Click here to watch a video of this episode. How banks make money came down, last year, to two things: rates went up and customers left their money where it was. That earned a record $1.3 trillion, and the market still values banks nearly 70% below every other industry. McKinsey counts about $23 trillion of consumer deposits sitting in accounts that pay nothing. If software moves even 5 to 10 per cent of it to the best rate on offer, a fifth of the industry's deposit profit goes with it. That is how banks make money today, and it is the part an AI agent has no reason to respect. Aman Narain and Zubin Vandrevala break down the future of banking in four acts, nine shifts and one flywheel, and answer the question the Future Of series has been building to: if how banks make money depends on inertia, what is a bank for when software does the shopping? Key takeaways: 1. Banks earned a record $1.3 trillion in 2025 and are still valued nearly 70% below other industries, because the market is pricing the day customer inertia ends. 2. Of the three legacies holding banks back, infrastructure is the one you can rent and mindset is the one you cannot procure. 3. Money that is portable, private, cheap and programmable carries data as well as value, and the next revenue line is a fee on that data. 4. The threat to banks is not that deposits leave for stablecoins. It is that the deposit stops being free. 5. In a life run by software agents, the bank's job shifts from serving you to standing guard: the last regulated name that can say no. Questions answered: - How do banks make money? In 2025, largely from higher interest rates on deposits that customers never moved. - How is AI changing the way banks handle your money? Software agents will move idle cash to the best rate and pay on your behalf, which leaves the bank as the regulated party that vouches for the agent and can refuse a payment. - When software does the shopping, who holds the money? A bank still does, but the value lies in who made the promise and whether they will say no on your behalf. Enjoying A2Z Fintech? Leave a rating and review on Apple Podcasts. It is the single biggest signal to the Apple algorithm and how new listeners in our world find us. Topics covered: - How banks make money in 2025: a record $1.3 trillion, priced nearly 70% lower - The banking hall, the pen on a chain, and John Reed's 1976 Memo from the Beach - Six refits in 300 years, from the Medici branch in Bruges to embedded finance, and a menu that never changed - The three legacies: business model, infrastructure and mindset, with Wayne and his COBOL - Nubank, Revolut and DBS as evidence that a bank can be built, or rebuilt, without them - A scorecard of four confident predictions about banking, and why each assumed the customer would do the work - The four Ps of money: portability, privacy, price and programmability - MINT: machine-made, instant, non-stop and tokenised, including the NYSE's planned 24/7 venue - Stablecoins against tokenised deposits, and the Deposit Concession - Five verbs, three I's, four rungs of data, and the household as the unit of banking - Cloned voices, the Arup deepfake, harvest now and decrypt later, and know your agent - The tenth idea: a flywheel with purpose and people at the hub - Two dated bets for 31 December 2027 Chapters: (00:00) - Cold open: Layla Haddad, Dubai, 2032 (00:48) - The banking hall and the pen on a chain (01:13) - John Reed's Memo from the Beach, 1976 (02:07) - Singapore to San Francisco: part three (03:13) - Disclaimer (03:49) - Shift 1: from a building to the moment (04:49) - Enfield 1967, a blizzard and Citi's $100 million bet (06:30) - From the phone in your pocket to embedded finance (07:26) - Six refits, one menu: the Medici in Bruges (09:40) - How banks make money: $1.3 trillion, priced 70% lower (10:57) - Shift 2: the three legacies and Wayne's COBOL (13:38) - Nubank, Revolut and DBS (14:21) - Scorecard: four predictions that missed (16:48) - Shift 3: the four Ps and money that carries data (19:15) - Programmable money in three steps (21:13) - New interchange: a fee on the data (22:35) - Shift 4: MINT, the kitchen that never closes (24:53) - Why a Friday payment clears on Tuesday (25:37) - Wall Street's hours and the NYSE's 24/7 venue (27:09) - Tokenised: the $200,000 velvet rope (28:24) - The Deposit Concession: stablecoin or deposit? (30:34) - Shifts 5 and 6: five verbs, three I's (32:54) - Shift 7: from hindsight to foresight (33:45) - Shift 8: from the account to the household (35:08) - Shift 9: the $23 trillion nap (36:36) - Doreen, Arup and fraud at machine speed (38:19) - Harvest now, decrypt later, and three defences (40:56) - Know your agent: compliance as a profit centre (42:47) - The tenth idea: from programmes to a flywheel (46:13) - The doom loop, and where to start (47:44) - What to do on Monday (48:22) - Closing wagers: weekend trading or a $1,000 fund? (49:15) - Keep the chain, give us back the pen Referenced in this episode: McKinsey Global Banking Annual Review, 2025 and 2026 editions; John Reed's "Memo from the Beach" (Citibank, 1976); Barclays' first cash machine, Enfield, 1967; Nubank second-quarter 2026 results; Revolut 2025 annual results; DBS on the value of AI in 2025; the NYSE tokenised securities venue; The Clearing House tokenised deposit network; Standard Chartered's stablecoin forecast; Hamilton Lane and Franklin Templeton tokenised funds; the UK's fraud reimbursement rules; the Arup deepfake fraud, Hong Kong, 2024; NIST's post-quantum timeline; Mastercard Agent Pay. Corrections: At 38:04 we said the Arup deepfake fraud was about $25 billion. It was about $25 million (HK$200 million). Related episodes: The Financial Invisible (Future of Work, part one); S2E32, The Future of Shopping (part two); S2E28, Revolut. Hosted by: Creators & Guests Aman Narain - Host Zubin Vandrevala - Host Aman Narain writes at amanwhoblogs.substack.com. Zubin Vandrevala is your payments provocateur. For information and entertainment only. Not financial advice. Transcript: Click here to view the episode transcript.
Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
Zubin: Aman, banks have never made more money than they did last year. So why does the market price them as if they're dying? Aman: The answer, Zubin, starts, if you'd like, in a kitchen in Dubai in 2032. Layla Haddad runs three physiotherapy clinics. While she slept, software she hired moved her spare cash twice to earn a better rate, paid a supplier in Rotterdam in seconds and put hundred dollars into a fund that used to ask for 200,000.
A bank stood behind every one of those moves. She hasn't opened its app since March. Layla is, of course, a composite and is a plausible future, not a fact. I think it's the future the market is pricing. To see how we got here, go back to the banking hall, Zubin, any city you like, any year before 1990. Marble floor, brass rails, a queue, and on the counter, a pen chained to the desk.
You've walked in to hand this institution everything you own and it doesn't trust you with a biro. Then let's go to a beach in St Croix. 1976. A 37-year-old banker called John Reed is meant to be on holiday. Instead, he writes 6 pages to the people who worked for him at Citibank. The bank still calls it the Memo from the Beach. One line in it has followed me around for years. The task is not to run branches, nor to issue cards. Reed told his bankers,
Zubin, the branch was a means, not the end. 50 years on, the bank itself is becoming the means. So hold on to three things: the number behind that valuation, the memo. And the question we left you with last time. When software does the shopping, who holds the money?
Aman: Welcome, welcome back to A2Z Fintech where we hunt for non obvious insights on finance, tech, people and power. I'm Aman in Singapore. Zubin: And I'm the Z Zubin in San Francisco. Aman: This is the last part of our future of series. Part one asked who pays you. Part two asked who you pay. And today, who holds the money? Zubin: And a question for you before we start. What does your current account pay you? Put that number in the comments now. You'll see why I'm asking. Same five verbs as the other two:
see, move, borrow, grow, and protect. In our first episode on the future of work, Divya couldn't get a mortgage on a dashboard. Two weeks ago, we covered in the future of shopping Tamika's phone did the weekly shop, they both end up at the same counter. And I'll tell you now where I'm gonna push. Your note says banking has only ever changed its shop windows. I think that's unfair to about four decades of financial engineering. Aman: Then keep score. Zubin: Man, I always do. Alright, the lawyers' bit's on the screen.
The short version is if an AI agent moves your savings at 3am in the morning for an extra 11 basis points, that's between you and your agent. We weren't there, we're just two dudes riffing. Stay smart. Aman: Okay, four acts today the past, the present, the future, and the rebuild. The 10 ideas along the way. 9 are shifts, each one a from and a to. And the 10th isn't a shift at all. It's the thing that turns the other 9. By the end of this, you'll have all 10 on one card.
So and I'm no Zubin's counting. Zubin: Somebody has to Aman: Yeah. So let's start with shift number one from a building you visit to the moment you're in. I said it at the end of part one that work, shopping, and banking were all built around a place. Banking is the purest of those cases. For 300 years, the only thing it really changed was where it met you. And and it did that in two revolutions, or should we say evolutions really.
Given how long they took. Some place, somewhere, and then something. Let's talk about someplace first. The the hall, the banking hall, went on to become the branch on the high street. You travelled to your money in your bank's hours and you queued for it. Zubin: Yeah, and when you finally reach the front, there it was. The pen on the chain from the top of the show. The chain tells you who the building was designed for. Your time was free. Their biro wasn't. Aman: Hold on to that pen, it comes back.
Zubin: So revolution, or should I say, evolution one gets us out of the queue with a hole in the wall, literally, right? 27th of June 1967, a Barclays branch in Enfield, North London, the world's first cash machine. 10 pounds at a time. Aman: And for 10 years, it's a curiosity, which is where my man on the beach comes in. John Reed is running Citibank's new consumer group. His memo says the business isn't branches or cards,
it's families' financial needs. Then he bets a reported hundred million dollars back then on wiring New York City with wire cash machines. And he insists on two in every branch. Zubin: Why two? Aman: Because machines break, Zubin, and the promise was a 24-hour day capability. His own colleagues thought it was ridiculous. Zubin: Yeah, the oldest rule in payments, right? One is none, two is one. Aman: I learned that in this in the research of this programme and and from you.
I love it. almost rhymes. And and new and then New York shrugs, you know, until February 1978, when a blizzard buries the city and every branch stays shut. The way Citi tells it, the use of the machines jumped. By about a fifth more or less overnight. They shot an advert of people wading through snow drifts to a hole in the wall and gave it a line. The Citi never sleeps. Zubin: Yeah. So so I guess the future of banking arrived because of the weather.
Aman: Well, it usually arrives because of an inconvenience. Zubin: Right, all right, all right, okay. So that's somewhere, right? The cash machines first, then the phone line, the desktop, and from I would say 2007, the thing in your pocket. The banks stopped being a building and became wherever you happen to be. So Revolution 2 or Evolution 2 is the something. You and I both did time at Google Pay. So we watched this one from the inside. It started with payments, Google Pay,
Apple Pay. Aman: Okay. Zubin: Nobody opens an app to do banking, right? You're paying for a coffee, splitting a bill, getting paid for a ride. The bank is inside the things you already are doing. And it's heading from embedded payments to embedded everything. Aman: Yeah, Zubin, my last job at HSBC was exactly this platforms and embedded finance. And and the pitch from clients was always the same. We don't want your portal, we want your balance sheet behind our screen. Zubin: Yep.
Aman: So so here's the non-obvious bit hall, branch, cash machine, website, app, embedded. 6 refits of the dining room in 300 years. And through all of it, the menu never changed. See your money, move it, borrow it, grow it, protect it. A client of the Medici family would recognise every single dish. Picture their branch in Bruges around 1450. A wool merchant checks his balance in a ledger kept by double entry.
