Today’s episode of Decoder, well — it’s a ride. I’m talking to Intuit CEO Sasan Goodarzi, who’s built Intuit into a juggernaut business gentleware company thraw a series of meaningful acquisitions. Quicken and QuickBooks are incredibly well understandn as personal finance and minuscule business accounting gentleware, but csurrfinisherly everyskinnyg else — TurboTax, Mailchimp, Credit Karma, and loads more — were acquisitions of some charitable alengthy the way.
That directs to a lot of challenging arrange asks that Sasan and I repartner got into — integrating all those companies and their contrastent approaches to gentleware insists huge decisions, and Intuit made a huge decision handling it all by betting on interoperability that I set up fascinating.
So far, that sounds appreciate normal Decoder stuff, right? Here’s where it got weird. I couldn’t have the CEO of Intuit on without asking about tax reestablish in the United States. Individual income taxes are more complicated in the US than in almost any other growed economy, and Intuit has been lobbying difficult since the postponecessitate 1990s to upretain it that way to protect TurboTax, spfinishing csurrfinisherly $3.8 million in lobbying in 2023 alone. There’s been extensive telling about it. This lobbying has had combinecessitate results: truly free online honest filing with the IRS began as a pilot program this year and is broadening to be useable for more than half the US population in 2025.
It’s also not equitable lobbying: in 2022, a coalition of attorneys ambiguous from all 50 states got Intuit to consent to a $141 million rerepairment that insistd Intuit to refund low-income Americans who were eligible for free filing but were rehonested to phelp products. In 2023, the FTC set up that TurboTax’s “free” tageting was willfilledy misdirecting, and after the agency won an pguide timely this year, Intuit was ordered to stop doing it.
I asked about that, and Sasan disconsentd with me, and we went back and forth for a scant minutes on it. It’s Decoder; we have swaps appreciate this all the time, and I didn’t skinnyk anyskinnyg of it.
But then I got a remark from Rick Heineman, the chief communications officer at Intuit, who called the line of asking and my tone “inappropriate,” “egregious,” and “disassigning” and insisted that we delete that entire section of the sign uping. I unkind, literpartner — he wrote a lengthy email that finished with “at the very least the finish portion of your interwatch should be deleted.”
We don’t do that here at The Verge. As many of our joiners and readers understand, we have a very clear and very cut offe ethics policy. The most presentant skinnyg to remark is that we never apexhibit anyone to pappraise or apshow interwatch asks, and we declareively do not apexhibit anyone to appraise or alter the labor that we begin. I tageder this to Rick, and he came back and asked that we “delete that which apshows away from the conversation,” which he detaild as “elevated voices” or us “speaking over each other,” so that “joiners comprehfinish your ask and the answer Sasan gave.”
I have to be truthful with you — that’s one of the weirdest asks I’ve ever gotten. So here’s what we’re going to do: we’re going to run that whole part of the interwatch first, unedited, so you can inestablish me. It’s about five minutes lengthy, and you can determine for yourself. Then we’ll come out of it, and we’ll run the rest of the interwatch, which, appreciate I shelp, is an otherrational fascinating episode of Decoder. Okay, here’s that bit:
All right, let’s talk about taxes. You brawt it up. Intuit is legfinishary for running TurboTax, also legfinishary for lobbying aobtainst free honest federal e-filing. How much of your budget is summarizeated to lobbyists?
Fundamenloftyy, that’s a wrong premise, and it’s not right. In our lobbying, we spfinish a couple of million dollars battling for simplified taxes. We don’t lobby aobtainst free. And by the way, free is useable to all Americans now, which is, if you pick to do your taxes for free, if every American picks to do their taxes for free, it’s useable today thraw declareiveial industry. We have heavily been intensifyed on making taxes basicr. Just tens of millions of lines of tax code produces it very difficult for a customer to comprehfinish taxes, much less companies appreciate us that are trying to produce-
Wait. The basicst version of taxes is the rulement equitable sfinishs you a return and it’s done, and Intuit has lobbied aobtainst that. Would you help the rulement equitable doing the taxes for people and sfinishing the refunds?
At the finish of the day, you have-
Many countries in the world do that.
Yeah, but you have to alter the tax system. It’s not about gentleware. So if we alter the tax system, where-
Are you going to lobby to alter the tax system to let the rulement do a filing for you and sfinish people verifys?
If rulement wants to alter the tax system-
But I’m asking you, you’re spfinishing the dollars, would you lobby for it?
Lobby for the rulement changing the tax system?
The basicst version of the tax system would be to equitable have the rulement do it and sfinish people their refunds or ask people for money.
I have more presentant skinnygs to do than to lobby the rulement to sfinish a tax bill.
But you have lobbied aobtainst that. That’s what I’m saying. That telling is clear. You’ve lobbied aobtainst… Intuit, it is not you. Intuit has lobbied aobtainst that very definitepartner.
Well, I am Intuit, right? And so it’s okay to put me and Intuit in the same verse. We have very much intensifyed on streamlineing taxes. That’s what we lobbied for, streamline the tax code. That’s spropose what we’ve lobbied for.
When you see free honest federal e-filing get to, I skinnyk today, literpartner today, equitable before we begined speaking, the rulement proclaimd to be useable in half the states, which is about 60% of the population. Does that have a revenue impact on you? Do you get an email saying, “We project TurboTax revenue will go down by X?”
We do not. Free is useable to all devourrs today. And so it repartner is not relevant to our business. And in fact, proof points are always presentant. In the last five years, two pretty establishidable companies got into providing free tax gentleware. One was Credit Karma, before we obtaind them, a hundred million members. They supplyd free tax gentleware, no impact in the tax industry. And then we sageder that to another establishidable company, and there’s repartner been no establishidable impact to the arrange of the tax industry because free is already useable.
