OpenAI Integrates Plaid Data Into ChatGPT Finances to Audit Expenses

OpenAI推出ChatGPT Finances以实现家庭支出自动审计 主动型对话式AI模型为普通家庭提供比传统财富管理机构面向千万富翁客户更为严格的税务和支出优化服务。 Ethan Bloch概述了如何通过Plaid集成银行账户并结合电子邮件上下文，将ChatGPT打造为一个自主、主动且能迅速收回成本的个人CFO。围绕六大核心使用场景，他展示了对经常性支出的审计、代办商户纠纷、主动发现税收抵免项、替换高费率投资基金、优化信用卡权益，以及为休假进修或购房构建交互式情景模型。Bloch还谈及了隐私安全防线——解释了只读财务数据如何与模型训练隔离开来——并分享了他的团队如何利用Codex实例在无需冗长规范文档的情况下，每周发布数十次产品更新。 OpenAI于2026年4月收购了Ethan Bloch的金融大语言模型初创公司以推出ChatGPT Finances，该功能通过Plaid集成银行动态并结合电子邮件上下文，以实现个人财务管理的自动化。 (0:44) 通过Plaid抓取的原始交易和余额数据均安全存储，绝不会用于训练OpenAI模型，用户还可以在设置中停用基于聊天的对话训练。 (9:22) ChatGPT Finances团队每周进行大约47次部署，他们摒弃了传统冗长的产品需求文档（PRD），转而采用工程师主导的Codex实例和可运行原型。 (36:58) 直接银行集成仍为只读且仅限于美国，但国际用户仍可通过上传对账单或关联Google Drive与电子邮件来进行税务规划和投资组合审查。 (23:40)

Ethan Bloch, product lead for ChatGPT Finances at OpenAI, argues that consumer AI should readily pay for itself by uncovering billing errors, lowering fund fees, and modeling complex financial decisions. By combining read-only banking feeds with inbox context, the platform shifts personal finance from static dashboards and reactive tax preparation to continuous automated monitoring and interactive scenario planning. Data Privacy and Read-Only Account Ingestion (7:14) ChatGPT Finances connects to financial institutions through an embedded, read-only Plaid interface, refreshing balances multiple times daily without handling user credentials, account routing numbers, or balance-moving permissions. Bloch states that raw transaction records and balances held in secure storage for categorization are never used to train OpenAI models. However, financial figures entered directly into chat conversations remain eligible for training runs by default unless users manually opt out within their settings. Users can optionally link personal email and Google Drive accounts. This context pairing allows the model to inspect vendor receipts, billing notices, and historical tax documents alongside banking feeds, bridging separate data silos that traditional personal finance applications leave disconnected. Recovering Cash via Subscriptions, Fees, and Rewards (3:32) Bloch designs ChatGPT Finances to recover at least its $20 monthly subscription fee by identifying recurring billing anomalies. During demonstrations, the engine flagged price increases, hidden account fees, and duplicate charges, drafting dispute emails directly to vendors to secure refunds without requiring users to negotiate with customer service representatives. The model applies identical audits to investment fee drag and credit card usage. By analyzing connected brokerage and retirement accounts, ChatGPT identifies bloated mutual funds and presents lower-cost index fund or ETF alternatives. For credit cards, it inspects multi-card transaction patterns and searches connected emails for unactivated benefits—such as TSA PreCheck or CLEAR credits—reallocating spending categories to ensure users maximize cash back and avoid hitting reward caps. Proactive Tax Strategy and Interactive Life Modeling (17:55) Contrasting with traditional CPAs who operate reactively at tax deadlines, ChatGPT Finances generates year-round tax mitigation strategies, quantifies potential savings by bracket, and maintains living to-do lists across chat sessions. Bloch asserts this provides continuous planning historically limited to expensive family offices. International users without US account linking can replicate similar workflows by uploading documents or connecting Google Drive. For major life transitions—such as taking a career sabbatical or weighing renting versus buying a home—the model generates interactive web applications with adjustable financial parameters. Users can test scenarios against strict reserve floors or analyze specific property links. When partners decline to link their bank accounts, users can manually enter asset holdings or upload PDF statements to integrate outside wealth into retirement projections. Scheduled Monitoring and Agentic Execution Limits (33:53) In addition to ad-hoc queries, the system supports background automation, including default weekly summaries and custom monthly audits to monitor recurring charges and email joint spending digests to spouses. Bloch also schedules checks to flag uninvested cash balances sitting in checking accounts that could be earning higher interest in savings accounts. However, autonomous execution remains constrained by financial institutions. While users might look to browser agents to automate investments on a recurring basis, Bloch notes that institutional security filters actively block automated bot traffic at authentication. OpenAI is working with US financial institutions to whitelist approved ChatGPT traffic, but downstream financial operations currently remain strictly read-only. Internal Codex Development and Voice Roadmap (36:58) Internally, the ChatGPT Finances team maintains high development velocity—shipping 47 updates in one week—by replacing traditional product requirement documents with agentic prototypes. Engineers guide Codex instances using Slack notes and roadmap goals, moving straight to dogfoodable software and reducing time spent talking about work from 30% to roughly 5%. Looking ahead, Bloch emphasizes removing prompt engineering barriers so mainstream users receive proactive insights automatically upon sign-up. Furthermore, OpenAI is preparing personal finance integration for its voice team, aiming to enable users to review spending alerts, track budgets, and evaluate complex portfolio trade-offs through real-time voice interactions. While ChatGPT Finances delivers high-end financial modeling and immediate expense recovery, its practical value currently depends on user initiative in crafting prompts and navigating read-only constraints. Realizing fully autonomous personal finance will require frictionless zero-prompt onboarding and formal institutional agreements that safely permit downstream financial execution.

[0:31] Ethan Bloch: I felt like a hero. And then by 2001, I sort of had my entire net worth wiped out. And so by the age of 15, I'd experienced what many people, you know, wait till later or maybe never experience. [9:22] Ethan Bloch: That raw financial data of your transactions or your balances is never used for training. That'll sort of sit in a secure store. [17:55] Ethan Bloch: Help me figure out the best way to optimize taxes based on my current situation. [17:55] Ethan Bloch: Make yourself a to-do list on what I need to do to actually get these next steps done. [24:58] Ethan Bloch: Analyze my ETFs, mutual funds. Are there any replacement funds that could be cheaper? [26:11] Ethan Bloch: Analyze how I've been using my credit cards. Am I using them in the best way? Right, to maximize my rewards. [28:20] Ethan Bloch: I'm thinking about taking six months off. Build me an interactive site to help me think through this. I want three scenarios. [28:20] Ethan Bloch: you can't just call them every, you know, on a Tuesday on a whim because you're thinking of something and have them crank a model for you. [32:35] Ethan Bloch: And like I have a degree in finance. This is like this is my area. Like it's too hard, [32:35] Ethan Bloch: I'm considering buying a house. Here's how much. Build me an interactive site to help me think through it. [39:46] Ethan Bloch: Can we stop talking about the work and do the actual work? [39:46] Ethan Bloch: it's like 5% talk about the work, 95% do the work. [42:44] Ethan Bloch: the goal for personal finance is to make it so ChatGPT can save you time, money, and stress when it comes to your money. [43:42] Peter Yang: you just turn it on... and then it just works for you.