See. He buys a bill of exchange so he can pay a supplier in Florence without carrying a coin over the Alps. Move. Zubin: And across the Channel, keeping with that same theme, the London branch is lending to Edward the Fourth in the middle of the Wars of the Roses. Borrow, I guess. And and back in Florence a depositor earns a discretionary gift rather than interest because Aman: Yeah. Zubin: interest was a sin. So that's the grow. Aman: And protect? Edward never repaid Zubin and in 1478 the London branch was wound up.
Zubin: So front of the house gets a refit every generation, and the kitchen is still cooking on the same stove. Aman: Until now. Zubin: Alright, y here's my push. That's too neat and it's unfair to the kitchen. The credit card, the credit score, securitisation, the money market funds, none of that was wallpaper. Those were new products and they moved trillions. Aman: that's fair. The kitchen has had new appliances. Zubin: Thank you, thank you, thank you. Aman: But look at what every one of them was made from.
A deposit, a ledger that lives at a bank, a working day with closing times, a minimum ticket, new recipes, same ingredients, Zubin. What's changing now is the ingredient, the money itself, and what the bank is built out of. Zubin: I'm I'm sorry you went too fast on that one. I want to go back. I'll I'll take the one half that I won, the Kitchen Concession. One Zubin, zero Aman. Aman: Don't get comfortable. Zubin: Well, there's a second one coming and it's about your deposit.
He doesn't know yet. Alright, Act Two, the present. And Aman you owe us the numbers. Aman: Yeah, so here it is. In 2025 the world's banks and get this made a record 1.3 trillion dollars of profit, Zubin. On McKinsey's numbers, no industry on earth earned more. Zubin: So the markets must love them. Aman: The same research has banks valued nearly 70% below the average of other industries. Zubin: Hang on, hang on. The best year in history of the industry,
and it's priced like a business in runoff. Aman: Yeah, because the market can read. Those profits came from interest rates going up and customers leaving their money where it was, not from anything, if you'd like, that the banks built. Zubin: So the market isn't pricing last year, it's pricing the day the customer stops being lazy. Aman: Yeah, that discount is the market's guess at what happens when the inertia goes. And the inertia is the one thing software doesn't have. Zubin: So if that number was news to you,
subscribe now. There's a bigger one coming. Here's what I don't get. Every bank chief executive can read that discount. They've had 20 years and a great deal of money. Why isn't it fixed? Aman: Because Zubin, the problems haven't changed. And this brings us to shift two, from the plumbing to the people. I I've worked inside three big banks, and it's the same problem every single time. I call them the three legacies. It's legacy business models, legacy infrastructure,
and legacy mindset. Let's start with the legacy business model. A bank is organised around its products, each with its own profit and loss account, and not around you, the consumer. The mortgage desk and the cards desk compete for the same customer. The whole arrangement is paid for by deposits that never move. Zubin: Legacy infrastructure, which I think means it's time you met Wayne. Aman: Yeah. Wayne. Wayne wrote his bank's interest engine,
if you remember, from an earlier series, in COBOL. He did it in 1984 and he's now on a golf course chilling in Boca Raton. He's fictional, but every banker watching this knows at least three of him. Zubin: All right, in in one industry survey, 62% of banks early in modernisation says old systems are still their main obstacle. Aman: And here's the non obvious bit. Infrastructure is the least constraining of the three. You can rent a modern core, AI can now read Wayne's code.
It's also the most dangerous of the three in the age of AI. Zubin: Right, because translating COBOL was never the hard part. The hard part is the rule nobody wrote down. Why does this account skip a day's interest in a leap year? The code says it does. Only Wayne knows why. Aman: So the machine can read the recipe, it can't ask the chef. Put software that acts in seconds on top of rules nobody understands, and you get your old mistakes at machine speed.
Zubin: And the third? Aman: Legacy mindset. this is the most constraining by some distance. Since 2008, a generation of bankers, 2008 being the global financial crisis, a generation of bankers have been paid for not getting it wrong. That isn't the same as being paid for getting it right. I've watched banks spend from August to January deciding what to fund. And calling it planning. You you can rent infrastructure, you can redraw a business model,
you can't procure a mindset. Zubin: Yeah, so the one nobody spends on is the least of the three. And the one no nobody's budgets for is the hardest. Aman: That's shift two. The plumbing you can buy. The people you have to change. Zubin: Yeah. And and you can see what a bank looks like without them. Nubank, nearly 139 million customers, its first billion dollar quarter, and about a dollar a month to serve each of them.
Revolut, we've covered this as well, 6 billion dollars of revenue last year, 2.3 billion in profit before taxes, and a full UK banking licence in March. Investors now value it about $115 billion. That's a private price, not a public one, of course. It's still more than most European listed banks are. Aman: Yeah, and an old bank can do it too. DBS says its AI is worth about 1 billion Singapore dollars last year, and it measures that against control groups.
Zubin: Now now a scorecard, as in part two, right? Four confident predictions about banking made by serious people, and how did they do? So let's talk about the first one. The bank branch is finished. Brett King called his Aman: Mm-hmm. Zubin: 2011 book, Branch Today, Gone Tomorrow, and predicted that American branches would halve by 2030. Aman: Zubin, I know Brett and I like him, but a title like that is lit written to sell books, not really to be checked. Yeah. Zubin: Touché, touché, but but verdict half right so far,
right? Britain has closed nearly about what 7,000 branches since 2015. Aman: True. Zubin: More than two-thirds of its networks says Which?, right? And and in America, JPMC is part way through, I guess, opening 500 new ones. So two, the second piece to this is big tech becoming your bank. Google shelved this the bank account in 2021. You and I were there around that time. And then Goldman, of course, Apple's card partner, headed for the exit. Aman: I should declare our interest.