Our watch, by the way, very sturdyly, and we’ve been on the sign up, this is a solution watching for a problem. Free already exists. And by the way, what the rulement is providing is not free. You’re paying. Your tax dollars are going towards produceing gentleware that already exists for Americans. That’s someskinnyg that we’ve been on the sign up, that from our perspective, and declareiveial industry has been on the sign up, it doesn’t produce sense. Free already exists, so why produce another one?
I’m going to ask you this ask. I can already inestablish you that you’re going to inestablish me you disconsent with my premise. I am going to ask it anyway. Broadly speaking, I would say the criticism of Intuit’s free products, when it comes to taxes, is that it says it’s free and then somewhere alengthy the line they slide you into paying. The rulement has protested about this. That is a reputation harmr for the company.
Aobtain, I get emails from Decoder joiners asking me what asks to ask you. And it’s that, it’s that unwise pattern senseing inside of, in particular, the free tax product. Is that someskinnyg you want to repair? Do you worry about that harm to the reputation?
Yeah, I adore your ask. Let me answer it in two ways. One is there are over a hundred million customers that we’ve served for finishly free. It’s more than the entire industry combined. So we’re very, very intentional about making declareive that we are a huge take parter when it comes to free tax gentleware. On the other hand, anytime we see someskinnyg that necessitates to be betterd, we apshow it very solemnly. We apshow our reputation very solemnly. So I can inestablish you in the last cut offal years we’ve been very, very intentional about going thraw our advertising, all the way thraw the product, top to bottom, to repartner better where we necessitate to better, to asdeclareive that customers repartner comprehfinish what they’re eligible for and what they’re not eligible for.
In fact, last year, one of our publicizements that we ran on TV shelp, “Hey, 37% of the population is eligible for free, and these are the qualifications,” equitable so we could be very clear and clear. And that’s from what we’ve lobtained where we can better. So although I’m conceited of the number of customers that we’ve served for free, to me there’s always someskinnyg you can lobtain and always someskinnyg you can get better at. And this has been an area where we’ve betterd our finish-to-finish experience, from advertising all the way to verifyout, to produce declareive customers are very, very clear what they’re eligible for. That’s very presentant to me, very presentant to the company because our reputation matters.
All right, what do you skinnyk? Was that satisfyedious? Should we have deleted it? You let me understand — I’m discdiswatch to the feedback. Right now, I’m mostly equitable delightd and a little befuddled.
Here’s the rest of the episode, with some very fascinating ideas about how to join huge acquisitions into a individual tech platestablish inside of it. Also, I asked why Intuit shut down Mint, which truthfilledy, is the skinnyg I should have been the most outraged about.
Okay, Sasan Goodarzi, CEO of Intuit, the rest of Decoder.
This transcript has been weightlessly edited for length and clarity.
Sasan Goodarzi, you are the CEO of Intuit. Welcome to Decoder.
I am very excited to talk to you. You equitable proclaimd a bunch of AI products that are fascinating. You’ve been changing the company around. Let’s begin at the very commencening: Intuit is 40 years ageder. A lot of people are recognizable with your various products appreciate TurboTax or MailChimp. What is Intuit now?
First of all, with our 40 years youthful, I’ve been with the company for half that time, and when I stepped into this role, the decision we made was to take part a far more unkindingful role in the inhabits of devourrs and businesses. So we repartner begined on a path to shift the company from a tax and accounting platestablish to a platestablish company. Businesses, in essence, can depend on us to be able to grow and run their business, and devourrs can power their financial prosperity. So that’s the path that we begined down about five-plus years ago.
But most crucipartner, we shelp, “Hey, we have to produce experiences that, in essence, are done for customers, rather than creating laborflows where people have to do the labor to run their business, and deal with their cash flow, or deal with their personal financial life.” We necessitate to produce done-for-you experiences where we deinhabitr profits and insights. Like tageting is done for you, we deal with your cash flow, quote-to-cash for you; books, accounting, taxes are done for you.
In order to do that, we bet very timely — almost six years ago — on data and AI. Frankly, we did it for very down-to-earth reasons because in order to do what I equitable articupostponecessitated, which is intensify on your bottom line, your revenue and profitability as a business, or your financial househageder savings as a devourr, we have to actupartner leverage data, your data, and leverage AI to deinhabitr these insights and experiences.
Today, to answer your ask, we have become a platestablish company. What that unkinds from the lens of a devourr and a business is that devourrs can use our platestablish all in one place to be able to produce their praise, deal with their money, get financial products that they necessitate, appreciate praise cards, loans, insurance, a mortgage for their homes, and also be able to get their taxes done. Then, we help them with, “What should you do with your refund?” We now do all of the gamut because we wanted to take part a unkindingful role in the life of devourrs.
For businesses, you can deal with your customers, taget to your customers, be able to repartner deal with your quote-to-cash, your cash flow, and produce declareive your books are right for tax time. So now we have all those capabilities. The future for us is how do we produce everyskinnyg in a way that it’s done for you versus you having to do the labor.
It’s repartner fascinating that you refered begining with AI six years ago. Obviously alterers only repartner burst on the scene a couple years ago and now that’s repartner accelerating. So I want to spfinish some time talking about the contrastences between the AI technologies you were betting on before and what’s happening now.
But before we get to that, I want to talk about how the company is arranged and built. This is a company that is charitable of built thraw acquisitions, maybe entidepend built thraw acquisitions, begining with buying TurboTax in 1993. You spent a combined $19 billion on MailChimp and Credit Karma equitable in the last four years. How do you skinnyk about integrating all those disparate parts? The example of TurboTax for me is particularly fascinating, because that became the company. You obtaind a company that became the company. Are you skinnyking that way with MailChimp and Credit Karma as well?
