A few days before this conversation was recorded, Himanshu Gupta announced that ClimateAi, the company he had spent eight years building, would close and return unused capital to its investors. ClimateAi helped food and agriculture companies make their supply chains more resilient to extreme weather. It worked with more than 70 of the world’s top 200 food, beverage and agriculture companies, reached over 10 million farmers, and was named one of TIME’s Best Inventions of 2022.

In this episode of Made to DO, Florian Hoffmann sits down with Himanshu in San Francisco’s Presidio. Himanshu grew up in a small town in north India, was the first generation in his family to go to college, and saw a computer for the first time in 2006. Ten years later he was running an AI company. Along the way he became the youngest person to draft the renewable energy chapter of India’s National Five-Year Plan, and he worked with Al Gore and Lord Nicholas Stern on climate economics.

This isn’t a post-mortem. It’s a conversation about where resilience comes from, why Himanshu plans his life in ten-year blocks, what worries him about who benefits from the AI boom, and why he’s now learning by building, starting with a small robot to help his toddler son learn to speak. His conviction runs through all of it: people aren’t left behind by technology. They’re left behind by access to information, access to means, and values.

Key takeaways

1. Plan in ten-year blocks, not quarters. Himanshu spends the first three or four years of each block learning, the next two turning that learning into impact, and the last four compounding it. Short-term wins can happen, but they’re luck, and luck doesn’t repeat.
2. Failure is a necessity, not an outcome. Closing ClimateAi was a deliberate choice. The company had helped create a new category, he had spent his energy building it, and he wanted to give the next few years to his son.
3. Resilience is built early, through exposure and action. His unpopular opinion is that resilience is mostly formed before adulthood. His came from his grandfather, a street vendor who supported a family of 14, and from small tests like being sent to open a bank account alone at age 12.
4. AI’s gains need to reach more people. When value flows to fewer and fewer companies, it stops trickling down to small businesses like his grandfather’s. Himanshu argues for a serious conversation about where AI investment goes, and about the values young people are being rewarded for.
5. Learn by building. Become the mechanic before the engineer. Buy a cheap robot kit, use open-source models, and learn from every debugging session. Each failure compounds into understanding.

Video chapters

0:00 — Opening
1:08 — Meet Himanshu Gupta
2:33 — Intentional or Lucky?
3:43 — Growing Up in a Small Town in North India
6:41 — The Ten-Year Rule
7:58 — Why He Closed ClimateAi
10:59 — The Confidence to Start Again
12:41 — What Real Resilience Looks Like
17:43 — Raising a Resilient Child
19:41 — AI, Wealth and Who Gets Left Out
24:23 — From Changing the World to Becoming a Billionaire
26:24 — Who Can Shift the Incentives?
30:33 — Learning by Building
32:58 — How to Get Started
34:47 — The Next Ten Years
37:47 — What Should We Do More Of, and Less Of?
39:14 — Closing

When Navrina Singh founded Credo AI six and a half years ago, "AI governance" wasn't a category. It wasn't a buzzword. It barely existed as a conversation. Today, it's one of the most urgent topics on the agenda of every board, every government, and every regulator trying to keep pace with a technology that reinvents itself every few months.

In this episode of Made to DO, Florian Hoffmann sits down with Navrina — founder and CEO of Credo AI, advisor to the White House, OECD, and UN on AI policy, and one of Time's 100 Most Influential People in AI — to talk about what it actually takes to make AI safe, compliant, and human-centered. Not in the abstract. In the messy, high-stakes reality of Fortune 500 companies deploying third-party models, regulators racing to write rules for systems that change faster than legislation can move, and leaders who know they need to act but don't know where to start.

What emerges is a conversation that refuses to pick a side between innovation and caution. Navrina's core thesis isn't about what could go wrong with AI — it's about what would go right when we get this technology right. That optimism, paired with a deeply practical understanding of how AI systems are built, deployed, and governed, makes this episode less of a debate and more of a playbook.