We were both part of that, as you said, at Google Pay for that. I I was sitting in that room in Mountain View where we convinced Citi it was the future. Big tech wants the consumer interaction and the data. It never really wanted the balance sheet or the regulator that comes along with it. Zubin: Right, that's right. So let's get to point three. Aman: Thank you. Zubin: Blockchain fixes trade finance. In 2018, Bain put a number on it: 1.1 trillion dollars of extra world trade by 2026.
Guess what? It's November 2026. The two big bank consortiums, we.trade and Marco Polo, both have folded. The technology works, but 190 legal systems did not. Aman: Again, I I love the folks at Bain, but not the Nostradamus of Fintech or tech for that matter. in in their defence they said could. Zubin: could, could. Noted, noted. All right. Four open banking arrives and everybody shifts, switches banks, Well, the data moved, the customers didn't, So every one of those predictions assumed the customers
would do the work. Aman: And nobody changes banks for fun. Zubin: So why is this time any different? Aman: Okay, because this time the customer doesn't have to do the work, importantly. Software does it for them. And let's hold on to that thought because it's where we finish. But first, what's changing underneath that? Act three, the future. In act one, the dining room got 6 refits and the kitchen never changed. Now, two things changed at once.
ingredient as we talked about money itself and the kitchen. The ingredient, which is money, Zubin, what did do you remember what your economics teacher used to tell you money was? Zubin: I better know four things. A medium of exchange, Aman: Yeah. Zubin: measure of value, a standard for payments you make later. And let me look up my notes. Ooh, a store of value, medium measure, standard store. I get a B Aman: I used to know that media measure standards. store. I I not many subjects where I got lots
of A's, but that was one. but all four of those are true, they've been true since the cowrie shell. And I think my generation or our generation is the one that watches money lose its body. Look, I was born the summer Reed wrote that memo by the beach. I've paid for things with coins, notes, cheques, plastic. And QR codes now, I expect to see the day none of it is physical. Which brings me to shift three from carrying value to carrying data.
It starts with what now sits on top of those four fundamental things your econ professor drove into you. And I like to call them the four Ps. They're portability, privacy, price, and programmability. Portability. Money used to live somewhere. A vault, a branch, a wallet. Now it lives in a phone. M-Pesa in Kenya, UPI in India, Pix, we talked about it in Brazil, and Alipay in China.
And here in Singapore, PayNow, where your phone number is your account. Zubin: Price, which is my world, the cost of moving money is heading for zero. UPI and Pix were built by central banks on purpose to do exactly that. And the rails have gone from national, China has its own, India is exporting UPI, Europe is building a rival to Visa and Mastercard. A payment system used to be plumbing, now it's foreign policy. Aman: That brings us to privacy. Every time you tap your phone, the shop never sees your real card number.
It's the trick you walked us through in part two. More of your money moves and less of you is exposed when it's built properly. Zubin: And programmability which needs the blackboard. Aman: Prof Z, as if I were a smart 12-year-old, please. Zubin: All right, all right, all right, all right, all right. three steps. Today money is a number in your bank's ledger. If you want to do something clever, For example, pay the rent of the first month, say you leave an instruction with the bank. The rule lives at the bank. The money is dumb.
Step two, put the money on a shared ledger as a token, and the rule can travel with it. And then step T, step three, now the money carries conditions. Pay the supplier when the goods are scanned in. Release this deposit when the key changes hands. Or the handyman from part one whose pay splits into tax, Aman: Yes. Zubin: insurance, and profit the moment the job ends. Aman: So a standing order is a rule about money. This is a rule inside the money. Zubin: That's it, Aman, right? Singapore's central bank calls one version
purpose-bound money. A voucher that knows which shops it works in and what day Aman: Okay. Zubin: it expires. It has run trials since we discussed this, right? 2022. India's piloting the same idea on the digital rupee. And the honest worry, Aman, right? Money with rules inside it is wonderful when you're the one writing the rules. YouTube is full of people frightened that it means somebody can switch theirs off, right? the fear is, I would argue, overcooked. The question underneath it is,
I think, a good one. Who writes the rules? And who do you appeal to? Aman: Now put the four Ps together because this is the part I think people miss. Money used to carry one thing, value. Money that's portable, private, cheap, and programmable carries two value and data. What was bought when, on whose instructions, under which rule? That's shift three. Zubin: And the industry has only ever charged for the first. Interchange,
a slice of the value every time a card moves. We by the way, we did a whole episode on it. Aman: And and you've just told us that the slice is heading for zero. So where's the money in the money? I so where's the money in money? I think it's in new interchange, Zubin, a fee on the data that travels with the payment shared along the chain, and for including the small merchant who has only ever paid interchange and never, if you'd like, earned it. Zubin: Hang on, a minute ago privacy was going up,
now you're selling the data. Aman: Privacy is what makes it worth paying for. Data that leaks is free. Data that moves only with your permission has a price. And programmable money can carry that permission. In part two, I said Meta was going from renting attention to taxing intent. banks hold the best record of intent there is. So far, they've either locked it up or giving it away. Zubin: I'll say what our evidence rule requires.
That's your thesis. Nobody has published a rate card for it. Aman: It's the it's a thesis, if as you say. I'd still rather own the revenue line in 2035 than the one it replaces. That's the ingredient. Now, now the kitchen. How banking actually gets made. Shift four from office hours to always on. Now, Zubin, part one, we had WHAM. Zubin: Oh. Aman: Part two, I brought you CART. Zubin: Where are Aman: so it's time for MINT, Zubin. I N T where money is made.