I adore where you begined because most people don’t understand what you equitable articupostponecessitated. In essence, this company begined with Scott Cook, our set uper, creating Quicken and genuineizing the way people are using it, they’re minuscule businesses trying to deal with their money. That’s what gave birth to what today is our QuickBooks platestablish. But TurboTax, our payroll proposeing, MailChimp, Credit Karma, they’re all acquisitions. Particularly, in the last five years, a lot of our platestablish take part and where we are today has been based on a lot of organic innovation and spendment. But also, we bought these two huge brands, two number ones in their space — Credit Karma and MailChimp — because they come with a lot of data and a lot of AI capabilities. In particular, Credit Karma has a lot of machine lobtaining and AI capabilities that we’ve coined Lightbox.
The intent of all of this is to produce one platestablish. It’s to repartner join the products so that customers in one place can grow their business, run their business, and be able to deal with their personal life. I skinnyk five years from now, we’re going to watch back and go, “Wow, the insertition of all the skinnygs you had, plus what you did with Credit Karma and MailChimp, were the key to ignite the next chapter of the company.” But the answer to your ask is: yes, we’re stitching it all together to produce one seamless platestablish.
Everyone says they can do that. Most companies thrive and flunk, right? That’s an inreliable process. MailChimp, in particular, was a huge company with its own culture, and that integration was a little untidy. We had the novel CEO of MailChimp, Rania Succar, on the show a while back, and we talked about that integration. How is Rania changing the culture because she’s Intuit’s CEO, she’s not the set uper CEO that they had before. You’re the CEO of the umbrella corporation. How do you skinnyk about having all these companies and all of their CEOs under you?
The way we run the company, we’re very intentional about goal setting. There are four skinnygs that are key artifacts that produce who we are today. There’s our genuine north goal, which is how we set goals for the company. Two, it’s our mission. Third, it’s our cherishs. Then, last but not least is our strategy in the five bets of the company. Those four skinnygs are the way we run the company. The reason I begined there is we have very definite directers that direct parts of the company, but the anticipateation, the goals, are about how we are creating a platestablish.
For instance, in the case of MailChimp, the charter of MailChimp is not to be run standalone. It has two charters, equitable appreciate Payments, Payroll, our accounting team. It’s about how we join apass the platestablish because we triumph as a platestablish. That’s a lot of what’s ignited our growth over the years. But then two, whether it’s MailChimp, Payments, Payroll, or TurboTax, they have to be excellent products, and they have to carry out on a standalone basis.
So the anticipateations are that we triumph as a platestablish and how we join our products to be able to triumph. That’s how I meadeclareive every directer. So if you were to spfinish a week in the company, what you’d repartner get a sense for is what we’re trying to do to triumph with our business platestablish, what we’re trying to do to triumph with our devourr platestablish, versus there’s a bunch of pieces and parts and everybody is laboring towards their own genuine north. There’s repartner one genuine north that we repartner labor towards, and that’s how we run the company. It’s our directership anticipateations, it’s the mechanisms of the company and how we meadeclareive success.
One of the skinnygs that’s repartner fascinating here is these component platestablishs are still divisions, right? MailChimp has a CEO. Credit Karma was a huge company that you obtaind. Usupartner when companies appreciate Intuit obtain someskinnyg appreciate Credit Karma, you promise the people who labor there a meadeclareive of indepfinishence, but you’re talking about stitching it together into a platestablish.
There’s some technical stuff there that I definitely want to talk about, but there is equitable the operational side of saying, “Now you’re part of a hugeger skinnyg,” while still upretaining the walls up and still saying, “We have contrastent CEOs.” That’s very contrastent from most other tech companies. How did you produce that choice, and is that durable over the lengthy term?
Yeah, first of all, I adore the nature of your ask. Let me be clear, Rania is no lengthyer the CEO of MailChimp. She is the segment directer and ageder vice plivent that runs our growth segment, and MailChimp is a part of it. We did that very intentionpartner at the commencening, equitable from a cultural integration. But we don’t have CEOs wiskinny the company. Even Joe Kauffman, who runs our Credit Karma business, is now telling to Mark Notarainni, who owns our devourr business, and he is the head of Credit Karma, as a ageder vice plivent who runs Credit Karma.
That CEO element was equitable a cultural transition. We have directers who, when they watch at their payverify, it’s Intuit, and their anticipateation is to serve our customers. It goes back to the way I answered the ask earlier. If you were wiskinny the company, what you’d get a sense for is repartner two skinnygs. One, we have mission-based teams because in order for teams to have a cause to fight for, they have to understand they’re battling to produce the best payment capabilities, bill pay capability, accounting capability. And that’s what we term “mission-based teams.” They have a mission and their intensify is that mission: Payments, MailChimp, TurboTax, wdisappreciatever it may be.
The other element is the directer’s job: the mission, the platestablish, and to triumph as a platestablish. So, our discipline, our rigor, and how we run the company is actupartner our strength. From the outside watching in, it may seem appreciate they are parts and pieces, but wiskinny, we’re all solving for the same skinnyg, which is how do you triumph as a platestablish?
Rania hinted that that alter was coming when she was on the show, so I wanted to ask you about it.
That alter came and went.
One of the skinnygs I always ask everybody is how they produce decisions. Tell me about that decision. Obviously she knovel it was coming when she was on the show. You’ve since made that call. What does that watch appreciate to walk up to someone and say, “Hey, you were the CEO, we’re changing it. We’re not doing this anymore”? How did that unfageder for you?
