Skip to content
About

An engineer-led Microsoft analytics consultancy

Vigiles AI is an engineer-led Microsoft data and analytics consultancy working across United Kingdom, Ireland, Germany, Switzerland, Belgium. We build semantic models, lakehouses, pipelines and the governance around them, and we say no to work that would not pay for itself.
  • Independent on licensing, no incentive to oversize a capacity
  • Documentation and handover written into every engagement
  • Distributed delivery, no nameplate offices we do not occupy
Hours of analytics engineering
30,000+
Analytics projects delivered
2,000+
Clients served
50+
Cities covered
239
Each with a published local view

The name

The vigiles were Rome's night watch: the people whose job was to notice something going wrong while everyone else slept. It is a reasonable description of what a data platform should do for a business, and of what a consultancy should do for a client, notice the problem early, say so plainly, and fix the cause rather than the symptom.

What we are

A specialist consultancy on the Microsoft data stack: Power BI, Microsoft Fabric, Azure Synapse and Data Factory, with Purview for governance. We are not a reseller, we have no product to licence to you, and we take no commission on your Microsoft spend. That matters, because it means we have no reason to recommend a larger capacity than your reader population justifies, and several times a quarter we tell someone the opposite of what a reseller would.

We work with organisations between roughly fifty and ten thousand employees, which is the range where analytics is genuinely load-bearing but there is rarely a dedicated platform team. Below that, the honest answer is usually a well-built import model and a licence. Above it you are probably buying from a systems integrator, and we are happy to say when that is the right call.

What we refuse

We do not take open-ended time-and-materials work for a first engagement, because neither side can judge value without a defined scope. We do not disable or skip tests to hit a date. We do not present fabricated figures as client evidence, where a case study is anonymised we say so, and where a number is illustrative we label it. And we will not take on support for a platform we have not been allowed to document first, because that is how people end up owning somebody else's outage.

How we price

Discovery is fixed-price and paid, typically in the low five figures, and the output is yours whether or not anything follows. Build work is fixed-scope wherever the scope can be fixed, and day-rate where it genuinely cannot, a migration wave, for example, once the inventory is known. Managed service is a monthly retainer sized to the estate, with unused change days rolling over for one month so nobody is pushed into make-work.

Where we work

The United Kingdom, Ireland, Germany, Switzerland and Belgium. Delivery is remote-first with on-site workshops at the points where being in the room changes the outcome: definition sessions, validation and go-live. We work in English and German, and handle Dutch and French-language stakeholder sessions through our network. We publish a local view for every city we cover, the industries, the metrics, the licensing arithmetic and the compliance position, rather than claiming offices we do not occupy.

Who you actually get

The person who scopes the work is an engineer who will also do some of it. We do not run a model where a principal sells and a graduate delivers; it produces the documentation problems we spend half our time repairing in other people's estates. On a typical engagement you get a lead analytics engineer, an architect part-time on larger programmes, and a named backup who already knows your estate.

Principles

How we work

Each of these is something a less disciplined consultancy would charge more for, or avoid saying out loud.
  1. Definitions are agreed before reports are built

    We do not build a revenue report until the business has agreed what revenue means, who owns that definition and where the exceptions sit. Most reporting disputes turn out to be definitional rather than technical.

  2. The smallest architecture that meets the requirement

    We propose the least expensive design that satisfies the requirement, and we say so when a capacity, a lakehouse or a migration is not justified. Overbuilt platforms are easier to sell and considerably harder to operate.

  3. Claims are supported by measurement

    Performance work is reported with before and after traces. Migrations are signed off against reconciliation evidence. AI features are assessed against a fixed evaluation set. Where something cannot be measured, we describe it as a judgement rather than a result.

  4. Documentation is produced during delivery

    Architecture decisions, measure definitions and runbooks are written as the work proceeds and kept in your own systems, rather than assembled in a final week before invoicing.

  5. Handover is planned from the start

    The work is complete when your team can change it without us. We are glad to continue on a small retainer, but the platform should not depend on our presence to keep running.

  6. We say when a purchase is not justified

    Several conversations a quarter end with a recommendation that is not a project: a licensing change, a process fix, or an internal hire. We would rather give that answer than sell work that will not pay for itself.

Engagement model

The engagement model

  1. Discovery

    2 to 3 weeks

    Decision inventory, source profiling, data quality findings and costed architecture options. Fixed fee, and the output is yours regardless of what follows.

  2. Foundation

    4 to 8 weeks

    Landing zone, environments, deployment pipeline and the first end to end reference implementation, which becomes the pattern everything else follows.

  3. Delivery

    Quarterly increments

    Domain-by-domain delivery, each validated against the source of record and signed off before the next begins. Value lands early and repeatedly.

  4. Operate

    Ongoing or handover

    Either a managed service with monitoring and agreed response times, or a documented handover to your team with a defined support window.

Stack

Technologies we work in

Analytics and semantic layer

  • Power BI
  • Tabular Editor
  • DAX Studio
  • Power BI Embedded
  • Paginated reports
  • Excel semantic connection

Platform and engineering

  • Microsoft Fabric
  • OneLake
  • Fabric Warehouse
  • Fabric Lakehouse
  • Azure Synapse Analytics
  • Azure Data Factory
  • Azure Databricks
  • Delta Lake
  • dbt
  • Spark

Data stores and integration

  • Azure SQL Database
  • SQL Server
  • ADLS Gen2
  • Cosmos DB
  • PostgreSQL
  • Dataverse
  • Azure Functions
  • Logic Apps

Governance and security

  • Microsoft Purview
  • Microsoft Entra ID
  • Sensitivity labels
  • Azure Key Vault
  • Azure Policy
  • Defender for Cloud

AI and data science

  • Power BI Copilot
  • Azure OpenAI Service
  • Azure AI Search
  • Azure Machine Learning
  • Fabric Data Science
  • MLflow
  • Python

Delivery and operations

  • Azure DevOps
  • GitHub Actions
  • Terraform
  • Bicep
  • Azure Monitor
  • Log Analytics
  • Power Automate
Next step

Arrange a technical review

A 45-minute working session with an engineer rather than a sales call. You describe the problem; we set out what we would build, in what order, and what it is likely to cost.

Answered within one working day, by the engineer who would do the work