Machine made, instant, nonstop, and tokenised. Zubin: You do know these take you longer than the actual research, but go on. I'm I'm saving something for the last letter. Aman: Okay, M is machine made. The product gets built and the decisions get taken by a model. McKinsey puts the gain from AI and software engineering at 20 to 25% and it reaches the product itself. Zubin: Let's go back to Divya from part one, right? 400,000 subscribers, $180,000 a year,
and a bank that told her it doesn't underwrite dashboards. Your bet was that by the end of 2027, a top-10 bank in Asia would approve home loans on platform earnings, not pay slips. Aman: And and this is how a model reads three years of her platform income and prices the swings. A bank puts its name on the loan and and the money increasingly comes from somewhere else, a private credit fund, if you'd like. Zubin: That last part is real, right? when we were researching this,
McKinsey has banks' share of the world's financial assets slipping from 44% to about 40% in three years. It's also untested. Private credit has never had to work out bad loans at scale through a long downturn. And machine made has a human cost. Both of your old shops have said so, right? Standard Chartered Aman: Yeah. Zubin: has said more than 7,000 corporate function roles go by 2030 as AI comes in. HSBC is reported to have weighed something like 20,000,
but really hasn't confirmed a number. Aman: those are real people, and I know some of them. The question I'd ask, and is something I'm putting to Sunil Setlur in our next episode on work in the age of AI. And and here's the question: if the junior work goes, where does the next generation learn what a wrong number smells like? Zubin: I for instant Aman: Yes, instant. Layla Zubin: Ha ha ha. Aman: Layla pays a supplier in Rotterdam on a Friday night. Why does that take until Tuesday to clear?
Swift's own data says three quarters of cross-border payments reach the receiving bank within 10 minutes. Most of the delay is in the last mile, inside the bank at the other end. So the network isn't slow, the bank is. And Capgemini, Zubin, reckons 4 trillion dollars sits parked in accounts banks hold with each other. Sort of the nostros and vostros. Just so those payments can clear. One source, so you can handle that with care.
N okay. N Zubin is for nonstop. What time, by the way, does the New York Stock Exchange open if if you know? Zubin: 9.30 am and it closes by 4 pm. Your notes said 9 to 5 by the way. Wall Street has never worked that hard. Aman: 6 and a half hours a day, five days a week since 1985. And those of us in Asia, that's like bedtime. Zubin: It was built for people and paper. clerks went home, certificates had to be carried across town, and the part everyone forgets,
the money stopped. The payment system underneath shut at night and at weekends. Here's what changed this year. In January, the New York Stock Exchange said it's building a venue for tokenised shares. 24 hours, 7 days, settling on the spot, funded with stablecoins. Last week it signed up the firms that keep the share registers. It still needs the regulator's approval and it has no launch date, so take this with a pinch of salt, but Nasdaq is going for round the clock trading on weekdays.
And we covered that in a previous episode. Aman: Now the tell is in the small print. The exchange's parent is working with two banks, BNY and Citi, on tokenised deposits. So its members can move money outside of banking hours. Zubin: Right, because a market can't stay open if the money has I guess gone. Aman: A clock with no hands. Zubin: Yeah and it cuts both ways, right? if a treasurer's software can empty accounts in 15 currencies in 10 seconds, a bank run doesn't wait for Monday either.
Aman: So T, the big one, tokenised. Start with the minimum. In Singapore, where I live, a fund can skip the retail rulebook if every investor puts in at least 200,000 Singapore dollars at a time. In America, one test for the private stuff is $200,000 of income. The number on the velvet rope is $200,000. Zubin: And tokenised changes that how? Aman: A token, well, because the minimum was never a law of nature, it was a cost of the paperwork,
essentially. Hamilton Lane offered a tokenised way into one of its private equity funds, and the minimum went from 5 million dollars to 20,000 dollars. Franklin Templeton's tokenised money fund starts at 20 dollars and it was launched here in Singapore with DBS. Zubin: Now the cold water. And by the way, I'm reading this from your own slide deck, right? Slide 14. Tokenisation tokenising an illiquid asset changes the record,
not the asset. Aman: I stand by slide 14. slice of a private loan that trades at two on a Sunday morning is still a private loan. A lower minimum is access. It isn't safety. Zubin: Then let's talk about the cash itself, right? A token of money issued by a company is a stablecoin issued by a bank, it's a tokenised deposit. Aman: And and this is where I get nervous on banks' behalf. If Layla's cash sits in a stablecoin and not in a deposit, her bank has less to lend. That's the agentic bank run.
That's the agentic bank run we flagged in part two as a scenario and not necessarily as a forecast. Zubin: Yeah, no, no, here's where I would push back with numbers, right? dollar stablecoins add up to about what 300 billion dollars. That's where they started the year, right? Your old shop, Standard Chartered, has pushed its $2 trillion forecast back from 2028 to 2030. America's new stablecoin law stopped issuers paying interest. And America's biggest banks plan a shared network for tokenised deposits
in the first half of next year. Aman: Yeah, so I guess the threat isn't the stablecoin, it's the deposit. Zubin: It's that the deposits has to grow, right? Open all night, programmable, and paying a proper rate. Aman: Which costs the bank its margin anyway, but you're right. And and I had it the wrong way around, I guess. The deposit doesn't leave the bank, it stops being free. That one's yours. Zubin: Like I said, I'm still counting. The Deposit Concession, Zubin two, Aman: I'm on zero. Let's go. Machine made
machine made instant nonstop tokenised. Da Vinci would call that making the complicated simple. Even the part where I lost. Zubin: Alright Aman can I can I take a sidestep right now? Because this is a big one. We just crossed 20,000 subscribers. Over 20,000 of you now subscribe to this show. Thank you. And if you've enjoyed watching him lose this one, as he always does, the like button is right there. Aman: Over 20,000 people, Zubin.
