Tax Reform Delivery Model
Brazil's largest tax reform had no ready-made delivery model. I helped build one with AI inside the workflow.
Context & Constraints
Brazil's Tax Reform is the largest in the country's history. Inside KPMG Tax Transformation, the team had to serve a massive, high-criticality demand without a specific methodology ready for this new kind of engagement.
Running client workshops was already hard. The reform added new stakeholders, new information flows and a new AI layer that did not exist in the old workflow. Materials arrived through email, chat, Teams, Meet transcripts, Miro boards and individual notes. The team needed one delivery model that could turn that mess into a validated action plan.
- No ready-made methodology existed for this reform. The team had to design the delivery model while demand was already high.
- Inputs came from many places: email, chat, Teams, Meet transcripts, Miro boards, stakeholder forms and individual notes.
- The Tax team delivered the validated mapping and action plan. Implementation moved to another practice, so the model needed a clear boundary.
“The old model was not broken. The problem was that Brazil had changed the tax system underneath it.”
Architecture
The delivery model had six phases. Business Context mapped the client's areas, stakeholders and initial material. Kickoff aligned objectives, teams and responsibilities. Pre-workshop collected structured information from mapped stakeholders. Workshop sessions mapped the work with the client. The Operational Impact Matrix turned activities, transcripts, notes and Miro outputs into a shared working artifact. The Action Plan / Executive Summary closed the Tax Transformation engagement and prepared the handoff to implementation teams. AI agents entered where information was hardest to consolidate: gathering materials for Business Context, reading forms before workshops, updating working spreadsheets, and drafting the Operational Impact Matrix from workshop evidence.
Decisions & Trade-offs
Start with Business Context
Map stakeholders, areas and scattered materials before the first client workshop.
Considered: Start directly with workshops · Collect context informally during kickoff
The workshops only worked when consultants arrived with enough context to be proactive. Business Context gave the team a common map before the client sessions began.
Put agents where information piled up
Use AI agents for emails, chat, Teams and Meet transcripts, Miro outputs, stakeholder forms and working spreadsheets.
Considered: One generic assistant for the whole project · Manual consolidation by each consultant
The hard part was not generating text. It was joining many sources into a working artifact the team could validate with the client.
Keep the delivery boundary clear
Tax Transformation delivered the impact matrix, action plan and executive summary. Implementation moved to another area.
Considered: Extend the same model into implementation
A clear boundary kept the methodology honest. The Tax team owned diagnosis and action planning; implementation required a different delivery practice.
- The AI layer had to work inside KPMG delivery artifacts, not as a separate tool consultants would forget to use.
- Coordinating areas and stakeholders was slower than building prompts, but it was the only way to make the model fit real delivery.
My Role
I translated the consulting methodology into an AI-enabled delivery model, designed the agents and templates, and wrote execution materials with the team.
Results & Validation
The model kept the consultants in control. AI did not replace the Tax work; it organized the information around it. Tax Transformation leadership later presented 30-60% efficiency gains on covered activities, and the team kept using the model across reform projects.
Stack
Building AI that has to work every time?
I design, ship and operate systems like this one.
Let's talk