[0:00] Chat GBT can just pay for itself. I have [0:02] one that will run monthly, [music] which [0:04] is checking all my subscriptions. Have [0:06] any changed in price recently. Are there [0:08] any you think I don't need anymore based [0:10] on what you know about me? And that will [0:12] just run each month and ping me. Help me [0:13] figure out the best way to optimize [0:14] taxes based on my current situation. I'm [0:16] considering of buying a house. Here's [0:18] how much. Build interactive site to help [0:19] me think through it. When I'm making [0:21] certain big financial decisions now, [0:23] there's nothing I will make a decision [0:24] about before I talk it through with [0:26] chat. And I want to show you also what [0:27] that looks like. [0:31] Hey everyone, my guest today is Ethan, [0:33] the product lead for Chhat GBT Finance. [0:35] Really excited to get Ethan to show us [0:36] how to use the product to save money and [0:39] hopefully, you know, grow a portfolio. [0:40] So, welcome Ethan. [0:41] >> What's up, Peter? Thanks for having me. [0:43] Really happy to be here. [0:44] >> Dude, you you spent most of your career [0:46] uh on this financial problem, right? [0:47] Like on solving like helping people with [0:49] their personal finances and everything. [0:50] So, maybe before we get to the product, [0:52] can you talk about your background and [0:53] what you think AI unlocks for personal [0:56] finance? [0:57] >> Yeah, for sure. For sure. [0:58] So, you know, for whatever reason, I [1:01] have been obsessed with finance for a [1:04] really long time. Actually, my first [1:05] love was a computer at like six. And [1:08] then my second love was finance for [1:10] whatever reason. By, you know, 10, you [1:12] know, I had had a credit card in my [1:14] mom's name. I was balancing a checkbook. [1:16] I had a bank account. And when I turned [1:19] 13, I started uh day trading my bar [1:22] mitzvah money. My parents, for some [1:24] reason, thought that was okay. It just [1:26] sort of kept pulling me, you know, and [1:27] so it's just been a common theme. Again, [1:29] when I was 13, this is sort of to set [1:32] the time, 1998, so it was sort of the [1:34] bull market of the internet bubble. I [1:36] felt like a hero. And then by 2001, I [1:38] sort of had my entire net worth wiped [1:41] out. And so by the age of 15, I'd [1:43] experienced what many people, you know, [1:45] wait till later or maybe never [1:46] experience. And again, it's always just [1:49] a passion that, you know, eventually led [1:52] me to where I am now. But I think most [1:55] relevant to what I work on today. Way [1:58] back in 2012, I'd actually started a [2:00] company called Digit that was an AI [2:02] financial chatbot that helped consumers [2:04] save money. It was text message based [2:07] and you would talk to it. And I really [2:09] thought the technology was good enough [2:11] to build an AI for everyone to help [2:13] manage their money. We ended up helping, [2:14] you know, millions of Americans save [2:16] billions of dollars of the digit [2:17] product. It was a 10-year journey, but a [2:20] lot of what we wanted to do just wasn't [2:22] possible. the tech wasn't good enough. [2:24] People would talk to Digit and it would [2:25] tell him, you know, I'm sorry, I'm [2:26] afraid I can't do that. And so I thought [2:28] that was that and Dig ended up getting [2:30] acquired in 2021 and I thought that [2:34] dream sort of died more or less. But as [2:37] we all know in 2022, Chad GBT came out [2:40] and for me it was another sort of aha of [2:44] maybe what we wanted to do at Digit [2:46] would now be possible. I really didn't [2:48] want to start another company. I just [2:50] done a 10-year journey and I didn't [2:52] really have to work and so I'm like what [2:53] am I going to do with my life ended up [2:55] starting another company focusing on [2:57] what's now possible with LLMs and [2:59] financial data with my co-founder [3:01] RushBub we started that for you know [3:05] really a year or so got an idea for a [3:08] product started building that fast [3:10] forward openai ended up acquiring us [3:12] earlier this year in April of 2026 [3:16] and so now I'm here within Cha GBT [3:18] bringing forth all this, you know, [3:20] passion and knowledge for building this [3:22] AI for consumers that I've been thinking [3:24] about for so long. [3:25] >> Got it. Okay. Why don't you show us the [3:26] product, man? I think a lot of people [3:27] still haven't tried Chad Finance. [3:29] >> Yeah, sounds great. [3:30] >> So, show us how to get to it. [3:32] >> So, I'm going to walk through a few [3:34] different sort of modes and things you [3:36] can do. I think the first mode that's [3:39] probably the, you know, the easiest to [3:40] get started with is I want to show you [3:42] how anyone can sort of watch ChatgBT pay [3:46] for itself. And what I mean by that is, [3:48] you know, today our finances product is [3:49] included in a plus and pro subscription [3:51] which costs $20 a month. We've been [3:53] working to make sure Chat GBT can just [3:55] pay for itself. So if you pay $20 a [3:57] month, how do you have it save you or [3:59] find $20 a month you otherwise didn't [4:01] know you could do? And so let me start [4:03] there and then we'll get progressively [4:04] more complex and more complex and you [4:06] can leave me wherever you want to [4:08] continue to go. Peter, [4:09] >> just have it cancel all the other AI [4:10] subscriptions. That's the way [laughter] [4:13] >> starting place. If you sign up for chat [4:16] GPT, sign up for a plus subscription, [4:18] connect your finances, right? So, use [4:20] our finances connector to connect your [4:22] accounts, which I can show you and we [4:24] can get into that in a sec. Also, I [4:26] would suggest you connect your email. [4:28] You can think about it like this. If you [4:29] connect your financial accounts and you [4:30] connect your email, chat has all this [4:32] context to help make sure, right, you're [4:34] not getting overbuild. When you do get [4:36] overbuild, it can email the merchant, [4:38] the biller, for you on your behalf. It [4:41] can make sure you're not getting double [4:43] build. it can look at all the [4:44] subscriptions to see what you're [4:45] actually using versus what you're not [4:46] using. And so this is a really important [4:48] point that these two sort of areas of [4:50] context, your financial accounts and [4:52] your email give Chad GPT superpowers to [4:56] help you get more dialed in with your [4:57] money to help you save money. And I'll [4:59] I'll show you what that looks like right [5:00] here. [5:01] >> So this is really simple prompt. Anyone [5:03] can do this, right? Review the last 90 [5:04] days for subscriptions. Look at price [5:06] price increases, duplicate charges, [5:08] refundable fees, you know, unusually [5:10] expensive household bills. So, right out [5:12] the gate, it in my case is like, hey, it [5:15] looks like this self- storage, this is a [5:17] fictional character, so bear with me. [5:18] And some of this data is also synthetic, [5:21] but uh go along the story with me. Just [5:23] pretend I'm Jim. [5:24] >> And hey, Jim, you have this potentially [5:27] duplicate charge, right, for your [5:29] storage. It looks like you might have [5:31] been duplicate charged for this hair [5:33] salon. In this case, I know this was me [5:34] and then my kids. Look, Evernote [5:37] recently increased their price. Is this [5:38] still worth it? Went up from 129 [5:40] to$2.49. and then checking fees and then [5:43] some other services that I use. And so [5:45] from here just very quickly gives you [5:47] that context. What would take you a lot [5:49] of time to pull through on your own and [5:51] I can go back and forth. But what's [5:53] really powerful and this is me asking [5:55] about my subscriptions. I'm going to [5:56] jump down. I then ask chat and I'm like, [5:58] "Hey, draft an email to self storage. [6:01] Tell them I've been build twice and then [6:02] send this off." Yeah. Yeah. And so, you [6:04] know, I had an example like this in my [6:06] own life with a hotel that I used points [6:09] to book it and then they still build me. [6:12] I had no idea it happened. Chat called [6:14] it. I had chat deal with their customer [6:16] service and I ended up getting [6:17] reimbursed. And this just another [6:19] example of that. [6:19] >> Nice. I remember I was trying to cancel [6:21] some hotels for my parents because [6:23] there's some health issue. And uh yeah, [6:25] I basically got chat GBT to email the [6:27] hotel and I I told her to like schedule [6:29] a task to monitor for any replies and [6:31] keep going back and forth with the hotel [6:32] to to get the money