It's an amazing achievement. They've all sat through WHAM, CART and now MINT. That's that's Zubin: Yeah. Aman: not an audience. It's a support group. Zubin: Alright, alright. Maybe it would have been a 50,000 if we didn't have that. But anyway, all right, act four. Four shifts down, the bank moved into the moment. The hard part moved to the people, money started carrying data, and the kitchen stopped closing. Five shifts to go, then the 10th idea, and these are the ones you can act on.
So pretend I run a bank and I give you 10 minutes. What do I do? Aman: Okay, four questions. What do you build? Who is it for? And what do you guard? Then the one most banks get wrong. How do you turn the thing? Stay for shift 9. It's the one about your savings. Now, what do you build? Shift five, from products to verbs. You go back to the primitives. In software, a primitive is the smallest building block, as we know. Banking. I believe has five. And they're the verbs that have run through this whole series.
A big bank sells hundreds of products. Every one of them is one of these five with paperwork attached. Zubin: Five verbs, one line each, like part two. See. The dashboard has left the bank. Open banking did that, and open finance and widens it to mortgages, pensions, and investments. Move instant and close to free. We've just done it, right? Borrow. Credit turns up where the work is. When Layla's clinic needs 60 days of working capital,
nobody rings a bank. The offer appears inside the software, she runs the clinic. Aman: In part one, I called that the CFO layer with the bank as the regulated infrastructure underneath. Grow. The next one is idle cash gets a job. Hold that one for five minutes and then protect somebody still has to make you whole. And hold on to that one too. Now let's go to shift 6 from what the bank offers to what the customer expects. And there are three I's here. Instant.
Integrated and insightful. Zubin: So so MINT is the kitchen and the three I's are what has to arrive at the table, right? The kitchen has Aman: Yeah. Zubin: can can make it instant, integrated is something, it turns up inside what you're already doing, and insightful is the one no kitchen can cook on its own. Aman: You're getting it. Because it isn't an ingredient. It's knowing the customer. Okay, so since we're in a hurry, shift 7. From hindsight to foresight. Zubin. Zubin: Four rungs, right? Reactive, your statement a month late,
real time, your balance is right, now. Aman: Proactive, your rent's due on Friday and you're short. And then comes predictive. On the 14th of next month, you'll be short. Here's the fix and it's already priced. Zubin: So so where's the average bank? Aman: Trying to get to rung two in my opinion. Most of the industry's efforts on data is going to being right now. And and the customer we're about to meet, the one made of software, starts at rung four.
Zubin: So the bank is building for real time and the bot is shopping for prediction. Aman: That's the gap, Zubin. An agent has no use for your statement, bank statement. So that brings us to shift 8 from the account to the household. And back to the beach with John Reed. He didn't write about a customer's needs. He wrote about a set of services that will Substantially satisfy a family's financial needs. A family in 1976
Zubin: And Aman 50 years on, your bank still can't see your family. It sees four accounts, a mortgage, a card, in four systems with four profits and loss accounts, that's your legacy business model again. Aman: It doesn't know that Divya the saver and Divya the media company are one person, or that Layla and her three clinics are one life. In part one, we call that one person enterprise. The unit of banking that was never the account,
it's the household, the person who is also a business. We've moved the bank twice and never finished Reed's sentence. Zubin: And here's the awkward part. A bot will see it that way from day one. An agent doesn't work for an account, it I think works for a household. Aman: So either the bank finishes the sentence or the agent finishes it for them. I wrote the long version on A Man Who Blogs, my blog, about this Memo from the Beach. It's also in the description. Zubin: All right, which brings us to shift 9,
the last of the shifts. Aman: Yeah, shift 9 is about going from serving you to standing guard. Go back to Act One, someplace, somewhere, something. Two revolutions to get from one to the other. I think the third has started, Zubin, from something that's done for you, banking that isn't done by humans. It's done by bots. Zubin: Somebody else some bot Aman: and here's the number that makes it matter. McKinsey estimates that 23 trillion dollars of the world's consumer
deposits sit in accounts paying nothing. Zubin: 23 trillion dollars earning zero. Now that's not a product. That's I would Aman: Yeah. Zubin: say, I don't know, a nap. And and McKinsey did Aman: Expensive nap. Zubin: something. It's an expensive nap, right? If software moves just 5 to 10 per cent of those deposits or balances to the best rate on offer, an industry's profit on deposits falls by a fifth or more. 5% of the money, 20% of the Aman: That's what Layla's software was doing in the cold open. It wasn't clever, it was just awake.
The lazy deposit paid for the free current account, the branch, and a good part of the dividend. When the money wakes up, all three get repriced. And this will be the slower. It's your savings, not your shopping. Medium to long term, but think about what the bank becomes in that world. Doreen Asante is 68, a retired midwife in Croydon and a composite. On a Thursday afternoon, her grandson rings in a panic.
He needs 18,000 pounds today. It's his voice, it isn't him. Zubin: Cloned voice. You know, fraud at machine speed. You know, in one industry survey, 77% of payment executives named fraud, and I would argue cybersecurity as the biggest thing slowing them down. And in Britain, since October 2024, banks generally have to make her whole up to 85,000 pounds, split between the bank that sent the money and the bank that received it. We said it in part two liability is a product
Aman: On Friday morning, Doreen doesn't want a chatbot. She wants a person and a bank that could have stopped it. Now make it Layla. And the thing moving her money at three in the morning is software she hired. When more and more of your life is run by bots, the bank is the final control wall, the last regulated name between your money and a machine's mistake. The one that can say no. Zubin: And Doreen is the retail version. Here's the corporate one, right?
