Build AI agents that already know your business.

Yes, core LDI membership is free, with no time limit and no hidden upsell to unlock the essentials. I'd simply rather have practitioners in the room who genuinely want to build than chase sign-ups.

Yes. There's a dedicated path for non-technical leaders, CFOs and senior commercial roles, who need to understand and commission Self-Service AI without writing a line of code, alongside deeper paths for finance analysts (starting from Excel and Power BI) and data practitioners. All three begin from first principles, so you join at your level.

This is the heart of why Self-Service AI is built on the Microsoft ecosystem. Your data never leaves your own environment: it stays in your organisation's Microsoft tenant and Fabric capacity, in the region you choose, with multi-geo support, so different datasets can be held in different regions to meet local requirements.

Access is governed end to end through Microsoft Entra (identity, conditional access and permissions), so an agent only ever sees what the person using it is allowed to see. You're not shipping your finance data off to a third-party AI provider, the intelligence comes to your data, inside your tenant, under the security and compliance controls you already run.

Microsoft Fabric is needed to run the LDI agent builder itself, though for workshops and demos I can set you up on a temporary account in my tenant. Each course also states exactly which tools and licences it needs, so you always know before you start.

It depends how much context you've built into your agent builder before you start. The quarterly build workshop is designed to get you to a working example in a single day, something your team can actually use, not a slide deck about what you might build. If your diary genuinely has no room for that, it's worth asking what's filling it, since it's often the exact kind of recurring manual work a day like this is meant to claw back.

Consultants build things for you; LDI teaches your team to build for themselves. The quarterly workshop is structured so your team owns every part of the agent they build, the data connections, the logic, the governance rules. The goal is a capability your business controls, not a dependency. Nothing here waits on a vendor or a tooling sign-off either, since the skill is judging where a model's output can be trusted and where a person has to check it, which your organisation needs someone to own regardless of what eventually gets procured.

Yes, and it isn't a you problem. Independent 2026 research puts M365 Copilot adoption below 4 in 10 licensed users, and around 3% of the wider M365 base. The tool isn't the bottleneck, the missing layer is the business context that turns a general assistant into something that knows your chart of accounts and your definitions. That's what the five-layer brain above is for.

Deciding what an AI model can actually see before you type anything to it, the process handbook, the approved definitions, the tools it's allowed to call, and where a human has to sign off. It's different from prompt engineering, which is just wording a single request well. Most agent failures trace back to missing context, not a weak prompt, which is why it's become the core skill in the Finance Forward Deployed Engineer pathway.

Sits inside the business, not in a separate technical function, and turns a real process straight into a working, governed skill. It's one of the fastest-growing roles in tech right now, generally requiring data engineering plus enough domain knowledge to translate a vague business need into something an agent can run. The Finance FDE pathway teaches exactly that, starting from a finance background rather than a computer science one.