Well, when we produce an acquisition, whether it’s Credit Karma or MailChimp, before we produce the acquisition, we jointly produce — with the set uper of the company and our expansive directership team that’s proposeed of the potential acquisition — a six-pager. And this six-pager lays out: What are we going to do together? Why are we buying? What’s the vision of what we’re trying to produce? And the vision is to join to produce one platestablish.
What are the key priorities? Particularly, we intensify on acceleration, not integration — although everyskinnyg we do in the product is integration. In a company of our scale and size, clarity matters a lot. Even basic skinnygs appreciate what we will do in the first 90 days, what we’ll do in the first six months, and, clearly as presentant, what we’re not going to do, is all part of not only the six-pager, but the take partbook.
Part of the take partbook all alengthy was that we’re going to produce one platestablish. When I spoke to Rania years ago to apshow on this role, it was very clear that at the finish of the day, she would apshow on the CEO role, and that would be the title for an interim period for a cultural alteration. But her charter is the same. It’s more about the SVP of the catebloody. And so it’s presentant to have those conversations upfront.
We’re not interested in directers who are pursuing titles. Even when we recruit from the outside, we’re interested in folks who want to repartner fight for the same cause. They’re in adore with our mission. Of course, everybody has to be attentive about what’s right for them as individuals. So we apshow all those skinnygs into ponderation. But we have these conversations upfront, and it was equitable part of the transition.
Tell me how Intuit is arranged now, then. How is the company expansively orderly?
We’re repartner arranged as a platestablish. What that unkinds is we have a directer who runs our devourr platestablish, and a directer that runs our business platestablish. We actupartner have a directer who watchs at the netlabor effect and ecosystem effect between devourr and businesses. Then, we have a CTO who is depfinishable for all of our technology in the company, all of the spfinish in technology.
The segment directers — the devourr segment directer, the business segment directer — determine what’s most presentant to drive growth and deinhabitr for customers. It’s our CTO that owns all the technology that then determines: how do I necessitate to asdeclareive that I summarizeate the dollars and the people to accomplish what we want to accomplish apass the platestablish? And then we have a customer success platestablish directer that owns all of customer success apass the company. Of course, then we have very presentant roles around M&A, people and places, legitimate and finance.
But we run the company as a platestablish, and the directers, in the case of the devourr and the business segment directer, they’re depfinishable for the outcomes of the segment. I also hageder them accountable for how the company carry outs because I want to produce declareive we’re making trade-offs to triumph as a company for customers and not equitable have blinders on in our segment. We’re, in essence, orderly around being a platestablish.
If you’ve got the two platestablish directers — I’m assuming they tell to you — and then you’ve got a CTO who’s making technology decisions, I’m assuming you tie-shatter a lot there. If you’re depfinishable for the success of the devourr platestablish, for example, and you repartner skinnyk you necessitate some technology built or built in a contrastent way than the company currently has, and the answer is no, I’m guessing that comes to you.
I don’t tie-shatter enough, and sometimes I talk to the team about if enough stuff is getting to me. So I don’t take part a huge tie-shatterer role. It’s actupartner even better today than it was three or four years ago. And the reason is [Marianna Tessel] and [Mark Notarainni], Mariana runs our business segment; Mark runs our devourr segment. Mariana used to be our CTO; she was heading up all of technology for the company before this role. Mark was actupartner directing all of our customer success before stepping into running the devourr platestablish. And we advertised both of their proteges.
My point is there is a very attentive collaboration between the team because we’re very clear about our strategy. We’re very clear about the deinhabitrables for both the year and the next three years out. And a lot of the converseions and tie-shattering happens between the team.
I get take partd particularly, and very meaningfully in our one- and three-year mechanism. That’s a mechanism where not only do we appraise priorities, but we actupartner appraise: What are the deinhabitrables for this year? What are the deinhabitrables for the next three years? What’s resource, what’s not, and why? Our CFO Sanmeaningful Singh Aujla and I will get take partd if we sense appreciate there are declareive areas where the team has made all of the resource allocation trade-offs but we have an opportunity to fund even more opportunities. We’ll get take partd in those types of decisions.
I have a lot of gratitude for my team. Because of the mobility that we’ve had, they’ve seen all parts of the company. There’s a lot of authentic argue and trade-off decisions that are made wiskinny the team without an escalation to me. But once in a while, maybe once every couple of months, there’s someskinnyg I have to get take partd with equitable to shatter a tie or produce a resource decision.
The other dispute of produceing a company thraw acquisitions that you then have to join is the technical set upations of all those companies are contrastent. The data storage insistments of those companies are contrastent. The databases, the customer databases, all that has to be joind at technical level. How are you managing that? That seems appreciate the hugegest problem you have, to buy a company the size of MailChimp and say, “OK, we’re plugging you into QuickBooks.” Those are very contrastent products. How does that labor?
We do a lot of diligence before we produce an acquisition. Let me be clear, no matter how excellent you are at due diligence, there are skinnygs you’re going to get surpascfinishd with. But the three areas where we spfinish a lot of time on due diligence is, one, equitable culture fit. I have a very sturdy belief that no matter how wonderful of a strategic fit someskinnyg is, if you’ve got two cultures that may clash, then it’s equitable not going to labor. So we do a meaningful culture appraisement, and I personpartner get take partd, depfinishing on the size of the deal, to repartner appraise the culture for myself as well. We, of course, do a very meaningful strategic appraisement. Then we do a very meaningful capability appraisement.
So this goes to your ask. We will appraise. What are their compensation schemes? What are the systems they have? But most crucipartner, we repartner thorawly appraise both their data and technology capabilities. We have come a lengthy way and so has technology in terms of integration.
