This AI Spreadsheet Just Replaced Years of Excel Training | Quadratic
TL;DR
Quadratic’s pitch is “AI, code, and connections,” not just chat on top of a spreadsheet — Cole Stark says the product was built from the ground up to let users query data in natural language while also generating formulas, Python, and SQL inside a browser-based spreadsheet.
The personal finance demo showed why direct integrations matter more than fancy prompts — by connecting Plaid to 12,000+ financial institutions, Quadratic pulled balances and transactions from multiple accounts to auto-build a net worth dashboard with waterfall charts, merchant breakdowns, and recurring charge analysis.
The key trust argument is that AI explains, but the spreadsheet remains the source of truth — Stark’s point is that chat can summarize and recommend, but the real value is that formulas and Python are visible in the grid, inspectable, and recalculable instead of living in a black box.
Quadratic’s marketing use case is about stitching together fragmented growth data — Stark described joining Google Ads, Mixpanel, Stripe, and Postgres data to answer questions startups often can’t confidently answer, like which channels drive feature adoption, paying customers, and higher LTV users.
One concrete growth insight from Quadratic’s own team: templates beat blog posts on LTV — after analyzing their SEO traffic, Stark says users landing on Quadratic templates had better retention and higher long-term value than visitors coming through blog content, changing where the company invests effort.
Model quality has been a major unlock for spreadsheet AI usability — Stark says newer models, especially Claude Sonnet/Opus-class systems, have sharply improved tool use and code generation, to the point where users can be lazier with prompts and still get the right spreadsheet actions.
The Breakdown
Why spreadsheets are the next AI skill to collapse
The host opens with a blunt framing: Excel used to take years to get good at, and now the right prompt can compress that learning curve into seconds. He doesn’t say spreadsheet pros disappear tomorrow, but he does suggest tools like Quadratic could make the old “spreadsheet master” path a lot less defensible within months.
Cole Stark’s core pitch: built for AI from day one
Cole Stark, Quadratic’s head of growth, describes the product as a modern spreadsheet centered on three things: AI, code, and connections. His big distinction is that Quadratic isn’t “a chatbot built onto an existing spreadsheet” — it’s meant to understand the data already in the sheet, generate formulas or Python, and connect directly to databases and tools like Mixpanel, Stripe, Google Analytics, and QuickBooks.
Personal finance demo: from raw bank exports to a net worth dashboard
The first walkthrough uses sandboxed bank data connected through Plaid, which Stark says supports over 12,000 financial institutions and can sync updates daily. He contrasts this with tools like NerdWallet: good for a simple trend line, but not flexible enough for deeper insight. Quadratic ingests the ugly raw balances and transaction tables, then turns them into a waterfall chart, net-worth-by-institution view, monthly inflow/outflow chart, merchant analysis, and recurring subscription detection.
The real magic is not just charts — it’s asking follow-up questions
Once the dashboard is built, Stark asks Quadratic for “five concrete actions” based on top merchants, recurring charges, and credit utilization. He compares the workflow to a long-running chat where the model improves as it accumulates context, then makes an important caveat: chat alone is not trustworthy enough, but spreadsheets plus formulas and Python are. That line lands hard — AI gives readable explanations, but the grid is where the answer becomes something you can actually inspect and trust.
Marketing analytics: stopping startup teams from “flying blind”
The second demo is more ambitious: a growth dashboard that ties together top-of-funnel clicks, sign-ups, product usage, and revenue. Stark says that a year ago, Quadratic’s own marketing, Mixpanel, and Stripe data lived in separate silos, so nobody could confidently say which efforts created long-term value; now he uses AI-written SQL and Python to blend Postgres, Google Ads, and Mixpanel data into one sheet.
A memorable internal insight: templates outperform blog posts
The strongest anecdote in the marketing section is about SEO. Stark says Quadratic had put serious effort into blog content, but the spreadsheet analysis showed that template pages brought in users with higher retention and better LTV, pushing the team to invest more in the company’s library of a few hundred templates instead.
Better models changed what spreadsheet AI can actually do
When the conversation shifts to model quality, both host and guest sound a little amazed by the pace. Stark says newer models have gotten dramatically better at understanding spreadsheet context, choosing the right tool calls, and writing usable Python or formulas, so users no longer need intricate prompts — they can “sort of type what you’re thinking” and get useful results.
The adoption challenge: not convincing people AI is interesting, but that it’s safe to use
Stark closes on the go-to-market problem: most people already have a spreadsheet workflow, even if it’s ugly, so the hurdle isn’t novelty but trust and friction. His answer is that spreadsheets are the ideal AI surface because they’re familiar, visual, and inspectable — not just a black-box chat — and his advice for new users is simple: connect some data, ask questions, and let the first chart or answer be the inflection point.