Why Personal Finance Is a Self-Driving Problem | Ambrish Tyagi of Rivo

From TwoSetAI: Decoding AI Builders · 53:03

Show by Angelina Yang

Ambrish Tyagi grew up in Delhi in a multi-generational, middle-income household and moved to the US in 2002 to start a PhD in AI at Ohio State. In the show's intro, Angelina says he spent eight years building AI at Amazon and helped put the first commercial robotaxi on San Francisco streets. He mentions the self-driving stack he worked on at Cruise. Today he is the founder and CEO of Rivo, a money app that moves idle cash out of your checking account and back in time for your bills. Ambrish says Rivo started building last year in stealth, announced a 2.7 million seed round earlier this year, and then opened to customers.In this episode, Ambrish joins TwoSetAI to explain the "inertia tax": the money people lose by leaving cash idle in checking accounts because they are too busy to move it. He traces the idea to his own story. After the pandemic he had hundreds of thousands of dollars sitting in cash with a financial advisor for four or five months while the federal government was paying 5% or 6%. When he asked about it, the advisor said he had not been told what to do. That was the moment he asked, "What am I paying you for?" His view is that the biggest problem in personal finance is not bad decisions but the effort of making and executing them every day.Try Rivo at rivofi.com.We go deep on how Rivo works and why Ambrish calls it a self-driving stack for finance. Users connect a bank account through Plaid, and Rivo reads bill history and liability data through a partner called Method. It sweeps idle cash into short-term US Treasuries and brings it back a day or two before a bill is due. Autopilot handles both directions and suits balances of $5,000 or more. Copilot starts at $200 and sweeps anything above a threshold you set, but you bring the money back yourself. Ambrish walks through the edge cases, such as a surprise first payment on a new credit card, and how Rivo generates synthetic near misses in the way self-driving teams do, with accounting rules as guardrails. He lays out five levels of money autonomy modeled on self-driving cars, and says Autopilot is an L4 product. He also covers the bank partner, Jiko, a federally chartered bank and broker-dealer, and why he avoided middlemen after the Synapse collapse. Rivo charges five basis points per month on the average balance it optimizes, with no subscription.We also get into the harder parts. Ambrish says error rates will never be zero, that Rivo has not published its rate, that today's level is under about one overdraft per person per year, and that he wants it 5X to 10X better. Angelina asks where a human stays in the loop. His answer is none: he calls most of today's agents "glorified reminders" and says he does not want to build an L2 system for finance. Asked about LLM mistakes, he says Rivo does not hand decisions to a model. LLMs supply signals and act as a veto, while layered checks work like slices of Swiss cheese. He also takes on whether a vibe coder could rebuild this with Claude, and why advice may get commoditized. He closes on evals, vibe coding and a book pick.We discuss:* Why $25,000 in a checking account earns about $2 a year, by Ambrish's math* The FOMO screen that shows new users their missed earnings* What counts as idle cash, and bringing money back a day or two before bills* Why Treasury bills, and why bill money is sacred* The inertia tax and how banks profit from it* Autopilot versus Copilot, and the minimum balances for each* Plaid, Method and the data behind predicting bills* Edge cases in personal finance and in self-driving cars* Synthetic near misses before an overdraft happens* Accounting rules as guardrails: you can't create or evaporate money* What happens when the system gets it wrong, and a goal of 5X to 10X fewer overdrafts than today* Why he will not keep a human in the loop* Why most agents are still "glorified reminders"* The five levels of money autonomy, L1 to L5* Beyond deposits: lending, refinancing and the 2030 vision* Finding a bank partner, and the Synapse lesson* SIPC and FDIC protection if Rivo disappeared* LLMs as the veto, not the decider, and the Swiss cheese model of safety* Could you vibe-code this with Claude yourself?* Growing up in Delhi, and the advisor who prompted Rivo* Why start with idle cash: the Waymo trust test and the cash buffer slider* Pricing for aligned incentives: five basis points per month* Where the first customers came from, including Reddit and Bogleheads* Vibe coding, specs and historical-simulation evals* A book pick: Traction by Gabriel WeinbergAmbrish Tyagi* LinkedIn: https://www.linkedin.com/in/ambrishtyagi/* Rivo: https://rivofi.com* Book Ambrish recommends in the episode: Traction by Gabriel WeinbergTimestamps00:00:00 The advisor who did nothing with his cash00:01:17 From robotaxis at Cruise to automating your money00:01:49 The seed round and what Rivo actually does00:02:50 The inertia tax: $6 trillion sitting idle00:04:00 Autopilot vs Copilot, and the minimums00:04:55…

More episodes of TwoSetAI: Decoding AI Builders

More episodes

PodSearch is a free, independent podcast search engine. Read about PodSearch, the privacy policy or contact us.