I already know you hate taxes
with Cy, Deduction.com
I already know you hate taxes
Show Notes
Cy is a British founder who came to the US on an O-1 Extraordinary Ability visa - earned through Caribou, a "Skype for kids" app acquired by Mattel and featured in Apple's iPad Air 2 keynote, generating 200,000 downloads in a single day. He spent seven years as Chief Product Officer at Latch, joining as employee #20 and helping grow it to 500 people and a $1.5B valuation before going public via SPAC in 2021. He recently became a US citizen. Now he runs Deduction.com , an AI tax agent for consumers who want something better than TurboTax DIY but without waiting four weeks for a reply from an overworked accountant.
Deduction's agent - Taylor - operates entirely via email (taylor@deduction.com). No app, no portal to learn, no new interface. You email the agent the way you emailed your last accountant. A human CPA reviews and signs your return. The price is $499 and Cy is personally onboarding every early customer.
The Pre-Digital Market Everyone Overlooked
55% of Americans choose a personal accountant over TurboTax. They collectively spend $16 billion per year on those accountants. And they wait weeks for responses, pay $1,000 or more, and get nothing digitally enabled in return. Cy's observation: this market is not just pre-AI. It is pre-digital. It has been operating on the same trust-based, paper-driven, human-bottlenecked model for decades - and the people working in it are fine with that, because 43% of CPAs turn away work and still cannot keep up.
This is what Cy calls the "dragon in the cave sitting on gold coins" dynamic. Accountants have leverage not because they are good at customer experience - they are objectively bad at it - but because clients are afraid of making a tax mistake. Fear is the moat. And it is a moat that AI can dismantle without the client even realizing how unnecessary their anxiety was.
Perceived Complexity Arbitrage
The most important insight in this episode: the gap between how complex consumers believe their taxes are and how complex they actually are is the arbitrage CPAs use to justify high fees, slow service, and enormous client anxiety. Cy calls it "perceived complexity." Most people's tax situations are more routine than they feel. The intricacy is not in the math - it is in the emotional weight the tax code has accumulated through decades of incomprehensible language and IRS-induced fear.
Deduction's value proposition is not just speed or price. It is leveling this information asymmetry. When the agent asks you questions and comes back with a clear picture of what your return actually requires, the perceived complexity collapses. The $499 product is not competing with TurboTax's interface. It is competing with the anxiety that sends people to $1,200 accountants they do not trust and cannot reach.
Meet Consumers Where They Are
The product decision that defines Deduction's interface: do not build a new UX. Figure out where your customers already are and operate there. How did you communicate with your last accountant? Email. How did you find your accountant? A friend or family member recommended them. How did you introduce them? Email introduction.
So Deduction operates in email. You email Taylor. Taylor BCCs your contact when someone refers you. There is no app to download, no portal to learn, no account to create before you can get started. The agent meets people in the medium they already associate with professional communication. Adoption friction is nearly zero because the interaction model is already familiar.
B2B Easy to Start, Hard to Finish - Consumer Opposite
Cy's framework for the B2B vs. B2C decision: B2B is easy to start and hard to finish. You can get design partners quickly. You can close your first $50K relatively fast. But selling to thousands of businesses is brutally difficult, and with 80%+ of Y Combinator companies now B2B, the competition for attention from enterprise buyers is intense. It is no longer the contrarian bet - it is the orthodoxy.
Consumer is the opposite. Hard to start - distribution is expensive, acquisition takes time, and getting the first 1,000 customers is a grind. But the market dynamics are radically different at scale. 60% of the S&P 500 by market cap are consumer or consumer-adjacent companies. The biggest outcomes in business history are predominantly consumer. The intellectual argument for B2B is sound. The historical data favors consumer if you can survive long enough to reach scale.
AI Should Be Deflationary
Cy's philosophical thesis for why Deduction exists: most AI today is being used to make businesses more efficient - cutting customer support staff, automating internal processes, increasing output per employee. The efficiency gains flow to business owners and shareholders. Almost none of it flows to consumers as lower prices or better service.
His conviction: AI should also be deflationary for the people it serves. The best AI products will not just make businesses more efficient - they will make previously expensive or inaccessible services available to people who could not afford them before. A $1,200 accountant becomes a $499 AI agent. The quality goes up. The price goes down. The access expands. That is the direction the technology should push - and most B2B AI is not pushing it.
