

The pilot worked. Production didn’t.

The demo ran on a clean sample. The real data sits across SaaS tools, spreadsheets and a database with no API, and nobody budgeted for connecting them.

AI costs arrived before AI value did

Licences and usage bills started on day one. Adoption stayed low, use cases were never ranked by payback, and nobody can show what the spend has returned.

Nobody knows what the AI is allowed to see

Access was granted to get things working, not to keep them safe. Once a copilot or agent connects, it can show salaries, contracts or customer records to anyone who asks the right question.

Every meeting starts with whose numbers are right

Every department pull from their own system and spreadsheet, so reports disagree. Answering a simple question wants an analyst to stitch exports together by hand, and the answer is outdated by the next meeting.

Four services. Start with the one you need.

Each service runs on its own, and you can engage for just one. If you are not sure where to begin, the readiness assessment is the quickest way to find out.
RECOMMENDED FIRST STEP

AI Readiness Index

Seven dimensions of your data and systems scored from evidence: data architecture, API surface, identity and security, observability, DevOps maturity, compute scalability and governance. The result shows which use cases you can support now, which need groundwork first.


Start here if you want AI but are not sure what your data and systems can support.

Explore
AI SOLUTIONS

AI Solutions: Architecture & Build

The production design for a chosen use case: boundaries, pattern, data path, evaluation and operations. Carried through to a working build where you need it, so it holds up on real data and real users.


Start here if you have picked a use case and need it to work beyond the demo.

Explore
STRATEGY & GOVERNANCE

AI Strategy & Operating Model

A short list of what to build, in what order, with the value and the risk written down. Alongside it, the rules for running AI safely: an inventory of what is already in use, risk classes set by use, the controls each class needs, and a named owner when something goes wrong.


Start here if you have more AI ideas than budget, or AI is already in use and nobody has set the rules.

Explore
DATA PLATFORM

Data Platform Architecture

The design of one governed data platform on Microsoft Fabric, with Purview for governance. Sources consolidated, ownership assigned, and a reporting layer the business can trust.


Start here if answering a simple question about your own business takes an analyst and a day.

Explore

What a working AI system sits on

Every AI use case rests on three layers: data you can trust, clear rules for how AI is used, and a design that holds up in production. The Readiness Index measures all three, so you know which layer needs work before money goes into the one above it. You can start at whichever layer is weakest.

Measure: across every layer

The AI Readiness Index scores seven dimensions of your data and systems from evidence, so each decision above it is made on facts. It also shows which layer to fix first, and why.


Readiness Index report

Layer 1: Data foundation

One governed platform for the data that matters, with clear ownership, a catalogue and controlled access. Usually the largest piece, and the one every AI use case reads from.


Data platform architecture

Layer 2: Priorities and rules

Which use cases are worth doing and in what order, plus the rules each one runs under: who approves it, what data it can touch, who answers for it, and how often it is reviewed


AI strategy and governance model

Layer 3: AI in production

The architecture for chosen use cases, with boundaries, evaluation and running costs decided before the first token is paid for, carried through to a working build where needed.


Solution design and build

How the work actually runs

Three shapes, all senior-only. The right one depends on whether you need a design, a second opinion on someone else's, or a standing decision-maker

Fixed-price design engagement


A defined scope, a written baseline, and a design with decision records, delivered in weeks. This is the usual entry point for Cloud Strategy & Architecture and for the governance and cost baselines. The number is agreed before it starts.

Standing design authority


A monthly retainer for the length of a programme or indefinitely: reviews, decisions, the gate, and a standing line for the questions that cannot wait for a meeting. Cancellable on a month's notice. Migration advisory almost always runs this way.

What senior-only means here


The person reading your policy assignments is the person writing your control baseline and the person your partner argues with at the gate. No associates, no handover, and no design that its author will not be around to defend in eighteen months.


What you walk away with

Whichever service you start with, the work ends in something your team can keep, act on and build from. Each service page lists its full set.

➖Readiness report. Seven scorecards with evidence, the use cases your data can support today, and the groundwork the rest need.


➖AI rules and inventory. Every AI tool in use, its risk level, the controls it needs and the person who owns it.


➖Solution design. Boundaries, data path, evaluation and running costs for a chosen use case, with the decisions behind them written down.

➖Use-case shortlist. What to build first, what to leave for later and why, and how each one will be measured.


➖Data platform architecture. The target design on Fabric and Purview, an ownership map, and a plan for moving there in stages.


➖Quarterly leadership summary. One page on what is working, what is at risk and what comes next. Included in any ongoing engagement.

Our Way

AI readiness grows out of well-governed data. That is the idea this practice is built on, and it means the most exciting part of the work usually comes later.

I have spent years on the unglamorous end of this: making three warehouses agree, finding out which of the four customer tables was the one finance actually trusted, and discovering that the reporting database everyone queried was a copy refreshed by a job nobody owned. Every one of those problems is something an AI system will run straight into before it can be trusted with a question.


So the practice measures first, and the result is allowed to say “not yet”. If the score points to a year of work on ownership and lineage, that year is what turns an AI project into something the business can rely on.

An AI system repeats whatever your data believes, only faster and with more confidence.


Yogesh, founder and principal consultant

Questions clients typically ask

Short answers to what usually comes up in the first conversation.

Yes. If you already know your data well, the work can start with the pilot. It will still begin with a short readiness check on the specific data that pilot needs, since that is where most pilots run into trouble. If the check finds gaps, you will know about them in days, before they surface in front of users.

The data platform work is built around Microsoft Fabric and Purview on Azure. The readiness, strategy and governance work use the same method on any platform, and the recommendation will say so if your situation points elsewhere. For AI solutions, the model is chosen per use case. Azure-hosted models are the usual starting point because identity, security and data residency are simpler to manage there.

Both are available. Most engagements lead with advice and design, and carry through to a working build where it helps. The aim is always a system your team can run, govern and extend after the engagement ends.

The work scales to the size of the business. For a smaller team, a readiness check and a short use-case list can take a few weeks and leave you with a clear first step. Nothing here assumes a large data team or an enterprise budget.

That is a common starting point. The work usually begins by finding out what is already in use and what data it can reach, then setting simple rules around it. From there it is easier to decide which uses are worth investing in properly.

Start with a conversation

Thirty minutes on what you are running, what you want AI to do for the business, and whether your data is ready for it. If the answer is “not yet”, you will hear that, along with what would change it.

Let’s Talk