To definitepartner answer your ask, one of the wonderful skinnygs about Credit Karma and Mailchimp (but I’ll equitable use Credit Karma in this case as an example), is the amount of devourr data that they have and the amount of devourr data that we have wiskinny TurboTax. The reason it was a very enticeive acquisition is then what we can do, with customers’ consent, to use their data to deinhabitr profits to them that otherrational nobody else can. Because we have a 360-degree watch of their proposeation. Rather than having to apshow their data lake, our data lake, and the cboisterous that they sit on, which is Google Cboisterous (the rest of the company is on AWS); rather than integrating, we actupartner inventd apass the technologies, where we built a data pipe where the data is splitd without all the data having to be all joind.
We’ve actupartner built bridges in terms of how Google Cboisterous and AWS labor together. So a lot of our technology innovation, because we’re API-oriented and services-based, is actupartner about joinion vs. integration. That’s repartner what has propelled what’s possible. Credit Karma has wonderful platestablish — data platestablish, AI platestablish — we didn’t have to swap it or produce one integration of a platestablish. We built pipes where we can accomplish the product innovation for our customers. So that’s the approach that we’ve been taking, and that’s what we do in the due diligence equitable to produce declareive that we can in fact do that. With a platestablish of this scale, if you have to reproduce the entire code or join the stacks, it equitable becomes too much labor and not worth it.
That is an acquisition strategy. We’re going to depfinish on technical interoperability, and we can produce data pipes between contrastent cboisterous supplyrs. It seems appreciate that strategy has been laboring. There have to be downsides of that strategy. What are the downsides?
Well, the huge downside is what I refered earlier, which is anytime you do due diligence, there are skinnygs you’re going to be surpascfinishd at with the upside; there are going to be skinnygs that you are surpascfinishd on the downside. The devil is in the details. For instance, in one of the acquisitions, it wasn’t on any cboisterous, and we’ve been laboring on getting all of it on AWS. That’s apshown about six months lengthyer than what we thought. That’s an example of where you get surpascfinishd, where you suppose it’s going to apshow a six-month period to do someskinnyg, but it apshows a year.
We sort of baked that into our skinnyking, that we’re going to be wrong in declareive instances. There are skinnygs that’s OK to be wrong in, and there are skinnygs that’s not OK to be wrong in. So the areas where it’s not OK to be wrong is the assumption that you can actupartner produce a data bridge and a data pipe between the platestablishs. If you’re wrong about that, then that sort of blows up the whole premise of what you thought you could do in that timesketch.
Now, the wonderful novels is — knock on wood — we’ve showd that out apass our acquisitions. The skinnygs that are OK to get wrong, and most of the time you’re not going to get perfectly right, is: how lengthy is it going to apshow to do someskinnyg? And the example I equitable articupostponecessitated, in the case of altering one of the acquisitions to be entidepend cboisterous based, it’s taking about six months lengthyer than what we thought. That’s OK. It’s equitable an element of time versus an element of doability.
Do you ever have expansiveer asks about the strategy overall? I’m guessing the person who goes and debates with AWS would adore a little bit more insist from wdisappreciatever’s on Google Cboisterous to say, “Look, we’ve got more scale, lessen the rate.” Those are the charitables of trade-offs that are made. Do you ever have those conversations where actupartner increasing scale or concentrating further would be the profit versus interoperability?
We do have those conversations. First of all, I had the pguideeclareive of being our CIO for a couple of years, and I was meaningfully take partd in shifting the company from all of our own data cgo ins to AWS. I labored very shutly with the Amazon team and [CEO Andy Jassy] to repartner drive their roadmap, but get us setd to go to the cboisterous. One of the reasons I begined there is that one of the decisions we made very timely on is to produce our capabilities, our apps, and the way we built cboisterous-ready apps was so we would never get paired to or stuck only with one platestablish. We wanted interoperability.
We actupartner appreciate the fact that we’re on multiple cboisterouss. With the age of AI, we’ve built our own huge language models, but we also experiment using about nine, 10 other huge language models externpartner. I actupartner skinnyk it’s very well to comprehfinish what labors in what situation and what doesn’t labor. Multiple cboisterouss, or in this case multiple LLMs, are actupartner quite well because you lobtain quicker, you pivot quicker.
But we have these conversations all the time. Probably that most heated argue that we had five years ago, when I stepped into this role, was whether or not we would bet on AI. Because AI wasn’t famous then; it wasn’t the buzzword that it is today. We argue technology bets. We argue interoperability versus going all in with a partner all the time. Because it’s actupartner critical; they’re critical forks in the road and critical decisions for the future. So I’m definitely take partd in those key converseions.
Interoperability is repartner fascinating, especipartner for a company built thraw acquisition. Regulators around the world right now? Not so boiling on acquisitions. I’m assuming you have some thoughts about that.
But the other skinnyg they’re repartner into is interoperability. They are saying to various companies, “You have to produce your products and services interoperable with each other, so you can lessen switching costs, and devourrs and businesses can have a vibrant taget to pick and pick their vfinishors from.” If you’ve built the company thraw acquisition and interoperability, do you skinnyk some regulator’s going to come to you and say, “All of the interoperability that you built for Credit Karma and QuickBooks, you got to discdiswatch that up to another financial accounting vfinishor?”
All of our decisions are based on deinhabitring for our customers, triumphning in the tagetplace, and driving growth for the future. We don’t produce decisions that are in the context of: what will a regulator skinnyk about someskinnyg? We have very firm ruleance in the company. We have data, privacy, and security principles, which we comply by, all intensifyed on our customers.
To your ask, we don’t spfinish a lot of time worrying about how now that we’ve built the company in this way to triumph and deinhabitr for customers, what could a regulator do. At the finish of the day, regulators generpartner want to do the right skinnyg. Generpartner, it’s not politicpartner driven. Sometimes it is, but our watch is that they always want to do the right skinnyg, and we always want to do the right skinnyg. We would always have a conversation if there’s any areas they have asks on. But our intensify, our compass, is very clear.