Hong Kong 2024. A finance worker at Arup, the British engineering firm, joins a video call with the company's chief financial officer and several colleagues. Familiar faces, familiar voices, he makes 15 transfers of about $25 billion. Everyone else on that call was a deepfake. Aman: It's crazy. When when I started in banking, security meant your mother's maiden name. Now it means proving you're a person. Zubin: And that's the threat you can see.
Aman: Prof Z, 60 seconds on the one you can't. Zubin: Alright. Almost everything a bank sends, your app, your card, the messages between bank is locked with math that today's computers cannot crack. A big enough quantum computer could. Nobody knows when those machines arrive, and the attackers aren't waiting. It's called harvest now, decrypt later, which we I think covered in a previous episode. Aman: Yes. Zubin: Copy the encrypted traffic today, read it when the machines turn up.
Aman: So the theft has already happened, we just haven't opened the envelope yet. Zubin: And the clock is running. America's standards body, NIST, has set the timeline. The locks most of finance uses RSA, elliptic curve, are on notice from 2030 and out to 2035. Aman: And a mortgage signed today runs to 2050. Its data has to outlive its lock. Zubin: Yeah, for those of us who lived it, it's Y2K without the date, right? At least with Y2K you knew which New Year's Eve to stay sober.
Aman: I did. I s I sure did, I was sit I was sitting on the banks of the of the Thames in London with this massive sat phone because I was one of those Y2K people. That phone that phone never went off and and probably gladly so because I had had a few drinks so so now how does a bank meet it? Three moves. One, machines against machines. A deepfake can fool a human ear, it struggles against a network that sees both ends of the payment.
Since 2023, Mastercard has scored the receiving account for 9 British banks. TSB was one of the first. And Mastercard's sum is that if every UK bank caught what TSB caught, nearly 100 million pounds of scams a year would be stopped Zubin: Yeah, because no single bank can see a mule account. The network can. Aman: Two is friction where it counts. A pause, a call-back, a question only Doreen's real grandson could answer.
Zubin: Yeah, and y you know, I'm on we've covered this in a previous episode. Agree to a family safe word and do it this weekend. Aman: Yeah. Zubin: And by the way, don't say it on this podcast, don't tell me. Aman: Three is change the locks before the deadline. Find everywhere the bank still uses the old math and build so you can swap the lock without rebuilding the door. The engineers call it crypto-agility. I call it the most boring board paper that will ever save a bank. Zubin: That's the threat. Now the opportunity,
because the same bots are about to become customers, right? We've covered this as well. Mastercard launched Agent Pay in 2025. And earlier this year, it ran live transactions made by AI agents in Hong Kong, Thailand, Australia. Visa has its own version. McKinsey puts the prize for agentic AI in payments, you know, up to $110 billion a year in productivity alone. Aman: So who vouches for the bot Zubin? Zubin: That's the business in the wall. Know your agent. Somebody has to vouch that this agent is real and acting
for this human. I think in part one, I said the banks that evolved go from deposit vaults to identity engines. Know your customer has been a cost for 30 years. It might be the first compliance department that ever turns a profit. Aman: Yeah, compliance as a profit centre. They won't have it to themselves. Your phone already knows your face. Zubin: Right, Apple and Google and others own the biometrics, the bank owns the liability, and no law yet says who pays when an agent sends
money to a wrong place. Aman: So so treat every agent like a new hire, a mandate, a spending limit, and a curfew. It can sweep the idle cash, it can't pay a stranger without asking. And the customer keeps a kill switch. Zubin: Spending limit and a curfew. You're just describing my teenage years. Aman: Skip that part of at your peril, I'd say. Gartner expects that by 2030 half of all failed AI agent projects will have failed on governance. Not the model,
the rules if you'd like, around the model. Zubin: So the research sketches five kinds of banks for 2035, and most of them end up as a licence behind somebody else's screens. The ones worth owning will be the walls. So let's go back. 9 shifts. There's a lot of change for an industry that still chains its pens. Aman: Yeah. Zubin: Alright. But every bank listening has seen all 9 in a consultant's deck, probably with a nicer font, better looking people in better suits.
So why do so few of them actually move? Aman: Because Zubin they run each one as a programme, digital data, fraud, each with its own budget, its own steering committee, and its own end date. A programme pushes one part of the bank while the rest of it pushes back. So the 10th idea isn't really a shift, it's the thing that turns the other 9 around from programmes to a flywheel.
Zubin: And I'll declare your interest for you. This is the framework that Aman uses in his, I guess, advisory work that you've been doing for a while. So yes, it's a pitch with the big banks, the big tech, and several startups worth of scar tissue behind it. Aman: Guilty. It's called the 6POS. And at the hub of these 6 Ps is purpose and people. If you'd like the psychology of change, purpose is the strategy plus the reason anyone should care.
People, of course, is the culture and the mindset that we've talked about so much in this episode and several others. On the rim around it, is product, platforms, process, and partnership. This is engineering that change. And what comes off the wheel is performance, measured three ways velocity, how fast you can launch, utilisation, whether anyone actually uses it, and then value, whether it actually pays.
Zubin: and Aman you skipped the hub, which I think everybody sort of skips. Aman: you're absolutely right, Zubin. And it carries the most amount of weight in physics and in banking. the the hub when it's weak, the entire wheel falls apart. That's shift two again. The plumbing you can buy, the people you have to change, the culture you have to curate. Zubin: And the rim is the rest of today, right? The four verbs are the product.
always on is the platform, foresight is the process, if of course you built the data in instead of bolting it on. The household and the walls. are the partnership because no bank does either. Aman: That's the whole episode on one wheel. Zubin: Alright, sceptic's question though. Every consultant has a flywheel. Amazon started Aman: Yeah. Zubin: on a napkin. What makes Aman yours different? Aman: Well, two things. Most flywheels Zubin put a product in the middle. This one puts people there because banks don't fail at change for the lack
of a platform or a product. They fail because they bought the platform and kept the mindset. I can't tell you how many examples I've seen of this. And it works from either end. Legacy banks with scale have to become modern. Modern banks without scale. Have to get big without getting old. Modern at scale is where both are heading. And very few have arrived there. Nubank to a degree Revolut Zubin: So what stops the big ones?
