Ask any auditor about their worst week and you will hear a version of the same story: a mountain of PDFs, a spreadsheet of selections, and a deadline that does not care how many pages a human can read before dawn. The work is essential - it is the reason financial statements can be trusted - and it is also, in large stretches, mechanical. Punchcard is a bet that the mechanical part no longer needs a human doing it by hand.
Founded in 2023 and part of Y Combinator's Summer 2023 batch, Punchcard - a company that spent its first stretch of life under the name Agentive - builds what it calls an assurance system of action. In plain terms: software that sends AI agents into the guts of an audit to read documents, extract the relevant fields, run comparisons, and produce a first draft of the workpaper, with citations pointing back to the source page. The human stays in the loop for the part machines are bad at - judgment - and hands off the part they are slow at.
The problem sits in the grunt work
Substantive testing is where audits get heavy. An associate pulls a sample of transactions, then goes line by line to prove each one is real and recorded correctly - matching an invoice to a shipping document to a bank statement. Multiply that across revenue, inventory, and a dozen other accounts and you have the profession's reputation for brutal busy seasons. The tooling, for the most part, has been spreadsheets and a document request list emailed to the client.
Punchcard's founders knew the pain from the inside. CEO Daniel Alberson started as an auditor and product manager at Ernst & Young after an accounting degree at BYU, then left a master's program at Northwestern to work as a product manager on augmented-reality glasses at Facebook. He came back to the industry most people in tech avoid on purpose.
How the work actually flows
The platform is organized around the shape of a real engagement rather than a generic chatbot bolted onto a file store. Four pieces do most of the work.
Request
Structured client document requests with instant feedback and zero training to upload.
CoAudit
Agents read documents, extract fields, and run the comparisons behind substantive testing.
Workpaper
First-draft workpapers for detail, control, and contract testing - with reasoning traces.
FS Review
TB-to-FS tie-outs, disclosure checklists, and cross-section consistency checks.
The design choice that matters most is the citation. In a regulated field, an answer without a source is worse than no answer at all - a reviewer cannot sign off on a number they cannot trace. Punchcard's agents are built to show their work, quoting the supporting section they relied on. That is less flashy than a magic "audit complete" button, and far more useful to a partner whose name goes on the opinion.
What the numbers are supposed to be
The company frames its value as throughput. It cites up to 10x efficiency on substantive testing and roughly 90% less review time. Those are vendor figures, not audited results, and worth reading as claims rather than facts - but the direction is the point. When an associate at JLK Rosenberger let the system scan 12,750 pages of offering memos and quote the supporting sections, the associate's own summary was telling: it let the team attempt work they would not have taken on by hand.
Where the hours are supposed to go (illustrative)
The people behind it
The founding team reads like a deliberately balanced hand. Alberson brings the domain scars and the product instinct. Nathan Turnbow, co-founder and CTO, built an early AI-powered education platform at Imagine Learning and worked as a cybersecurity engineer at Dell - handy pedigree for a company whose customers care a great deal about where their data goes. Jason Draper, the third co-founder, is a second-time founder who led teams as an officer in the US Air Force.
The company is small and works in person from Provo, a short drive from BYU's dense pipeline of accounting graduates. Its backers are not small: Y Combinator, Dell Technologies Capital, General Catalyst, Foundation Capital, Liquid 2 Ventures, and Ritual Capital have all put money in.
| Also known as | Agentive (rebranded to Punchcard) |
| Founded | 2023 · YC Summer 2023 batch |
| Headquarters | Provo, Utah, United States |
| Category | AI-native audit & assurance platform (B2B SaaS) |
| Named users | Richey May · JLK Rosenberger · Larson & Company · RSM Ebner Stolz · Stambaugh Ness |
| Security | SOC 2 certified · GDPR compliant |
Where it sits in the market
AI for accounting has become a crowded pitch, from document-testing tools like DataSnipper to anomaly platforms like MindBridge and lease-and-revenue specialists like Trullion. Incumbent suites from CaseWare, Thomson Reuters, and Wolters Kluwer still run most of the world's audits. Punchcard's wager is narrower and, in a way, more ambitious: not a feature bolted onto the old workflow, but the workflow itself rebuilt around agents that do a unit of audit work end to end and show the citation.
Whether that becomes the default way audits get done or one of several tools in a firm's kit is the open question. What is clear is the shape of the opportunity - a large, conservative profession, a workload nobody enjoys, and a technology finally good enough to read a document and explain what it found.