back and uh it [6:35] worked. [6:36] >> Yeah. [6:36] >> Yes, that's exactly. Yeah. I I'm at this [6:39] point now where I will almost not [6:43] interact with any service providers [6:45] myself. [6:46] >> It's all chat GPT over email through [6:50] their browser, right? when you have to [6:52] talk to the support agent. Um anyway, we [6:54] can talk more about that later. [6:57] >> So, this one again is just a really [7:01] great way just right off the bat, you [7:03] know, to see how chat can pay for itself [7:06] and how chat can just get you money that [7:09] you otherwise just might have left [7:11] leaving around or don't have the energy [7:12] to deal with or don't want to deal with. [7:14] Now, let me take a step back and I want [7:16] to show you some of the finances sort of [7:19] dashboard you get when you set this up. [7:21] Does that sound good, Peter? [7:22] >> Yeah, that sounds great. [7:23] >> And so, when you do this, you're going [7:24] to want to tag out finances. You can [7:26] also go to more and you'll see finances [7:30] in here. I don't have it because I have [7:31] it pinned here. That's the other way. If [7:33] you tap that, it'll help onboard you [7:35] into the product. But from chat, you can [7:37] just ask it. [7:40] And in this case, I have accounts [7:41] connected. But if you don't have any, [7:43] you'll see this sort of institution [7:45] connection widget where you can tap here [7:48] and connect for the first time. You [7:50] know, we use Plaid for all of our bank [7:53] integrations. Plaid's the gold standard [7:56] for securely connecting your bank to a [7:58] service. When you're entering in, you [8:00] know, your login or your password to [8:02] connect to your bank, none of that's [8:03] going into chat GPT. That's passing [8:06] directly to securely to Plaid and then [8:08] Plaid is interfacing with your bank. I [8:11] think it's also worth noting when you [8:12] connect your bank to chat GPT, ChatGpt [8:15] never gets access to your routing number [8:17] or account number. It can't move any [8:19] money. This is sort of just a readonly [8:21] data connection that chat then has to [8:24] your bank. I'm not going to connect [8:26] here, but this kind of shows you some of [8:27] the familiar Plaid UI to do that. [8:30] >> Yeah. [8:31] >> Once your bank account is connected, you [8:34] could continue to go back and forth in [8:37] chat to talk about it. Then you also can [8:39] just go to the finances tab to get this [8:41] sort of bird's eyee view of what's [8:42] happening with your money. This episode [8:45] is brought to you by Whisperflow. [8:46] Whisperflow saves me at least 3 hours a [8:48] week and is one of my favorite AI apps [8:51] by far. It's just so much faster to [8:52] dictate to AI using your voice than to [8:55] type. You just talk naturally and it [8:57] outputs clean, ready to send text. [8:59] Whisper Flow even removes filler words [9:01] and [music] formats your sentences for [9:03] you. I use Whisperflow for everything, [9:05] including drafting newsletter posts, [9:06] writing product specs, replying on [9:08] Slack, and more. It works on Mac, [9:10] Windows, iPhone, and Android across all [9:12] of your favorite apps. Try a free at [9:14] whisperflow.com and use my code peter [9:18] whispererflow to get 6 months free. [9:20] That's Peter Whisperflow. Now, back to [9:22] our episode. Let me ask you this before [9:24] we move on. So, Chad uses Plaid, which [9:27] is used by a lot of other fintech [9:28] providers, too. I think just like normal [9:30] people are just worried that like chatbt [9:32] will learn that I only have $8,000 in my [9:34] bank account or [laughter] something or [9:36] like I'm spend a lot of money on this [9:37] and that and um can you give some [9:39] reassurances about like you know [9:40] training on this financial data or [9:41] something is that [9:42] >> yeah yeah I'm happy to happy to talk [9:44] through this a little bit so there's a [9:46] couple things to just sort of pull apart [9:48] when you connect your bank account to [9:50] chat GPT you know chatbt will download [9:53] those transactions from your bank [9:55] account or your balances in order to [9:57] enrich them and enriching literally [9:59] means give proper categorization. So you [10:01] can have a great you know spending [10:02] categorization widget like this. It will [10:05] detect subscriptions and sort of just [10:07] using intelligence to connect dots [10:09] amongst your transactions. That raw [10:12] financial data of your transactions or [10:14] your balances is never used for [10:16] training. That'll sort of sit in a [10:18] secure store. It sits there and it's not [10:22] not used in any way for training. Now [10:24] there's a nuance here where if you're in [10:26] a conversation talking about you know [10:29] spending on Door Dash and talking about [10:31] Door Dash enters a conversation if it in [10:34] a conversation that could be used for [10:36] training. So I think it's an important [10:38] distinction and as a consumer you have [10:40] the ability to turn training off and so [10:42] you can go to settings and you can turn [10:44] training off for your conversations [10:46] which would then ensure that we also [10:48] can't train even when that data would [10:51] enter your conversation. And so you have [10:53] full control over what you're [10:54] comfortable with. [10:55] >> Great. Yeah. Thanks. Thanks for that [10:56] reassurance. [10:57] >> Yeah. And so from the dashboard, you [10:59] know, there's a couple familiar things [11:00] that if you used a financial product [11:02] before will seem familiar, right? You [11:04] can see your spending by category, which [11:07] we're sort of breaking down for you. You [11:09] can change this if you want to see it [11:11] month over month. [11:12] >> You can drill into a category. So I'll [11:14] like go back to August. Okay. What's [11:16] dining and drinks? I'll collapse this so [11:18] we can see more. [11:20] And this will give me a readout of, you [11:25] know, what's basically been happening in [11:26] this category, even calling out like [11:29] what my larger subcategory was. And then [11:30] you can see the subcategories within [11:32] there. And you can see recent [11:33] transactions it considered to be uh [11:35] dining and drinks. [11:37] >> Nice. [11:38] >> If you keep going down, you'll get a net [11:39] worth widget. Um, and it's worth [11:41] mentioning if you're in just regular [11:43] chat talking, you'll see these widgets [11:46] also show up. You don't have to come to [11:48] the finances tab to see these, but the [11:50] finances tab is a helpful way to quickly [11:51] get that bird's eye view of what's [11:53] happening with your money. [11:56] >> If I've connected investment accounts, [11:57] which I did in this case, you sort of [11:59] get this daily mover widget. [12:03] You can see sort of your upcoming [12:05] incoming and outgoing cash flow or or [12:08] subscriptions or u bills. [12:16] If it's not sure about a transaction and [12:19] what category it is, it will do its best [12:21] job guessing. But if it's can't figure [12:23] it out, you'll get this category review [12:25] widget where you can tell it, you know, [12:27] what this is. [12:29] >> And I guess tell through chat or [12:30] something or [12:31] >> you could also do it through chat. Th [12:33] >> this [clears throat] one actually will [12:34] be through UI. You can, you know, make a [12:37] change here. But if I was doing this in [12:39] chat, I could literally just say like ax [12:41] find hotels. that's hotels and it would [12:43] update that category for me. [12:45] >> I remember I don't think it was CHB [12:46] finance. I remember using like another [12:48] provider and they classified Mancinis as [12:50] Italian restaurant but it's actually a [12:52] mattress orderer that I got. [laughter] [12:54] It's like a $4,000 bill. So yeah, I had [12:58] to for sure. [12:59] >> Yeah, we you know I've been working as I [13:02] said in in consumer fintech for a long [13:04] time because we're using models and sort [13:07] of frontier intelligence to categorize. [13:08] It's the best I've ever seen out of the [13:11] box. like connect your account [13:13] >> and you'll just see accurate [13:14] categorizations. That being said, there [13:16] are oneoffs like that that will still [13:18] happen and we're working hard to make [13:20] this even better and better. And there's [13:22] also cases where, you know, you might [13:24] think something is shopping, but I think [13:26] it's services and that's just really [13:28] hard for us to guess and get right. And [13:29] so there will still be, you know, some [13:31] categorization if you want there to be. [13:33] But out of the box, you'll get