Human Touchpoints in Agentic Products
Cy's counterintuitive operational bet: the more AI products people interact with, the more they will value arbitrary human touchpoints. He does personal 15-minute onboarding calls with every customer - not to talk about taxes, but to give them a face and a story behind the product. A woman in Santa Fe texted his personal cell, they spoke for 15 minutes, she signed up. She did not text a robot. She texted a person.
His analogy to medicine is sharp: the ideal future of AI-assisted healthcare is not an AI that replaces doctors but an AI that handles the diagnostic labor so the 13 minutes with the doctor are spent understanding the patient as a person, not reading test results. The same structure applies to tax: Taylor handles the technical work; Cy (and eventually a customer success team) handles the trust layer. They are not competing functions - they are complementary.
How to Get a Great Domain Before You're Famous
Deduction.com was acquired for under $50,000. The two rules Cy shares: start before anyone knows who you are (the moment you raise a Series A, domain sellers see your valuation and price accordingly), and use a specialized broker - he used Snagged.com . Most .com and .ai domains are owned by two or three companies, and brokers have the relationships and negotiating leverage that cold outreach cannot replicate.
His philosophical case for .com over .ai: ".ai" sounds like a test product or a sandbox. In two years, AI will be assumed to be inside every product - the same way nobody put "internet-enabled" in their company name in 2010. The .com signals permanence, prestige, and the expectation of long-term existence. For a consumer product competing on trust, that signal matters enormously.
Decks Get Checks
Cy's motto on pitch decks: decks get checks. Before starting Deduction, he created 10–15 decks for different ideas and shared them with friends, watching for reactions. One of those decks was compelling enough that a friend picked up the idea and started the company. The deck is not just a fundraising tool - it is a thought-clarifying, idea-testing, relationship-opening mechanism.
His three deck rules: follow the standard structure (team, market, problem - do not reinvent the format), invest in a real designer (not Fiverr), and treat the brand loosely at first (logo in a few hours, one revision from the designer). The insight from Latch's founder that stuck with him: "render to reality." If you can make something feel real - give it form, weight, and visual existence - it starts to manifest.
- B2B Easy to Start, Hard to Finish - B2B is easy to start (fast design partners, quick first revenue) but hard to finish (selling to thousands of businesses at scale is brutal, especially now that 80%+ of YC companies are B2B). Consumer is hard to start and easier to finish - distribution friction is high early, but market size and best historical outcomes favor consumer at scale. See Frameworks.
- AI Should Be Deflationary - Most AI today captures efficiency for businesses, not consumers. The underbuilt opportunity is AI that makes previously expensive or inaccessible services cheaper and better for the people who use them. Products that lower the price of professional services while improving quality are competing in an almost uncontested market. See Frameworks.
- Meet Consumers Where They Are - Do not build a new interface and ask customers to learn it. Identify the medium they already use to interact with this type of service (email, text, call) and build the product to operate there. Lower adoption friction = higher trust = faster growth. See Frameworks.
- Deduction.com - AI tax agent (Taylor) for consumers and founders. Email taylor@deduction.com with your documents - the agent works through your return, flags questions, and a licensed CPA reviews and signs the final filing. $499/year. Cy personally onboards every early customer. The next-generation H&R Block.
- Snagged.com - Domain brokerage Cy used to acquire deduction.com for under $50K. Specializes in premium domain acquisition. Most valuable .com and .ai domains are held by a small number of portfolio owners - a broker with existing relationships can get deals that cold outreach cannot.
- O-1 Visa (Extraordinary Ability) - A US visa originally designed for Nobel Prize-level talent, broadened to include entrepreneurs, athletes, and others with documented extraordinary achievement. Cy qualified through the Caribou acquisition by Mattel and Apple keynote feature. The irony: being rejected from every graduate job in the UK made him eligible - and without that rejection, he would never have been. See Glossary.
- Perceived Complexity Arbitrage - The gap between how complex consumers believe their tax situation to be and how complex it actually is. This gap is the primary mechanism CPAs use to justify high fees and slow turnaround - not maliciously, but structurally. AI that collapses perceived complexity can disrupt incumbents without needing to outcompete them technically. See Glossary.
- Pre-Digital Market - An industry that has not meaningfully adopted digital tools or workflows - beyond having a website or email. Tax preparation serving individuals is characterized as pre-digital, not just pre-AI. Pre-digital markets require less displacement of existing technology and more replacement of human processes. See Glossary.
- Agentic Professional Services - Professional services (accounting, legal, medical, financial advice) delivered primarily through AI agents with human review and verification touchpoints, rather than direct human labor throughout. Deduction is the canonical example: AI agent handles the work, human CPA reviews and signs. See Glossary.