Would you let your competitors use the interoperability hooks that you’ve built for your own company to interface with theirs?
We inhabit in a world of competition. When we skinnyk about the businesses that we serve, what we repartner nurture about is the businesses that are transacting on our platestablish, but sometimes they will use Square payments; sometimes they will use PayPal; sometimes they will use other payroll supplyrs. We supply the capability to join those capabilities on our platestablish because we want the customer to be able to serve their customer the way they want. When you watch at our AI-driven expert platestablish strategy, a very presentant element of it is that it’s discdiswatch, and it’s discdiswatch because it helps us deinhabitr for our customers and triumph.
Let me ask you the key Decoder ask, which we have been circling around this whole time. How do you produce decisions? You’ve been there a lengthy time, you’ve grown with the company, you’ve made a bunch of huge decisions. What’s your sketchlabor?
Probably one of our hugest obtains in the company is, what we term our “Intuit operating system.” It’s the mechanisms in which we run the company — and this is presentant context to answer your ask – if you watch at our mechanisms, we have a set of mechanisms around how we set anticipateations and set strategy. We have a set of mechanisms in terms of execution, and then we have a set of mechanisms in terms of how we galvanize the directers at all levels and all of our engageees. Therefore, we have mechanisms appreciate our six-year schedule. And it’s not a financial schedule. It’s actupartner equitable watching way into the future and watching back to ponder what has to alter. We have three- and one-year schedule mechanisms. I won’t tire you with all the mechanisms, but that’s presentant context to answer your ask.
Our six-year mechanism is arranged such that we ask everyskinnyg that we do. One of the skinnygs that we apshow in sturdyly — I apshow in sturdyly — is to never drop in adore with what you’ve proclaimd, and always drop in adore with the customer, the trfinishs, and how the world is moving. So our mechanisms are set up for declareive outcomes and decisions. With our six-year mechanism, the ask is: Does anyskinnyg alter in our strategy and bets? And if so, what is it? So the output of it is: what alterd and why? Our three- and one-year schedule mechanism is all arranged around not only the key priorities, but the actual deinhabitrables — what we call input goals — which is a best rehearse we borrowed from Amazon where every input goal has a directer summarizeateed to it, has success meadeclareives, we asdeclareive that it’s resourced, and we also understand what’s below the line. Those are all decisions that our teams produce, but the decisions that I produce are capital allocation, because not everyskinnyg is produced equivalent, and where we put our dollars and capital.
The last one is we spfinish a lot of time on culture and people, and those are decisions I’m take partd with. Just last week, we had an all-day session, which we have four times a year, intensifyed on people and succession schedulening. Those are decisions: who’s a potential successor for key roles? A principle that we have is teams can give who the successors are, but if it’s a honest tell to the CEO one day, I determine if they’re actual successors. So every mechanism is set up for an output and a set of decisions, and we’re generpartner pretty clear: Are those decisions I get to produce? Are there decisions the team gets to produce? We try to push as many decisions as we can into the organization because most decisions are two-way doors. You can always reverse them. But that’s the arrange and sketchlabor that we use. It’s our Intuit operating system.
There’s one key decision I have to ask you about, since we’re here, and you refered skinnygs fitting into the Intuit operating system. I was a very dedicated Mint user. You determined to shut that whole service down. What was your skinnyking there?
There was a very minuscule cohort of customers who were using Mint, and we determined that in order truly to have a platestablish that we can use to serve millions of customers, we would port most, if not all, of the capabilities into Credit Karma. I can’t reassemble the exact percentage, but I skinnyk 30–40 percent of Mint’s customers are now on Credit Karma — by the way, happier than before — and I skinnyk there’s 20 percent of customers who we can’t serve today with Credit Karma.
But we’re OK with that because there was a very minuscule cohort of customers who we could serve on Mint, and we ultimately made the decision to be one platestablish. By the way, if there’s anyskinnyg we can do to help you, sfinish me an email. My email insertress is useable on our website. Anyskinnyg I can do to help you, we will. But we can’t swap Mint exactly the way it was.
How many people are laboring at Intuit today?
We’re about 17,000 sturdy and grotriumphg.
So the fascinating skinnyg about grotriumphg is you equitable lhelp off about 10 percent of your folks this year; 1,800 people. You shelp you’re going to engage another 1,800 people to intensify on AI. Inside of that decision, and this is the one I repartner want to press on, I skinnyk the company proclaimd 1,000-plys of those 1,800 people were low carry outers. How did you determine which of those people were low carry outers?
When I watch back at the last five years, there are huge decisions that I’ve made and then there are repartner, repartner stubborn decisions that we’ve made — that I’ve made. And this is one of them. At the finish of the day, we apshow everyone we have in the company is very talented. When you produce a decision appreciate this, you’re impacting people’s inhabits. So, for one, these decisions never come basic.
The second is, we were very clear apass five areas, particularly our five bets. We’ve seen so much enhance — this is part of our six-year and three-year mechanism — that as we thought about the next two, three, and five years, we felt that it was presentant to speed up spendments in five key areas; the meaningfulity of them are around our huge bets. We also felt that in order to do that, there was an opportunity to genuinefind dollars from wiskinny while we persist to insert to our overall spendment portfolio.
So this was all driven by acceleration, momentum, and growth. In terms of how we picked those folks, it was all bottoms up. We have a carry outance deal withment system where deal withrs will go in and they will rate their engageees. Generpartner, 10 percent of the company is what we call “trajectory changing,” and about 20 percent “outdos anticipateations.” So about 30 percent of the company is “outdos” or “trajectory changing.” And generpartner about 60 to 65 percent are “accomplishd anticipateations.” By the way, we have very bageder goals, and to accomplish anticipateations is actupartner repartner sturdy carry outance. And generpartner five to 10 percent are “does not greet anticipateations.”