a lot of [13:34] accuracy even if you do nothing. [13:37] >> Awesome. you know, we'll we'll surface [13:39] fees and interests and then you'll sort [13:41] of start getting this account breakdown [13:43] and asset allocation widgets as well. [13:45] Um, in this case, I don't think I've [13:47] connected enough accounts to go to the [13:49] next detail. [13:51] So, [13:52] from a dashboard, you know, a lot of [13:55] what you'd want is here, but this [13:58] requires you to come into the app, go to [14:02] this dashboard, and ask these questions. [14:04] So, it sort of requires more work from [14:06] you. And so what we've also done is when [14:08] you sign up for the finances product in [14:10] chat GPT, you'll automatically get a [14:13] sort of weekly finances update that will [14:15] get sent to you proactively. And this [14:18] will get either pushed to you over email [14:21] andor you'll get a push notification. [14:22] And you can turn this off if you don't [14:24] want it. But we find that users love [14:26] this because, you know, it'll be the [14:28] weekend, they'll have a moment, they'll [14:30] get this push and all of a sudden [14:33] they're in this place and I'll just show [14:34] you what it looks like where they're [14:36] getting an update sort of week over [14:37] week. How's my spending so far this [14:40] month versus last month? We'll just come [14:42] to you [14:44] in terms of my spending. What have I [14:45] been spending on? And you'll kind of get [14:47] this similar widget. And then also, [14:49] what's been happening with my net worth [14:51] and my investments? [14:53] And then lastly, if it thinks there's [14:55] things you could be doing or take action [14:56] on, it'll give you suggestions like you [14:58] see here at the bottom. [14:59] >> Yeah, this is great. Yeah. [15:02] >> So, so I'll just get like a notification [15:04] on chat GP on my phone and like check [15:06] check this out. Right. [15:07] >> Exactly. You don't need to do anything. [15:09] You just sit back. [15:10] >> And what's cool is some of these [15:11] suggestions [15:13] you could actually have chat GPT do or [15:15] follow up for you on. I mean, in this [15:18] case, I don't think these would work, [15:19] but sometimes the suggested action will [15:21] be applying for something, cancelling [15:23] something, and you can be like, "Hey, [15:24] like, take care of that for me." And it [15:26] will go off and it will attempt to do [15:27] that for you on your behalf. [15:28] >> Got it. What if I want to set up a a [15:30] triggered a triggered task where if my [15:32] wife buys too much Amazon.com stuff, it [15:34] just sensor [laughter] it just sensor [15:37] ego saying like, "Hey, stop stop buying [15:38] stuff." [15:39] >> You could you could do that. You could [15:40] do that. Even in fact, I'll show you [15:42] this really quick example. Um, [15:45] this Door Dash one. So, I spend a lot on [15:48] Door Dash or Jim does in this case. So, [15:50] I'm just like first talking to Chad. I'm [15:51] like, "Hey, how much have I been [15:52] spending on Door Dash?" Okay, here's [15:54] this graph that it's built for me over [15:56] the last six months. Okay, TLDDR, too [15:58] much money spent on Door Dash. [16:00] >> So, down here, I'm like, "Okay, please [16:02] notify me when I'm approaching, 1500 in [16:05] a month." And, you know, give me a heads [16:07] up. And so, that's what it's doing right [16:09] here. [16:09] >> I see. [16:10] >> And you could create I mean, you could [16:11] create this notification for anything. [16:12] It could be spending category. It could [16:15] be transaction. It could be balances in [16:17] account. It is very flexible based on [16:20] what's important to you. And this goes [16:21] back to what you said at the beginning. [16:22] You know, what's possible now that [16:23] wasn't it's meeting everyone where they [16:26] are doing exactly what they need when it [16:28] comes to their money. [16:29] >> And and how often uh is this data [16:31] updated like my bank account data or [16:33] something like is it based on my [16:35] requests or is it updated on a regular [16:37] basis? [16:38] >> So, it's updated on a regular basis. [16:40] Each bank's a little bit different. You [16:42] know, we again work with rely on Plaid [16:44] for all bank data and I believe Plaid [16:47] will re try and refresh your account [16:49] several times per day and when they get [16:51] a refresh, they'll sort of send us a [16:52] ping and we'll come and grab new data [16:54] when it comes in. [16:55] >> So, it should be updating multiple times [16:57] a day. [16:57] >> And um just for the people who are like [16:59] token conscious like so you don't have [17:00] to use like so max to do all this stuff. [17:02] You can just use because I guess it's [17:04] just reading financial data, right? It's [17:05] not actually making any changes. [17:07] >> Yeah. [17:08] >> Yeah. Yeah. Everything I'm showing you [17:10] so far, again, if you're on a plus plan, [17:12] it's not going to any come anywhere [17:14] close to hitting any of your usage [17:16] limits. [17:17] >> Okay, got it. Great. Cool. [snorts] [17:19] >> So, let me move from, you know, we we [17:22] covered this sort of really quick ways [17:26] if you sign up for this to have chat GPT [17:28] pay for itself. I want to move from [17:31] there and talk about things that are [17:34] other ways to have chat help you save [17:36] money or be a little bit smarter with [17:38] your money. And these sort of somewhat [17:40] get progressively more difficult. And [17:42] then I want to go over this bucket of [17:44] just when you're making like for me [17:46] personally when I'm making certain big [17:48] financial decisions. Now [17:50] >> there's nothing I will sort of make a [17:52] decision about before I talk it through [17:54] with chat. And I want to show you also [17:55] what that looks like if that sounds [17:57] good. [17:57] >> That sounds great. Yeah. I I can't wait [17:58] to do the tax plan by anyone. Yeah, [18:00] >> that's where we're going next. That's [18:02] where we're going next. You read my [18:03] mind. Everyone's favorite thing to do. [18:06] No. So, really basic prompt, right? [18:10] Again, I've connected my financial [18:11] accounts. In this case, um this is not [18:14] as reliant on email, but again, having [18:16] email does help because it can use that [18:17] for cross contact understanding, [18:19] especially if you have a CPA or you're [18:21] getting, you know, um tax d tax sort of [18:24] information sent to your email. just [18:26] makes chat even smarter at helping you. [18:28] So, really straightforward prompt, [18:30] right? Help me figure out the best way [18:31] to optimize taxes based on my current [18:32] situation. [18:33] >> Okay, it crunched in this case for two [18:35] minutes and it comes back and it has [18:39] immediate recommendations for my case. [18:42] >> And you know, for me, I don't have the [18:46] attention span to always read all of [18:48] this. And so I'll flip [clears throat] [18:49] through this and I'll look for the big [18:50] buckets, but then I'm actually um asking [18:54] like, okay, how much money could this [18:56] actually save me? I want to know if this [18:57] is worth my time, right? So this case, [18:59] it's telling me how much this might [19:00] actually save me in taxes depending on [19:02] my tax bracket. Then I'm like, okay, [19:04] that's all great. Like, make me an image [19:06] and tell me what's up here. [19:08] >> And so this is me just like quickly [19:10] visualizing here are the different ways [19:12] I could save on taxes ordered by [19:15] potential impact. And even it will show [19:17] me down here there's things I shouldn't [19:19] even consider because the impact may not [19:21] be big enough. [19:22] >> Wow, this is great. Yeah. Is this a [19:24] pretty common pattern that you use to [19:25] like understand something without read [19:27] through everything? [19:28] >> Yes. Yes. I'll either do Yeah, I'll do [19:32] an image um and I'll show in a different [19:34] one or I'll have chat actually visualize [19:37] it and that will just be like in the [19:38] chat conversation it can use these rich [19:40] visualizations to help make things [19:42] easier to digest. [19:44] >> Got it. And then if it's more complex, [19:46] I'll have it make me a site. [19:48] >> Or if it's something I'm going to sit [19:49] down and talk through with my wife, I'll [19:51] also have it make me a site. [19:52] >> That that makes sense. That makes sense. [19:53] Yeah. [19:54] >> So, but still for me, right, this still [19:56] feels like too much work. Like, okay, [19:58] great. I'll potentially save, you know, [20:00] $6,000, which is insane, but still too [20:03] much