Key takeaways

1. AI governance is an enabler, not a hindrance. Companies that build a foundation of risk and compliance adopt AI faster, not slower — because they can evaluate, procure, and deploy new systems with eyes wide open.
2. Context matters more than capability. The guardrails for a marketing flyer generator and a claims-processing system should be drastically different. Governance has to happen at the use-case level, not the model level.
3. Leaders need to go back to building roots. You can't govern what you don't understand. Navrina's advice to CEOs and board members: build an AI system yourself before you try to set policy for one.
4. AI literacy is the bridge between public and private sector. Policymakers who have never used an LLM and builders who have never navigated a regulatory regime both need to meet in the middle. Adaptive policymaking — flexible enough to evolve with the technology — is the way forward.
5. The 10–15 age window is critical for girls in tech. Navrina encourages her 12-year-old to use AI tools freely (social media is off-limits). The confidence gained from building at this age shapes whether women show up at the tables where AI decisions get made.

Video chapters

0:00 — Opening

0:45 — Meet Navrina Singh

2:10 — What Is AI Governance?

5:06 — Governance Across the AI Lifecycle

6:07 — Building a New Category

7:26 — Keeping Up with Rapidly Evolving Models

9:13 — Bridging Tech and Policy

13:04 — What Separates Good AI Leadership from Bad

15:36 — Existential Risk: Responding to Harari

17:13 — Governing AGI

18:11 — Who Should Govern AI?

21:54 — AI Literacy and Adaptive Policymaking

24:25 — First Steps for Leaders

27:23 — Raising Girls in the AI Age

30:37 — Women in AI: Funding, Investing, Celebrating

33:20 — Final Thoughts: More Human Connections

Bet on the Humans: Laurie Richardson on Trust, Safety and Leading Through AI

Every day, someone has to decide what "safe" looks like for the technology billions of people use. At Google, that person is Laurie Richardson. As Vice President of Trust and Safety, she leads the global team that writes and enforces the policies covering Search, Ads, Play, Cloud and Google's generative AI products. Her work spans fraud and scams, child safety and the new harms that AI agents bring with them.

In this episode of Made to DO, Florian Hoffmann sits down with Laurie in San Francisco's Presidio. Her path started a long way from Silicon Valley: she was a lawyer, then chief of staff at the US Treasury, where she worked through the aftermath of the financial crisis and Brexit. Later she helped write Google's original AI principles. That mix gives her an unusual view of how good decisions get made when there's no perfect option on the table.

The conversation covers a lot of ground: staged product launches, regulation, demo slams, entrepreneurship and the redwoods. Through all of it, Laurie keeps coming back to one belief. However much AI changes the world, she bets on people's ability to adapt.

Key takeaways

1. Most big decisions aren't binary. Laurie separates what her team knows from what it hasn't been able to test, then uses staged rollouts and provisional calls she can adjust instead of a single yes-or-no launch.
2. Doing right by users tends to keep you ahead of regulation. When the Digital Services Act and the AI Act arrived, much of what they asked for matched what her team was already doing. Mostly they had to formalise it.
3. You can't plan for every future, but you can prepare for it. Critical thinking, a beginner's mindset and the courage to challenge each other matter more than any single plan. So do structures like her "strategic command" team, which takes charge when a crisis breaks.
4. Measure impact, not adoption. Her team dropped "AI touch rate" as a metric because using AI isn't the same as transforming the work. Competitive, playful demo slams have done more for real progress than mandates have.
5. Leaders need to use the technology themselves. Laurie's advice to CEOs, ministers and board members is to start with something small, like an agent that manages your inbox, and build literacy before setting direction.

Video chapters

0:00 — Opening
0:25 — Meet Laurie Richardson
1:45 — What Government Teaches Silicon Valley
4:50 — What Silicon Valley Teaches Government
7:00 — Deciding Without Perfect Information
9:10 — Regulation, the DSA and the AI Act
11:55 — Preparing Teams for an Unpredictable Future
14:05 — How AI Is Changing Trust and Safety
16:00 — The Human in the Loop
17:55 — Lifting Teams Up: Metrics and Demo Slams
22:10 — Overhyped or Underestimated?
25:15 — Skills, Opportunity and Entrepreneurship
28:10 — Questions Every Leader Should Ask
30:30 — Staying Grounded in Turbulent Times
31:40 — Closing