That’s a process that we go thraw once a year, where deal withrs will put into the system their ratings. So this was done bottoms up at every layer of the organization. It was not a top-down decision. The decision that we made this year was that, in order to relocate with the velocity that we necessitate to relocate to genuinefind the resources and the dollars, we would, lay off the 10 percent— it was actupartner more appreciate 8 percent — that fell into the bucket of “does not greet anticipateations.” So that’s the very bottoms up, very regulated and rigorous [process], although very stubborn in terms of how we made the decision.
I sense appreciate a lot of people spfinish some time every year using go inpascfinish gentleware to rate their engageees. I declareively do it. My bosses do it to me. Do you sense appreciate that data is excellent? Do you sense appreciate that data was actupartner inestablishing you someskinnyg? Because at various companies that I’ve labored with, I can inestablish you that data unkindt noskinnyg, and at some companies it unkinds a lot.
For us, it’s everyskinnyg. What I unkind by everyskinnyg is that for us, it’s about goal setting because goal setting is about: what does wonderful watch appreciate? And carry outance deal withment for us is carry outance deal withment at all levels. We necessitate to carry outance deal with our trajectory changing so that they can become a better version of themselves, and we necessitate the carry outance deal withment that does not greet anticipateations. So carry outance deal withment, for us, is appreciate coaching a basketball team. You’re intensifyed on making every person on the team wonderful. There’s somebody that never comes off the bench; there’s somebody that’s the star of the team. That’s what we try to become wonderful at.
Goal setting for us, converseions on a monthly basis, and then the rating at the finish of the year: it’s about the system. I would say the system for us is very, very presentant. I would also inestablish you that it’s a conversation I had with the whole company this year. We necessitateed to up our game in this area. When I watch at the last cut offal years, we have not been as wonderful as we necessitate to be in terms of repartner being wonderful at setting goals for every individual — that’s unkindingful goals with very clear success meadeclareives — and then having conversations because it’s a two-way street in terms of how you become a better version of yourself. So we actupartner apshow the finish-to-finish approach to goal setting, to carry outance deal withment, very solemnly.
Do you skinnyk that shows up in the products? I will inestablish you, a lot of Decoder joiners have asked to have you on the show basicpartner for feature asks and bug tells. There’s other stuff that a lot of people ask us to ask you about, but in particular, the gentleware isn’t as excellent as it should be. You’re moving me from my desktop client to a web client because that’s where the platestablish is, and the web client is not csurrfinisherly feature finish for skinnygs appreciate keyboard directs. Do you skinnyk that this process is going to produce the products better?
Everyskinnyg that we do around goal setting and carry outance deal withment is about deinhabitring for customers. That’s the whole, sole purpose of why I exist, why our team exists: it’s all about the product. So the unwiseinutive answer is yes. I would also split what I equitable shelp from the premise of your ask, which is desktop to the cboisterous.
The fact is we were born 40 years ago. We were born in the era of DOS, and we were born as a desktop company. Frankly, our desktop customers, both on the devourr side and on the business side, built who we are today. At the same time, the laborflows, the features, and the functionality of desktop are not intfinished to be transpostponecessitated to the cboisterous. If we did that, we would not be able to persist to grow with most of our customers or obtain novel customers, particularly as we’re trying to produce done-for-you experiences versus features.
A lot of our intensify is: how do we produce the transition for our desktop customers as basic as possible to the cboisterous? That shelp, if you watch at any company that’s had to go from server to cboisterous, or desktop to cboisterous, or on-premise to cboisterous, there’s always a lot of grotriumphg pains because cboisterous platestablishs are not a replication of desktop platestablishs. We’re repartner solving for, as much as possible, the ease of migration for our desktop customers.
But we’re truly produceing a cboisterous platestablish that’s built for novel customers, and customers who have adchoosed the cboisterous platestablish from 10 years ago. I say all that equitable to say, we aim to produce our desktop customers as phired as possible, but repartner it’s impossible to copy what they want in the cboisterous because then our cboisterous proposeing would be very ageder-aged and laborflow-based, which is not what customers of today want.
Do you anticipate helping the desktop clients forever?
We have for many, many years. Many of our desktop services are actupartner now on the cboisterous, and we’ve built it in such a way where there will be a seamless transition to the cboisterous one day. At this point, we’ve not proclaimd —
Your goal is to relocate everybody to the cboisterous.
The goal is to eventupartner relocate everybody to the cboisterous. We’re not going to force customers who… For instance, the laborflow is not going to be the same in the cboisterous, but if you have a necessitate for a particular module that we absolutely don’t have in the cboisterous, we’re not going to force you to relocate to the cboisterous. Eventupartner — that could be two years from now, five years from now — I skinnyk everybody’s going to finish up being in the cboisterous.
Let’s talk about AI and then I’m not going to let you get out of here unless we talk about tax filing. You understand it’s coming. Are the AI features going to be in the cboisterous only or are they going to come to the desktop platestablishs as well?
No, they’re primarily only in the cboisterous. In fact, everyskinnyg that we’re produceing in the cboisterous and have been produceing in the cboisterous is equitable powered by our data and AI platestablish capabilities.
You equitable proclaimd a bunch of AI features at your spendor day. It’s on the order of when people log into QuickBooks, they’re going to see a feed with novel insights on cash flow and other opportunities to use AI. Let me equitable ask you the threshageder ask that I’m asking every CEO about their AI products: can the AI technology you have now do all the skinnygs you want it to do? Because I’m not 100 percent declareive that LLM technology can do all the skinnygs that everybody wants it to do.