work. So, I'm like, chat, make [20:05] yourself a to-do list on what I need to [20:08] do to actually get these next steps [20:10] done. And so here it is making me my [20:12] to-do list to go through this. And then [20:14] what I would do from here is on my phone [20:18] or you know in these sort of micro [20:19] moments in the day I would ask it to [20:22] sort of take that next step on the to-do [20:23] list. Some of that might be asking me [20:25] for information giving it access to [20:27] documents or letting it go do something [20:29] with cloud browser to go take those next [20:31] actions. But I would push chat as far as [20:33] it can until it needs me to final review [20:36] something or file something or contact [20:38] someone. [20:39] >> Got it. So, you make a to-do list, but [20:41] then you tell to do all the work, [20:42] basically. [laughter] Yes. Do [20:44] >> the to-do list. [20:45] >> And the to-do Exactly. And the to-do [20:47] list helps me, you know, I'm I'm not [20:50] maybe a tax expert. I'm not, but it [20:52] helps me understand what it's suggesting [20:54] in a more digestible format. And it [20:57] helps me keep track of where we are in [20:58] the process. And it helps chat GBT keep [21:01] track of where it is in the process [21:03] because this is now an artifact, you [21:05] know, you could reference from any tax, [21:07] any chat conversation, excuse me. And [21:09] you can come back into the same chat and [21:11] tell it to go the next step or you could [21:12] be somewhere else and it can reference [21:13] it and go to the next step. [21:14] >> Got it. And do you also like uh let your [21:16] accountant see some of the stuff or like [21:18] kind of like how do you share this stuff [21:19] with your accountant? [21:20] >> Yeah, typically my workflow will be like [21:22] my accountant is emailing me has some [21:23] questions. I will actually have you know [21:26] chat uh read that email, tell me what it [21:29] needs for me. Sometimes it needs some [21:31] documents, sometimes it needs an answer. [21:33] I'll have it grab the documents, draft [21:36] the response. In some cases, I'll just [21:38] let it send the response because it [21:39] makes sense to me. In other cases, I'll [21:40] check it and then let it send to my CPA. [21:43] >> Got it. Yeah, dude. Like just from my [21:46] personal taxes, like yeah, I have no [21:47] idea about how to save money on taxes [21:48] either. And a couple of areas where it [21:50] saved me time is my accountant asked me [21:53] for like all these like income [21:54] statements and blah blah blah, right? [21:55] Like here's like a list of all these [21:57] income statements. So, I usually ask uh [21:59] Chad P looking in my Google Drive to [22:01] find all the stuff or like try to find [22:02] it from the websites that I use. That [22:04] took a ton of time. Um, [22:07] >> amazing. You're on the frontier. [22:08] >> On the frontier. Yeah. And another time [22:11] is like this. He sent me like a This is [22:12] probably bad, but he sent me like a my [22:14] my business's tax statement. It's like [22:17] 50 plus pages and it's like, "Hey, [22:19] Peter, can you review this? I have no [22:21] idea what was going on here." So, just [22:23] >> we all get CPA. [22:25] >> Yeah. [22:25] >> Yeah. [22:26] >> Like, I'm gonna look through 50 pages of [22:27] this, you know? [22:28] >> Yes. And even if we even when we do, [22:31] >> we don't know what we're reading. [22:33] >> Yeah. You know what I mean? Like we we [22:35] don't have the time to figure this out [22:36] to depth and get Yeah. [22:40] >> Your your pattern and your usage is is [22:43] common among people who are comfortable [22:44] right with these tools. [22:46] >> Everyone's doing this now [22:48] >> because it's more rigorous and it saves [22:51] us time. I also think like uh [22:53] accountants depending on how good your [22:54] accountant is like they don't actually [22:56] proactively help you save money on taxes [22:57] throughout the year, right? They come at [22:59] the end and they're like, "Hey, you got [23:00] to pay your t taxes now." So like so [23:02] actually like chat GP can be more [23:04] proactive. [23:05] >> Yeah, [23:05] >> that's right. 100%. There's like you [23:08] know we talk about this a little [23:09] internally. If you're using you know [23:11] chat to its max in some ways you're [23:14] getting you know you're getting a whole [23:16] bunch of abilities that even people with [23:19] you know 5 10 20 plus million dollars [23:22] aren't getting from their service [23:25] providers. Like that level of detail, [23:29] care, rigor, and help is available to [23:31] everyone now. [23:32] >> Yeah, that's amazing. Yeah, maybe you [23:34] can prompt it like, "Hey, pretend I'm a [23:36] billionaire. Help me save [laughter] [23:38] all the money on my taxes." Yeah. [23:40] >> I did. Yeah. And in tax, I'll just like [23:42] highlight this as well just from my own [23:44] experience. You know, a lot of a lot of [23:46] people will focus on, you know, their [23:48] investments and where am I going to get [23:50] alpha and how's it allocated and they'll [23:52] sweat all this stuff because it's sexy. [23:54] I mean, investing is interesting. It's [23:55] the reason, you know, people like it. [23:57] But for most people, the most money that [24:00] they could be saving or impacting their [24:02] life is all in their taxes. [24:04] >> People are underoptimizing their taxes. [24:06] And it compounds every single year. And [24:08] there's no risk with those decisions in [24:11] the sort of classic risk sense like [24:12] there is an investment that you're [24:14] making. Mhm. And just so like for people [24:16] who are not in the US, the way it's able [24:18] to find all this stuff is just looking [24:20] at the IRS website or something and like [24:22] you know if I'm in Canada can can also [24:24] get us benefit. [24:25] >> Yeah. [24:26] >> Yeah. Yeah. I think so caveat that the [24:29] finances tool today and the finances app [24:32] is only available to users. So you won't [24:35] be able to connect your financial [24:36] accounts to chat but you would still [24:38] right be able to connect Google Drive [24:40] that has stuff in it. You would be able [24:41] to connect your email. You would be able [24:44] to use, you know, model intelligence to [24:45] help answer questions. Again, the model [24:46] will use web search to help answer [24:48] questions more accurately that are more [24:49] relevant and timely. So, all of that's [24:51] available, you know, wherever you are in [24:53] the world. [24:53] >> Got it. Okay, cool, man. This is super [24:55] helpful. Yeah. [24:56] >> Okay. So, let's move to So, this is [24:59] really easy, really fast. If you connect [25:01] your brokerage accounts, your IRA, send [25:04] chat this prompt, right? Analyze my [25:06] ETFs, mutual funds. Are there any [25:08] replacement funds that could be cheaper? [25:10] This is going to be a really fast way to [25:12] know if you're, you know, buying some [25:14] mutual fund or ETF where there's a [25:16] comparable one that's the exact same but [25:19] lower fee and becomes almost a [25:21] no-brainer to swap into it. [25:24] >> Uh, and we also see depending if you [25:25] have a human advisor or not or a human [25:28] broker, [25:29] >> sometimes they don't even know that [25:31] you're in some fund of fund that has [25:33] these ETFs that have really high fees. [25:35] And so this is another quick check and [25:36] it'll compare it. It'll look for [25:38] comparables, right? It'll estimate your [25:40] annual savings uh if you were to flip. [25:43] But again, really good check just to [25:44] make sure you're not leaving money on [25:46] the table in regards to any of your [25:48] brokerage accounts or IAS. [25:49] >> Yeah, my my my parents are investing in [25:51] mutual funds and they can probably save [25:52] a lot of money in this. Just switch to [25:54] ETFs. [25:55] >> They should you should run this for [25:56] them. [25:56] >> Yeah. [25:57] >> See them see them next. [25:59] >> Yeah. [26:00] >> Okay. So let's move to a little bit more [26:07] I wouldn't say complex but sort of less [26:09] frequent sort of higher stakes. Actually [26:11] before I do that I I lied. Let me go to [26:13] this other one. So a lot of people like [26:15] optimizing their credit card rewards [26:18] even if you're not going to spend a lot [26:19] of time. There's some people that will [26:20] go extremely deep and spend an [26:22] incredible amount of time but I would [26:24] encourage people to just spend you know [26:26] five minutes six minutes took to run. [26:28] Another prompt, right? Analyze how I've [26:30] been using my credit cards. Am I using [26:32] them in the best