AI Strategy & Readiness

Most AI projects don't fail on the model. They fail on the data that wasn't ready, the workflow nobody mapped, or the guardrail nobody specified until the demo went wrong. Our assessment exists to find those problems before you spend on a build.

It is run by the engineers who would build the system, not a separate advisory team. Everything in the report is something we would be willing to implement.

Who it's for

• Companies with a real operation (sales, support, operations, compliance, engineering) and a hunch that AI could take work off it
• Product teams deciding whether to add an AI assistant, an agent or an MCP server to what they already sell
• Leaders who have seen a prototype and need to know what it would take to run it safely in production

What you get

A one-to-two-week engagement with five deliverables:

1. Opportunity map. Your workflows, ranked by how much an agent or model could take on, how costly a wrong answer is, and what data each one needs.
2. Data and guardrail requirements. What has to be true of your data, access control and review process before an AI system can act on it, including privacy (PIPEDA) and audit considerations.
3. Architecture options. Two or three ways to build the first system, with the trade-offs spelled out: Azure OpenAI or another provider, agent framework, retrieval design, hosting, and the monthly running cost of each.
4. A prioritised build plan. What to build first, what to defer, and what not to build at all, with estimates we will stand behind.
5. A prototype scope. A two-to-six-week prototype on your real data, specified closely enough that you could hand it to any team.

How it's done

• Week 1. Interviews with the people doing the work, a review of the systems and data involved, and a look at anything you have already tried.
• Week 2. We draft the architecture options, test the riskiest assumption (usually retrieval quality or data access) with a small experiment, and write the plan.
• Readout. A working session with your team, then the written report.

What we bring

• Hands-on AI engineering: agents, MCP servers, retrieval systems and applied machine learning, on Microsoft Agent Framework, LangChain, Azure OpenAI and OpenAI-compatible models
• Twenty years of production systems where downtime and data loss were not acceptable: banking, federal defence R&D, online travel, lending
• Current research: our founder publishes comparisons of MCP tooling and AI host requirements, so the plan reflects how the ecosystem works today, not last year

See how this plays out in practice in MayAI, a consumer agent taken from a chat message to a working prototype, and in our other case studies.