Hands-on courses taught from live programmes, online or at your offices, and shaped around your people and your platforms.
New · Packt · 2026Build Production-Ready AI Agents with the Microsoft StackGround, secure and evaluate agents with Copilot Studio, Foundry and Agent Framework.
Second edition · Packt · 2025Power Platform Solutions Architect's HandbookExtend Power Platform with Azure, AI and Copilot, updated for today's Dataverse and ALM.
Where I test the latest technologies and techniques, pushing the boundaries of frontier models, before they reach a client programme.
New · Packt · 2026Build Production-Ready AI Agents with the Microsoft StackGround, secure and evaluate agents with Copilot Studio, Foundry and Agent Framework.
Second edition · Packt · 2025Power Platform Solutions Architect's HandbookExtend Power Platform with Azure, AI and Copilot, updated for today's Dataverse and ALM.
First edition · Packt · 2022Power Platform Solution Architect's HandbookThe end-to-end reference for architecting enterprise Power Platform solutions.
A six-day, hands-on course that trains your developers, consultants and IT engineers to take a legacy system all the way to a live replacement. AI agents do the heavy work, and your people approve every step.
Every lesson comes from fifteen years of high-stakes, large-scale enterprise delivery for Microsoft, its partners and their clients.
6 days, or 3 condensed · Up to 12 attendees · Claude Code and GitHub Copilot
Created and taught by Hugo Herrera, Microsoft Certified: Agentic AI Business Solutions Architect Expert
6 to 8 people
1 or 2 people, with agents doing the routine
9 to 12 months
Live months sooner
Spot checks, by hand
Every record, every column
One-off load: days of outage
Delta load: hours
Leaves with the partner
Stays in your team
Each approach comes from Hugo’s own programmes, refined over fifteen years of high-stakes, large-scale enterprise delivery for Microsoft, its partners and their clients. Your team practises them on Lunvale Bank, a fictional bank with a legacy CRM 4.0 system, taking it all the way to Dynamics 365 Customer Service.
Before an AI agent touches anything, your team builds a tool that anonymises the data inside production. A rule is set for every column, people try to re-identify the result, and only then does an agent connect: to the safe copy, read-only.
It’s the step that lets your security and data protection teams say yes.
Agents run read-only queries against the safe copy and turn the answers into a visual report for business owners: what’s used, what’s growing, and what stopped years ago.
Decisions rest on the system’s own data, not on memory.
Agents read the code, the scheduled jobs and the connection strings, then compare them with what really runs, from the logs. The differences are the findings.
Nothing connected to the old system is discovered on go-live weekend.
Your team weighs the options, agrees a blueprint that uses standard features first, and turns it into exact instructions. The agent proposes each change, and a person reads and approves it before it’s built.
Nothing changes in the new system until someone has read the change, and every change is kept in Git.
An agent extends a delta migration tool: the data moves weeks before go-live, then only the changes, every few minutes. Every row and column is compared with the source.
Go-live moves hours of changes, not years of history, and the proof is on screen.
Deployments run through approved pipelines, with a person signing off each environment. After go-live the old system is switched off step by step, with evidence at each step.
The old system is retired with proof that nothing still depends on it.
A visual report, from the system’s own data, of what’s used, what’s growing and what can be retired.
Found in the code, confirmed in the logs, and each one given a plan: rebuild, replace or retire.
Dynamics 365 Customer Service built in approved stages, with less custom code and a test for every process you keep.
Every row and every column compared with the old system, with a control panel that shows the proof.
A rehearsed cutover, then evidence at each step that nothing still depends on the old system.
Your people keep the method, the tools they built and the confidence to do it again on the next system.
Large IT projects often run over and deliver less than planned, and legacy systems carry years of features nobody uses. The method starts from how the system is really used and proves every step before the next.
over budget and 7% over time, on average, for large IT projects1
less value than predicted from the same projects1
of features in the average software product are rarely or never used2
CRM 4.0 left Microsoft support. CRM 2016 followed in January 20263
Every lesson in the course was learned on live programmes: fifteen years of delivering high-stakes, large-scale enterprise solutions for Microsoft, its partners and their clients. Hugo is an independent architect for Microsoft AI and Dynamics 365 and the author of three Packt books.
Tell me about your team and your systems, and I’ll come back with dates and a proposal.