- Render to Reality - The practice of creating detailed visual or written representations of a future state - decks, prototypes, mockups - to make a vision feel real enough to act on, share, and fund. From Latch CEO's philosophy: if you can render something and make it feel real, it begins to manifest. The mechanism behind "decks get checks." See Glossary.
- SPAC (Special Purpose Acquisition Company) - A blank-check shell company created to take a private company public without a traditional IPO, by merging with it after the SPAC raises funds on a public exchange. Latch went public via SPAC in 2021 during the SPAC boom, at a $1.5B valuation. Chamath Palihapitiya led the deal. See Glossary.
- Pay Per Transaction - A pricing model in which users pay per use or per action rather than a recurring subscription. Cy advocates for more AI tools offering this model - many tools would see higher adoption if users could pay $1–5 for a single use rather than committing to $20/month subscriptions they'll feel guilty about not maximizing. See Glossary.
What makes tax preparation the right market for an AI consumer product?
Three things converging: 55% of Americans pay a personal accountant (not TurboTax), spending $16 billion per year - and waiting weeks for responses, paying $1,000 or more, for a service that is pre-digital, not just pre-AI. The incumbents (TurboTax, H&R Block) own the DIY end. The high end is owned by overworked CPAs who turn away 43% of potential clients. The middle - people who want expert guidance without the friction and cost - is almost unoccupied. And the labor shortage on the CPA side means the existing providers are actually welcoming the technology rather than fighting it.
What is perceived complexity arbitrage, and how does Deduction exploit it?
Perceived complexity is the gap between how complex consumers believe their taxes are and how complex they actually are. CPAs have always profited from this gap - the more intimidating taxes feel, the more justified the high fees and the more captive the client relationship. Most people's tax situations are more routine than they feel. Deduction's AI agent asks questions, reviews documents, and comes back with a clear picture of what the return actually requires. When the perceived complexity collapses, the $1,200 accountant becomes replaceable by a $499 agent.
Why does Deduction operate through email rather than building an app?
Because that's where people already communicate with their accountants. How did you reach your last accountant? Email. How did you find them? A referral from a friend or family member - typically delivered as an email introduction. Deduction meets people in the medium they already associate with professional service. No app to download, no portal to learn, no new account to create before getting started. The agent even supports email introductions - you CC a friend, Taylor responds to both of you. Adoption friction is nearly zero because the interaction model is already familiar.
What is Cy's framework for deciding between B2B and B2C?
B2B is easy to start, hard to finish. Design partners are accessible, first revenue comes relatively quickly, and you can reach $2–3M without massive distribution. But selling to thousands of businesses is brutal - long sales cycles, high CAC, and now 80% of Y Combinator companies competing for the same enterprise buyers. Consumer is hard to start and easier to finish. Distribution costs are high early, but 60% of S&P 500 market cap is consumer or consumer-adjacent. The biggest outcomes in business history favor consumer. His advice: start B2B only if you genuinely love the customer, not because it's intellectually correct. Start consumer if you're willing to grind through the early distribution problem.
Why should AI be deflationary for consumers?
Because most AI today is capturing efficiency gains for businesses - cutting labor costs, increasing output per employee, improving margins - without passing any of those gains to the people being served. The more important and underbuilt use of AI is to make previously expensive or inaccessible services cheaper and better for end users. A $1,200 accountant becomes a $499 AI agent. A $10,000 lawyer becomes a $200 AI contract reviewer. The technology should be compressing the price of professional expertise, not just the headcount required to deliver it.
How did Cy acquire deduction.com for under $50K?
Two rules: start before anyone knows who you are, and use a broker. The moment you raise a Series A and your valuation is public, domain sellers will price accordingly. As a nobody with nothing on your LinkedIn, you are negotiating at the lowest possible moment of perceived leverage. And most premium .com and .ai domains are held by two or three portfolio companies - cold outreach gets you nowhere. A specialized broker (Cy used Snagged.com / Rob) has existing relationships with those portfolio holders and can negotiate deals that aren't available through direct contact.
What is 'render to reality' and why does it matter for founders?
It's the practice of creating detailed visual representations of a future state - decks, prototypes, mockups - to make a vision feel real enough to act on, fund, and share. Cy learned it from Latch's founder, who would painstakingly create beautiful decks for every major fundraise, customer pitch, and all-hands. The principle: if you can render something and make it feel real, it begins to manifest. It's the mechanism behind 'decks get checks.' A rough Apple Notes idea is just a thought. A designed deck with a market slide and a team page is something a person can react to, fund, or steal - and all three are useful outcomes.