way? Right, to maximize [26:34] my rewards. [26:35] >> And chat will look at your cards, [26:38] >> look at your spending, look at the [26:40] categories you're spending on, [26:42] understand the member benefits for your [26:44] card, understand its reward structure, [26:46] understand points value, and then tell [26:49] you if there's anything you should be [26:50] changing about the credit cards you [26:52] already have so you can get more more [26:54] value out of them. [26:55] >> Oh, that that's amazing. So, it will [26:56] look at all the card statements and [26:57] everything and try to figure this out, [26:59] right? Yeah. [26:59] >> Yeah. In this case, it's really just [27:01] looking at transactions and uh if you [27:03] have your email connected, it can also [27:05] be looking at member benefits like AX [27:07] has these benefits. Are you using them? [27:09] Well, it'll know if you're not because [27:10] you didn't get the email saying you [27:11] activated it. And it'll say, oh, you [27:13] should turn this clear benefit on or [27:14] this TSA pre-check benefit. [27:16] >> Got it. Got it. [27:17] >> And from here is what I was talking to. [27:18] I'm scrolling through here. Let me just [27:20] go to this last one where I was [27:21] prompting. But in this case, I wanted a [27:23] visualization. And so I just ask it, [27:25] hey, please visualize this for me. And [27:27] now you get sort of this visualization [27:30] of I have this sort of intricate case [27:32] where I have a credit card that has a 4% [27:35] reward structure, but only up to [27:37] $10,000. And it's telling me earlier [27:40] last year I was going way above that. [27:42] And so I wasn't getting the rewards. And [27:44] you can see I figured that on my own, [27:45] then I brought that spend down. And now [27:47] I'm almost never hitting that limit. [27:49] >> So I might be leaving something on the [27:50] table here. [27:52] >> And so that's one part. And then it's [27:53] telling me where to move spend, right? [27:56] In terms of this AAX general spend, I [27:59] should move to a cashback cart. And [28:01] again, if you go down to here, it's [28:02] showing me why it's making those [28:04] recommendations [28:05] and how much money. [28:07] >> Yeah. Could be better off with, right? [28:08] This is like a $2,900 potential gain [28:11] >> just on rewards in this case. [28:13] >> Wow. Wow. I I should make a Chach site [28:15] with like, you know, credit card reward [28:17] powered by Chacht. [28:19] >> You could. [28:19] >> That would be very useful. Yeah. What [28:21] else? What about this thing? take time [28:22] off. [28:23] >> Does that resonate with you? I mean, [28:24] you're not really taking time off, [28:25] though. [28:26] >> You're working your ass off. [28:28] >> Uh, you know, my my schedule's pretty [28:30] flexible. So, yeah. [28:32] >> So, you know, for for people who may or [28:35] may not watch The Office, there's a case [28:36] where like Jim's thinking about taking [28:38] time off uh to pursue a startup. He's [28:40] not taking time off. He's about changing [28:41] his job. So, this isn't the perfect [28:43] analog for that, but I basically [28:45] created, you know, a prompt, which is, [28:46] I'm thinking about taking six months [28:48] off. Build me an interactive site to [28:50] help me think through this. I want three [28:52] scenarios. I gave it a little bit of [28:54] detail on how much I think I would be [28:55] spending and still making in that new [28:57] world being that my wife would still be [28:59] working. [29:00] >> Okay. [29:00] >> And you know, it runs this and then it [29:03] gives me this. I'll flip over to here. [29:05] This is another one. And it's a really [29:09] easy way to visually understand what [29:12] these different paths mean for your [29:14] savings. And in this case, I want to [29:16] make sure I don't ever go below this [29:17] $21,000 in savings. And I can see like [29:20] not only can I take that six months off, [29:23] I can use some of my cash to fund my [29:26] small business or startup idea, which [29:28] again is this still orange one. I get [29:29] the lowest, but I still don't get down [29:30] to three-month reserve. Over here, I'm [29:33] like, you know what? What if it was [29:34] seven months off? What if it was eight [29:36] months off, nine months? And because I [29:39] have these toggles, it's a really quick [29:40] way I might sit down, you know, with Pam [29:42] or my wife and talk through this so we [29:44] can have something that really helps us [29:46] think exactly how we think about our [29:48] finances and make this really important [29:49] decision together. [29:50] >> I guess you can build something similar [29:51] for like retirement planning, you know? [29:53] >> Oh, exactly. [29:54] >> Yeah. Mhm. [29:55] >> Again, this goes back again to the [29:58] beginning when we were talking about [29:59] personalization. Like whatever decision [30:01] you're thinking through, you can really [30:04] build or or quickly get chat to build [30:07] the sort of model interface, decision [30:10] tree, you know, just just paring back to [30:14] you what you're thinking so you can [30:16] think it through in a deeper way that [30:18] before this was just way harder to do [30:20] and way harder to provide this level of [30:21] service. Right? This is a level of [30:23] service you might have gotten if you [30:25] paid for a fee advisor or even an AUM [30:28] based advisor. And even then, [30:31] >> you can't just call them every, you [30:33] know, on a Tuesday on a whim because [30:34] you're thinking of something and have [30:35] them crank a model for you. You know, [30:37] you can now be on the comfort of your [30:38] couch, have some a hairrained idea [30:40] maybe, and quickly start to think [30:42] through Chad GBT. [30:43] >> Dude, you know what's the number one [30:44] question I ask Ch Finance? [30:46] >> Tell me. [30:47] >> I think I think you don't have this [30:49] problem, but like I always ask it. When [30:51] can I fat fire? [laughter] Fat fire. And [30:54] I should have build a dashboard for me. [30:55] Yeah, [30:56] >> that's it. Exactly. And I It's funny. I [30:58] actually did that um for Jim. [31:01] >> Yeah, for Jim. But because of Jim's [31:04] situation, it's a little bit [31:06] underwhelming. Like he can fire full [31:09] fire at 76 with his current financial [31:12] situation, [31:13] >> which is a little bit late. [31:14] >> A little bit a little bit late. Yeah, [31:16] >> you could lean fire at 66. [31:17] >> Yeah, but you can't lean fire in the [31:19] barrier. You have to you have to [31:20] [laughter] [31:21] >> Well, yeah. Jim lives in Pennsylvania [31:23] though, right? Again, if we like we're [31:24] living with the I think Pennsylvania. [31:26] We're living with the um the the the [31:28] sort of fantasy here. [31:29] >> Let me give you some product fee [31:30] feedback. So, [31:31] >> please [31:31] >> I asked her when can I fat firefire, [31:33] right? But to properly fatfire, I need [31:34] to connect my wife's financial accounts [31:36] too and she doesn't want to do it. [31:38] [laughter] [31:39] I have to convince her or like someone [31:40] to share with the stuff with with her to [31:43] have her connect her accounts. [31:44] >> Yeah. Do you do you know some of the [31:46] financial details of her accounts, but [31:48] she just won't connect them? [31:49] >> I mean, yeah. I mean, she has like a lot [31:51] of Microsoft stock and other stuff, so [31:53] it'll bring my fatfire age down a lot, [31:55] you know? So, [31:56] >> so you could just you you honestly could [31:58] just tell chat. [31:59] >> Yeah. [32:02] Oh, I see. Just tell her how how much [32:04] she has. [32:05] >> Yeah. [32:05] >> Yeah. [32:05] >> And it will add it to your net worth. [32:07] It'll add it to your assets and then it [32:09] would reflect right in um it would [32:12] reflect for you. So you don't actually [32:14] for some of these you don't need to [32:15] connect accounts necessarily. You can [32:17] just tell chat you could upload a [32:18] statement you know if you had one or you [32:21] just tell it like this and now it'll [32:22] track it. [32:22] >> I I see. I think it's still better to [32:24] get my wife's accounts somehow. I'll [32:26] figure out [32:27] >> I totally agree [laughter] the sort of [32:28] the Yeah. Yeah. I'll let you work on [32:30] that. [32:31] >> I can help. Yeah. [laughter] [32:33] >> Got it. [32:34] >> Cool. [32:35] >> All right. Let me let me do one more if [32:38] you think that'd be useful, Peter. On [32:40] home buying. [32:40] >> Yeah, sure. Yeah. So again, just another [32:42] example. Again, I used the office for [32:45] inspiration if people might remember [32:47] this episode. [32:49] Um Jim's going