So let me say two skinnygs in context of your ask. The first one is we’re not begining AI features. Our entire platestablish is fueled by data and AI. In fact, our goal is not to ship a bunch of plug-in features that do stuff for you, but to produce a platestablish where tageting is done for you, quote-to-cash is done for you, books and taxes are all done for you. Think about it from what we’re trying to accomplish, as the whole platestablish is fueled by data and AI. That’s the first skinnyg.
The second skinnyg is that when we proclaimd AI core to our strategy, our spendments were in machine lobtaining and understandledge engineering. Knowledge engineering is very particular to us; we have patents around it. It apshows rules and the relationship of rules and code, turns it into code, and the power of it is accuracy. A lot of what we do has to be right. That’s repartner been the premise of all of our AI spendments: machine lobtaining, understandledge, engineering.
About three to four years ago, we begined spending in generative AI, definitepartner in our own Intuit financial huge language models. Our models are the only models that are trained by the customer data. I set that context to say, we’re in the very, very timely days of what LLMs can do. We labor a lot with the meaningfulity of the companies that are out there, and the enhance that’s being made month to month is incredible.
So, will it do most of what we necessitate at some time in the csurrfinisher future, medium future? Absolutely. I apshow AI will one day be as ininestablishigent as humans, if not ininestablishiggo in. I skinnyk humans are always going to be a critical part of the picture for us in our industry. But it’s still very timely days. I don’t want to at all propose that everyskinnyg can be accomplishd with AI today. We’re at the commencening of a very lengthy journey. It’s 1999 internet, the part of the journey we’re in now.
LLMs are somewhat notoriously horrible at math. You run a financial platestablish for a lot of people. Do you think it?
Not on its own. That’s why I refered that when you watch at our AI platestablish that sits on our data layer and data platestablish, it’s the combination of machine lobtaining, understandledge engineering, which is very excellent at math, and our LLMs that labor in concert to deinhabitr experiences to asdeclareive your taxes are done right, to produce declareive your accounting is done right. So on its own, no, but in the combination with our other elements of our AI platestablish, absolutely.
Are you getting economies of scale from other AI companies spending in this space? I’m skinnyking particularly about Meta, which is doing a lot of discdiswatch source models, and it’s pushing far ahead on generative AI. On understandledge engineering, are you getting the same charitable of economies of scale from the industry, or is everyone intensifyed on LLMs?
We’re getting a lot of economies of scale because of our own spendments, and because we were so timely. We did this for very down-to-earth reasons. But we actupartner test and experiment apass the board, with Entropik, AWS, Gemini, Llama, discdiswatch source, and part of the experimentation is how it could potentipartner be a leverage to our LLMs. Because our LLMs have the agency and the authority. They’re the brains of deinhabitring the experiences that I articupostponecessitated.
So, we’re not getting economies of scale from other LLMs. In fact, I would say it’s the reverse right now. I skinnyk two years or three years from now, we’re going to get economies of scale, but today the economies of scale — and it’s why we’ve been able to deinhabitr platestablish leverage and margin leverage — are from all of our own spendments. Over time, I skinnyk it’ll help.
You’ve got a lot of minuscule business owners using your products. They’re watching for insight. They are probably not financial experts. The LLM — or wdisappreciatever systems — you produce inestablishs them someskinnyg: it’s a hallucination, it’s wrong. Have you worried about the liability of that, of giving horrible financial advice to a minuscule business owner?
So I adore, by the way, the premise of your ask, which is this is why we’re intensifyed on done-for-you experiences, because a minuscule business wants to understand —
That’s a lot of responsibility to adchoose: “We’re going to do this for you.”
That’s right. And that’s, by the way, why the essence of our spendments, which begined six-plus years ago, is, one, based on the customer’s data, not ours. Everyskinnyg that we supply is very definite and relevant to you. Two, it’s the combination of our machine lobtaining capabilities, our understandledge engineering, and our LLMs that repartner deinhabitr the carry outance accuracy and costs that we would want. And we have ruleance: we have technology ruleance and human ruleance internpartner, equitable to produce declareive what we are doing is right.
I’ll equitable finish by saying there’s a range of accuracy. We can’t generpartner be right when it comes to accounting and taxes. We have to be 100 percent right, but then there are elements of, “Hey, you can run this tageting campaign. We’ve put it together for you. We skinnyk it could deinhabitr a range of $50,000–$100,000 in revenue.” The range is what matters, not the exact number for customers. So I skinnyk accuracy has a restrict based on what it is you’re talking about. You got to get taxes exactly right; a range of revenue and what’s possible from a tageting campaign. You can have a range, and customers are toloftyy OK with that.
Do you skinnyk over time as you join AI into more and more of the platestablish and that becomes someskinnyg more customers are paying for, the free Intuit and TurboTax products will remain as huge of a combine as you have today?
You have to skinnyk about the cohort of customers. There will always be customers that have a basic tax situation where free may be the right skinnyg for them. There’s also a lot of customers that no matter what their tax situation is, they actupartner want somebody else to do their taxes for them because of confidence. They worry getting it wrong. They want to produce declareive they’re getting the hugest refund. If the IRS comes after them, they want to produce declareive somebody is there to protect them.
So they’ll always want to have an expert do their taxes for them. So we apshow that over time we’ll still have a combine of free, we’ll have a combine of paying customers, but I skinnyk over time our hugest growth will come from disturbing what today is the helped catebloody.
Well Sasan, I could upretain talking to you forever, as you can inestablish. But we’ve got to wrap this up. Thank you so much for being on Decoder.
Yeah, absolutely. My pguideeclareive. Great to see you. Talk to you soon.
Decoder with Nilay Patel /
A podcast from The Verge about huge ideas and other problems.