Before an organisation builds AI on Microsoft Azure, its board usually wants three questions answered. Are we ready? What should we build first? And what will it cost to run safely? An AI readiness assessment answers them with evidence from the organisation’s own estate, before money is committed to a build.
What an AI readiness assessment is
An AI readiness assessment is a short review, with a fixed scope, of an organisation’s technology estate, data, governance and use cases, measured against what AI agents need in order to run in production. It ends with a prioritised plan and a decision for the board.
What it should cover
Microsoft’s Cloud Adoption Framework sets out AI adoption in six phases: strategy, plan, ready, govern, secure and manage. Its guidance for the ready phase, for example, recommends separate management groups for internet-facing and internal AI workloads, with AI-specific Azure Policy applied to each. An assessment should check the estate against this guidance, so that the plan it produces is one the security team can approve.
Talastron scores six areas, so that each gap comes with its place in the order of work:
- Infrastructure: the Azure estate, identity, network and landing zones.
- Data: the data platform, its governance and its quality.
- AI and agents: the AI already in use, and the knowledge it would draw on.
- Business value: use cases scored on value and feasibility, with one chosen to prove first.
- Alignment with Microsoft’s frameworks: the estate against the Cloud Adoption Framework and the Well-Architected Framework.
- Risk, compliance and governance: security posture, data boundaries and AI policy.
What you should own at the end
Talastron’s assessment leaves seven artefacts with the client. Whichever provider you choose, ask for the equivalent of each:
- a current-state audit of the Azure estate, identity, network, data governance and security posture;
- up to five use cases scored on value, feasibility and risk, with one named candidate to prove first;
- a target architecture, including where the AI runs, how it is isolated, and whether each use case is built in Copilot Studio or in code;
- an AI governance and data-boundary baseline: a usage policy, an AI register and a data classification map;
- a run-cost model, with the forecast cost of model usage at pilot and at scale;
- a phased roadmap covering 90 days, six months and twelve months;
- a board readout with a recommendation and a budget request.
How long it takes
Talastron runs the assessment over four weeks, after a short mobilisation:
- Week 1: interviews with the CIO, CISO, data owner, enterprise architect and two or three business leads, and a scan of the Azure estate.
- Week 2: a use-case workshop, the governance baseline and the data-boundary map.
- Week 3: the target architecture and the run-cost model.
- Week 4: the roadmap, the risk register and the board readout.
It needs read-only access to the Azure tenant, and the estate is left as it is during the assessment.
How it should be priced
In Talastron’s view, an assessment with defined deliverables should be sold at a fixed fee, set by its scope: the size of the estate, the number of use cases and the regulation that applies.
Questions to ask a provider
- Who will do the work, and are they the people presenting the proposal?
- Which deliverables will we own, and in what form?
- Will the assessment end with a decision on what to build first, and its cost to run?
- What access does it need to our tenant, and is that access read-only?
- Is the fee fixed?
About Talastron’s Readiness & Blueprint
Every Talastron engagement starts with a four-week, fixed-fee Readiness & Blueprint, led by the Chief AI Architect with the Chief AI Strategist. To discuss one, book a consultation.
Sources
Microsoft Learn, AI strategy and AI Ready, Cloud Adoption Framework for Azure.