Aman: A doom loop. A legacy business model attracts a legacy mindset, which starves the infrastructure, which keeps everyone thinking short term. Banks are paid to protect the downside, not to chase the upside, as we talked about. And since 2008, compliance has overtaken ownership. Compliance lets people disown outcomes. A flywheel is the opposite. Anti-gravity. Every turn makes the next one a little bit easier. Zubin: Compliance lets people disown outcomes.
That's going on a mug. So where do you start? Aman: Not where you think, Zubin, product and process have the longest lead times or the long poles. Start them now, but they won't move the quarter, if you'd like. Partnerships and performance, you can pull on hard today. And all of it needs muscle you have to build. A bank can run, if you'd like, a credit committee in its sleep, but ask it to ship a product every week. or run a partnership or or end one and it ends up pulling something.
Every partnership I've seen go wrong had no pre-nup. Zubin: Department of Breakups. Alright, I'll I'll I'd watch that show. Aman: So that's the 10th idea. 9 shifts tell you where the bank is going. The flywheel decides whether your bank actually gets there. Zubin: All 10 are on the screen now. 9 shifts and the flywheel that turns them. Screenshot them by the way, or ping Aman or me and we'll be happy to share this. But 10 is a lot to carry out of the room. So what does anybody do on Monday?
Customers first. Three questions for your bank. What is my idle cash earning? who makes me whole? And can you show my family or my business on one screen? Aman: And if you run the bank three for you, find the product that isn't one of the five verbs and ask why it exists. Ask your data team which rung you're on, and time how long the answer takes, by the way. And then decide what you'd refuse to do on a customer's behalf
and say it in public. Zubin: All right, bet's time. Same rules as part one and two, dated, checkable, and the loser eats humble pie on air. Mine by December 31st 2027 A share in an American listed company changes hands on a venue run by the New York Stock Exchange on or Nasdaq on a Saturday or Sunday. Aman: Mine by 31st December 2027 an ordinary retail customer of a major bank in Singapore, the UK or America can buy into a tokenised private market fund,
private credit or private equity inside their own bank app for 1,000 dollars or less. Zubin: I was aiming to go 1,000 dollars on this bet. I'm gonna stick to 10 dollars. That's my cap. For something that has asked for 200,000. Aman: And if I'm wrong, I'll wear the pie. Zubin: Alright, okay. I can feel it coming. Come on, Aman, you know this is you. Aman: Okay. So go back to the banking hall, the marble, the queue the pen on the chain. For 300 years, the bank's controls pointed at you.
Come to us in our hours and leave the pen. The number, 70%. The most profitable industry on earth, marked down by more than two-thirds, because the market has worked out what the profit was made of. Our inertia. And software doesn't have any. The Memo from the Beach? Reed said the task isn't to run branches or issue cards. It's a family's financial needs. 50 years on, we've moved the bank twice and never finished his sentence.
So who holds the money when the software does the shopping? A bank does. But holding it was never the valuable part. Money has always been a promise. Take away the paper, the plastic, the building, and what's left is whoever made the promise. and whether they say no on your behalf. Part 1 asked you who pays you? Part 2 asked who you pay. All end in the same place. In a life run by bots, the bank is the last wall.
It still needs a chain, but the chain belongs on the machine. Keep the chain, give us back the pen. Zubin: Alright, nice one. That wasn't a mic drop, that's I'll call it a pen drop. Aman: Ha ha ha. Zubin: Alright, that's the future of work, shopping, and banking. Three parts, five verbs final score, two concessions to me, and one bet each on the clock? Aman: One question for the comments. What would it take before you let software move your savings?
A better rate? A guarantee? A human you can shout at? Zubin: And if you've stuck around this far, subscribe and send this to the one person at your bank who still thinks the app is the strategy. All three parts are in one playlist on screen now and in the description. Start with work. Aman: Next time, work in the age of AI with Sunil. Stay curious and Zubin: Stay purposeful and stay solvent. By the way, one more thing. Aman: Go on. Zubin: Somewhere in Boca Raton on the 9th green,
Wayne's phone rings. It's an AI. It has read all 4 million lines of his code and it has one question about line 4,012. Aman: Does he pick up? Zubin: No no no. He lets it go to voicemail. He's still putting Aman: Smart man.
Click here to watch a video of this episode. The Jio IPO raised $3.2 billion, the biggest listing in Indian history, and nobody sold a single share. Three weeks earlier Elon Musk had asked whether Mukesh Ambani is "the...
Click here to watch a video of this episode. Amazon blocked Meta's shopping agent in September; a month later Walmart opened its doors to approved AI agents with spending limits. The fight is not really about the...
Click here to watch a video of this episode. Imagine a YouTuber with 400,000 subscribers and about $180,000 a year who cannot get a mortgage, because her bank does not underwrite dashboards. The future of work has...
Click here to watch a video of this episode. Four personal AI agents landed in seven weeks, and three of them can already spend your money. Meta's help page puts the liability in eleven words, and Instinct's terms bind...
Click here to watch a video of this episode. Andreessen Horowitz has published 90 slides arguing that AI is not a bubble, one month after closing a $1.1 billion fund for AI hardware. On slide 88, holographic Pokémon...