to surprise. He's going [32:51] to buy a house, buy his parents house. [32:53] And a lot of thing, a lot of times in my [32:56] own life, I would build a model, take me [32:59] hours. Should I buy or should I rent? [33:02] And like I have a degree in finance. [33:04] This is like this is my area. Like it's [33:06] too hard. And so with again a sort of [33:09] prompt here, I'm considering of buying a [33:12] house. Here's how much. Build me [33:13] interactive site to help me think [33:14] through it. [33:15] >> Yeah. [33:16] >> And then I get this really great way to [33:19] see the difference between buying versus [33:21] renting. And I have these toggles over [33:23] here that obviously will adjust it to [33:25] again give me some way to think about [33:27] this more deeply in a more personalized [33:29] way than I could have otherwise. [33:31] >> And you can probably like uh pull [33:32] information for a specific house that [33:34] you want to buy, right? Like you just [33:35] point to some ref and link or [33:37] >> totally could. Yeah, this case I just [33:38] gave it an amount. You could give it the [33:40] link to the exact house you're [33:42] considering, [33:43] >> have it build interactive site for you. [33:45] If you wanted, you could have it build a [33:46] spreadsheet that you you could work on [33:48] as well. You could just keep it in text [33:50] and help you think through what it means [33:51] for you financially. [33:52] >> Got it. Super helpful, dude. So, let me [33:54] ask this. So, we just walked through [33:55] like five or six different use cases and [33:57] one of them is a weekly finance update [33:58] that happens automatically, but like I'm [34:00] always looking for Chad GP to do more [34:02] work for me proactively. So, I'm just [34:04] curious like if you have any other [34:05] scheduled jobs like maybe you can just [34:07] talk about them like that that actually [34:08] help you manage your finances. [34:10] >> So, we right now again when you sign up [34:12] for finances we just create one of those [34:14] for you and we do that because there is [34:15] a limit on how many a plus user and pro [34:17] user can have. We don't want to take too [34:19] many of those up by default. But what [34:22] we'll see users do and what I will do, [34:24] you know, I'll personally have a handful [34:26] of these. So, I have the weekly finances [34:28] update. I have another one that will run [34:30] monthly, which is checking all my [34:32] subscriptions. What are all my [34:34] subscriptions? Have any changed in price [34:37] recently? Are there any you think I [34:39] don't need anymore based on what you [34:40] know about me? And that will just run [34:42] each month and ping me. I have another [34:44] monthly that runs that actually sends to [34:47] me and my wife and we'll show spending [34:50] this month versus last month and we'll [34:52] also look at what's happened with our [34:53] investments and we'll use that as a way [34:55] to just stay a breast and actually email [34:58] back and forth is actually pretty great [34:59] because it'll send an email to both of [35:00] us and we can engage on that surface [35:02] which is really awesome. I think those [35:04] sorry and then I have another one which [35:06] will sort of be a cash monitor and so [35:10] what this basically is is like hey again [35:12] I think I have this run I think monthly [35:15] but it's basically do I have too much [35:19] idle cash sitting around in any accounts [35:22] that isn't earning enough in terms of [35:24] interest or that I should consider [35:26] investing because sometimes we just [35:28] forget [35:28] >> yeah that specifically like we had some [35:31] employer account we forgot to roll it [35:34] over or we did and it's sitting in cash [35:35] is one sort of extreme example. But the [35:38] other one is maybe I have a couple [35:39] thousand dollars too much in my checking [35:41] account that I could be earning, you [35:42] know, three and a half% in my savings [35:44] account and it'll monitor and tell me [35:46] that as well. [35:46] >> And the one with your wife, you just get [35:48] to send your email or do you get just [35:50] pin chatbt or something or [35:52] >> I haven't sent an email to both of us. [35:54] >> Okay. I think like technically like all [35:55] the stuff is read only but technically [35:58] if you really wanted to like for example [35:59] if I want to invest in like the index [36:01] fund every quarter I could technically [36:03] use browser use to like try to get to [36:05] invest it right to just keep keep the [36:08] >> I think yeah so I I believe with and we [36:11] talk you know we are in in partnership [36:14] and conversation with most of the [36:15] financial companies in the US most of [36:18] them will for good reason have at the [36:22] off step something that blocks agents [36:24] because they're protecting themselves [36:25] from people nefariously doing that. [36:28] >> I see. I see. [36:28] >> So that step for some will get blocked. [36:31] There are others we're working with to [36:33] help whitelist when it's coming from [36:34] chatbt. [36:36] >> So it's off and it's approved and you [36:38] could do that but it will be hit or [36:40] miss. [36:40] >> Got it. Okay. That that makes sense. [36:42] Yeah. Because I've uh for better or for [36:43] worse I I just started like chat GPT is [36:45] basically my operating system now. Like [36:47] I don't actually use any websites or [36:48] apps separately if I can avoid it. So [36:52] yeah, but I understand financials is [36:54] more uh sensitive. [36:55] >> Yeah, exactly. Exactly. [36:57] >> All right. Well, this is super helpful, [36:58] dude. Cool. So, quickly, let's talk [36:59] about working on a chatb finances team, [37:02] right? In some of these other teams, [37:03] like, you know, in in Codex Chat app, I [37:05] I see them shipping like multiple times [37:06] a day and just like doing all kinds of [37:09] stuff. Are you guys more careful on the [37:10] chatb finances team or how do you guys [37:12] work? [37:13] >> Well, yeah. So, you know, we're [37:14] certainly extremely careful when, you [37:17] know, there's different parts of the [37:18] stack that can have different [37:19] ramifications for users, but you know, [37:22] there's different modes of us being [37:23] careful across that stack, especially [37:25] when it comes to dealing with someone's [37:27] data or the data pipeline or the data [37:30] store, but we've been fortunate to build [37:32] our system in a way that, you know, [37:34] we'll ship all day long. So, just in the [37:37] last seven days, we've shipped 47 [37:38] things. [37:39] >> Wow. [37:39] >> For our small team. And so, do the math [37:42] right. multiple times a day. Some are [37:43] really small bug fixes, polish, [37:45] improvements, latency. Some are like [37:47] smaller features that will show up and [37:49] obviously some bigger features as you [37:50] know like will get broken down into [37:52] smaller features that will be shipping. [37:53] >> Got it. [37:54] >> This is true everywhere here. You know [37:56] we as long as we have you know the right [38:00] uh testing the right checking the right [38:03] actual like you know automated testing. [38:05] The goal is to build as fast as we can [38:07] in a way that again is integrating the [38:09] user experience in a way that's secure. [38:11] all that has to be met and then ship as [38:14] fast as you can. [38:15] >> Okay. And uh you know I haven't been a [38:16] PM for a few months now but like and now [38:18] that I'm solo I just talk to you know AI [38:20] the whole day but like you know you [38:22] actually have a team and other people. [38:23] So how do you bring chat GPT into your [38:25] meetings and your conversations or like [38:27] how do you balance talking to the AI [38:28] versus real humans while doing your job? [38:30] >> [laughter] [38:30] >> Yeah, it's interesting. What I find is, [38:34] you know, collaboration with real humans [38:36] is just as important as it's ever been, [38:39] but you're actually collaborating kind [38:41] of higher up the stack in a sense. [38:44] Instead of, you know, me collaborating [38:47] with design on this small little thing, [38:49] we're, you know, on what color this [38:51] might be. I trust, you know, Rich in our [38:53] case, he's gonna he's a pro. He's going [38:56] to figure that out. we're collaborating [38:57] on like that next feature that's [39:00] somewhat tricky in terms of the user [39:01] experience. So we're like higher up the [39:03] stack or I'm collaborating with an [39:04] engineer on a new feature but all the [39:07] way up at the new feature level not like [39:09] oh this little bit down here or let me [39:11] write a PRD that's seven