Bruno Caldas has built dashboards on a laptop in the African jungle, trained 300 people on SAP, and is now building AI agent teams for one of the world's largest mining companies. The throughline across 18 years in the industry is always the same: the data foundation comes first.

Bruno Caldas is a project controls and business intelligence professional at Rio Tinto, sitting at the intersection of capital projects, finance, systems, and AI. With 16 years in mining across Guinea, Brazil, and Canada - and roles spanning site engineering, FP&A, and PMO leadership at Vale and Rio Tinto - he has built the reporting infrastructure behind some of the industry's most complex billion-dollar portfolios. He also teaches business intelligence in Northwestern University's MBA program.

Bruno Caldas is a project controls and business intelligence professional at Rio Tinto with 18 years of experience, 16 in mining. Having built PMOs for iron ore, copper, and nickel portfolios at Vale and Rio Tinto - across Africa, Brazil, and Canada - he now sits at the frontier of AI-driven portfolio management. He teaches BI in Northwestern University's MBA program.

https://www.linkedin.com/in/brunodcaldas/

1. Start with your most mature data source - and earn every layer from there

Every PMO Bruno built began the same way: find the data that's already clean (usually SAP), build dashboards people actually want to see, and use that success to earn sponsors for the next layer. He didn't start with governance frameworks or org design. He started with what worked and let the wins spread organically. His rule: prove value with what's mature, then go to the piece that needs fixing. Bottom-up, always - even when top-down funding is available.

2. Two reports, not one: the 3-page executive version changes everything

On every major project Bruno worked, there were two deliverables: a 200-page full report and a 3-page visual executive summary. Leadership wanted fast, clear, and visual. That discipline - ruthlessly separating what the full team needs from what the boardroom needs - built trust with executives faster than any comprehensive report would have. The 3-page version was how Bruno got sponsors. The 200-page version was how the project stayed on track.

3. Mining operations are five years ahead of corporate on AI - and the math explains why

In iron ore, cutting unit cost by just $1 generates $400 million in EBITDA. That return on technology is why operations have been running AI teams, autonomous trucks, and sensor-driven predictive maintenance for years. The S11D project Bruno worked on is entirely truckless - conveyor belts run directly from the pit to the processing plant, no diesel, no CO2. Corporate PMOs are a year or two behind. The gap is real, and closing it is where the opportunity sits.

4. AI agents are only as useful as the data beneath them

Bruno is building a team of AI agents - one for risk, one for HSE, one for community relations - on a medallion architecture: raw data piped into a processed layer for agents, then a gold layer for Power BI dashboards. The Microsoft CEO said it plainly: your AI strategy and your data strategy are the same strategy. Bruno's been living that for a decade. His corporate training program teaches everyone up to the markdown level. The agent-building stays with the people who want to go further.

• 00:00 — Introduction
• 02:00 — Building computers at 14 to mechatronics engineering: Bruno's tech-first origin story
• 08:00 — The Simandou Project, Guinea: $20B iron ore, a railroad through 50% of a country
• 15:00 — S11D, North Brazil: drone photography, walking 1,500km of railroad, performance at scale
• 19:00 — FP&A at Vale: seeing billion-dollar projects through the finance lens
• 26:00 — Power BI, SAP, and the CFO meeting that moved Bruno to Canada
• 35:00 — Building a PMO from scratch: start with what's mature, earn sponsors, scale up
• 44:00 — Operations vs. corporate: why mining AI is years ahead - and the truckless mine
• 55:00 — AI agents, medallion architecture, and the one-stop-shop PMO control tower
• 1:00:00 — Teaching AI inside the company: level 1 chatbots to level 8 VS Code
• 1:10:00 — What PMO teams look like in five years - and why data is the only moat

• Primavera P6 / MS Project (scheduling tools)
• Power BI - Microsoft (launched 2017)
• SAP (PS, FM, MM modules)
• S11D - Vale truckless iron ore mine, North Brazil
• Simandou Project - Guinea, West Africa
• Medallion architecture (data engineering framework)
• Northwestern University MBA program (Bruno teaches BI)