pages and we [39:13] can talk about it that largely doesn't [39:16] exist unless it's an extremely complex [39:19] feature. we're just able to collaborate, [39:22] you know, higher up that I don't know I [39:24] don't know the exact description, but [39:25] higher up that stack because we can own [39:27] so much more ourselves and do so much [39:29] more ourselves. We're like almost like [39:32] managers [39:35] interacting higher up the stack and we [39:36] both have teams of people. [39:38] >> Yeah. [39:38] >> And so it's more of like a coarse grain [39:40] collaboration than that fine grain [39:42] collaboration that might happen in you [39:45] know historically on a product team. And [39:46] and we talked about this on Twitter a [39:48] little bit, but like what's the artifact [39:49] that comes out of that collaboration? [39:51] You know, like the artifact for the [39:52] agent to actually build, right? Like [39:54] [laughter] or is just like a is this [39:56] like a transcript of the meeting or [39:57] something or like what's the artifact [39:59] that comes? [40:00] >> It really depends. It really depends, [40:03] you know. So in obviously in meetings [40:05] we'll have noteaker notetaker will take [40:07] notes. Notes get pumped into slack. We [40:09] can revisit them. Our codeexes, our [40:11] tachibbts can reference slack and [40:13] reference the notes and so it has that [40:15] context. It obviously can reference [40:16] linear. It can reference some of our [40:18] strategy docs. And so it really depends. [40:20] But you know when an engineer is [40:22] starting a feature, it might start with [40:24] a conversation right of just like what [40:27] are we thinking about that feature? And [40:28] really it's probably in our road map [40:30] somewhere at a pretty high level of [40:32] granularity, not a super detail level [40:34] granularity. And then that engineer will [40:36] likely spin off one or many codeexes [40:39] talking to it, explaining it to get it [40:41] going on the actual feature. And then [40:43] they'll see what it did. And you know [40:45] the engineers on our team, they have [40:47] high product taste, high product sense. [40:49] They kind of know what I'm thinking for [40:51] the feature and if they don't, they'll [40:52] ask from that high level and they're [40:54] able to direct from there codecs to get [40:56] closer and closer to what it should be. [40:58] And then they'll have a dog food version [40:59] and then we'll look at it together. [41:00] >> Okay. [41:00] >> And so long way saying like there's no [41:02] artifact written in many cases. There's [41:05] a quick chat and they're off building. [41:07] >> I guess because the agents can just [41:09] build the the artifact is just whatever [41:10] prototype or whatever they come up with, [41:11] right? But you can play with with it. [41:13] >> Exactly. Exactly. For us, it's like chat [41:16] is not chat like talking about the [41:18] feature and we've we felt like this on [41:20] product teams forever and like great [41:21] engineers and designers forever would [41:23] always say can we stop talking about the [41:26] work and do the actual work. [41:28] >> Yeah. [41:28] >> You know, and so now I think that ratio [41:31] is shifted even more. You're just doing [41:33] the work and you know, maybe it used to [41:35] be like 30% talk about the work, 70% do [41:38] the work or some other ratio depending [41:39] on the company. Now it's like 5% talk [41:41] about the work, 95% do the work. And and [41:44] to your point, that's exactly how it [41:46] feels. [41:46] >> I I think that's a great feeling, right? [41:48] To actually do the work. [laughter] [41:50] >> If you're a maker, if you're a designer, [41:51] if you're an engineer, if you're, you [41:52] know, product manager like myself. [41:55] >> Yeah. [41:55] >> Who also gets to ship to production by [41:57] the way. That's why we're here. We want [41:59] to make the product. But you probably as [42:00] since the engineers are shipping end to [42:02] end to a certain extent you probably [42:04] spend a lot of time with customer [42:05] conversations or like you know bank [42:06] conversations and other stuff maybe not [42:09] shipping a feature right like yes that's [42:11] my guess yeah we'll do I mean depending [42:13] on the thing lots of partner [42:15] conversations depending on the area the [42:17] product and what we're doing especially [42:18] if it's assisted or powered by a partner [42:20] >> we'll do a lot of you know actual more [42:23] traditional UXR to understand how people [42:25] are using the features what's good [42:27] what's bad qualitative quantitative [42:29] that'll all come in and then just we [42:32] know we're always listening. I mean you [42:33] see this with the folks open. We're [42:35] always listening what people are saying [42:36] out there about the product about the [42:38] feature you know there's no voice we [42:40] don't want to hear to incorporate into [42:42] the product to make the product better. [42:44] >> Got it. So what what what's next for [42:46] chat finances man? Maybe we don't have [42:47] to share too much but like what is your [42:48] vision for where this can go like six [42:50] months or like a year from now you know? [42:52] >> Yeah I mean I can't share anything about [42:54] what what is coming. I don't have [42:55] anything to announce today. [42:57] >> Yeah. But high level, you know, the goal [43:00] for personal finance is to make it so [43:02] chat GPT can save you time, money, and [43:05] stress when it comes to your money. And [43:07] even what we have today is really good [43:10] at doing that. And so a lot of our focus [43:12] right now is making the features easier [43:15] to use, easier to find, easier to [43:18] discover, and getting the word out about [43:20] them. And again, improving the product, [43:23] polishing the product. It's a financial [43:24] experience. We want it to be perfect. [43:26] It's really good. good. It's not perfect [43:27] yet. There's going to be continued work [43:28] to make it perfect on every dimension. [43:31] And as we continue to do that, then [43:33] we'll start thinking about, you know, [43:34] what other areas of someone's financial [43:36] life should we be helping with? Should [43:37] we be making easier to do, faster to do, [43:39] less effort, save them money, uh, and [43:42] such? But a lot of our focus right now [43:43] is, yeah, that product that's out that [43:45] we're seeing a lot of great traction on, [43:47] making it even better and getting it [43:49] more aware that it even exists. [43:51] >> Yeah, I think that's key, man. I think [43:52] the onboarding is like super important [43:54] cuz like you know we just walk through [43:55] like six really cool use cases but that [43:57] require like us knowing to submit the [43:59] prompt and like tell to build a site and [44:01] all kind of stuff. I think most nor most [44:03] people just don't get the stuff so you [44:04] have to like onboard them. [44:06] >> Yeah. [44:06] >> Exactly. And they shouldn't it should be [44:08] so easy you know you just turn it on [44:10] easy way to turn it on and then it just [44:11] works for you. [44:12] >> Yeah. [44:13] >> That that's that's the dream. And I [44:15] think there's like a lot of potential [44:16] with the voice feature too where you can [44:17] get like, you know, like people have [44:18] their chief of staff coordinate all the [44:20] other stuff, but I can imagine like a [44:22] financial advisor coordinate all my [44:23] other stuff too. [44:25] >> Totally. Totally. Yeah. And I think Addy [44:27] has shared publicly, you know, they're [44:28] working, voice team is working on [44:30] plugins and finance is one of our [44:32] plugins and so we're excited about that. [44:33] >> Awesome, dude. Well, I think between uh [44:35] finances and health, I think there's so [44:37] much more that Chad can help with than [44:39] just like trying to manage your emails [44:40] and stuff. So really excited about the [44:42] future. [44:42] >> Me too. Me too. [44:43] >> All right, Ethan. Well, thanks so much [44:44] for your time, man. And if people want [44:45] to give you feedback, where can people [44:47] find you online? [44:48] >> You can find me at eblock. Eb oc onx is [44:51] probably the best place and where I'm [44:53] most active. Feel free [44:54] >> to any and anything you have feedback, [44:56] product ideas. Uh I love hearing from [44:58] people using the product and want to [45:00] hear from everyone. [45:01] >> Cool. And uh thanks for making this your [45:02] life life's work. I I think you're [45:04] really going to transform uh millions of [45:05] people's lives. So, thanks so much. [45:08] >> Yeah. Thanks so much. Thanks for having [45:10] me.