# Jeff Epstein

Jeff Epstein is an investor at BVP and a former CFO who writes about how finance teams can use AI, measure its business impact, and set controls for AI agents.

- Profile: https://b2binfluencer.marketing/creators/jeff-epstein/
- LinkedIn: https://www.linkedin.com/in/jeffepstein1
- Topics: artificial-intelligence, ai-in-finance
- Based on 12 dated posts, latest 2026-10-07

## Summary

Jeff Epstein helps finance leaders test what AI changed in finance work and what controls it needs.

- Best for: B2B finance software campaign for CFOs evaluating AI in finance operations.
- Not for: Consumer or small-business accounting apps seeking product demos, guaranteed attendance, or paid media placement.
- Check first: Ask what paid partnership terms are available and whether a paid partnership route exists.

### Takeaways

- A faster close does not necessarily mean a faster business.
- Define what AI agents may do before naming their permissions.
- The real cost of AI in finance is time people spend checking it.

## Main ideas

- **"How are you using AI" has been replaced by "what actually changed"**: Epstein argues that a couple of years ago, the finance question was mostly how a team was using AI; now it is what actually changed as a result. The test they set out is deliberately financial: did the team get more productive, did the economics change, are decisions faster or better, and can any of it be measured? The processes they name are forecasting, reporting, close, collections and analysis; in a later post, the same test extends to margins, pricing and retention. They describe this from both sides of the table: proving technology investments were worth what was paid for them as a CFO, and now asking founders to do the same as an investor.
- **Define what agents may do before naming their permissions**: Epstein's governance argument starts with a gap: finance controls were designed around people — someone approves a payment, someone owns the close, someone signs off on the numbers — but no equivalent exists for software agents that touch cash, close, collections, reporting, procurement or pricing and make decisions. Their position is not to slow or stop AI, but to set boundaries first: who owns the decision, what agents may act on, and what must be stopped and escalated to a person. In a related post, they make the same point: finance has to help decide where machines can act on their own and where human judgment still matters.
- **A faster close is not a faster operating model**: Epstein stacks automation gains against the state of finance's underlying data and controls to argue that finance is getting better at producing information without changing how the business decides. If the numbers are ready Tuesday instead of Friday but pricing decisions still wait for the monthly meeting, or the close is automated while the board cadence stays the same, the old operating model has simply been sped up. The question they say they are most interested in is what happens when finance stops being a function that reports and starts operating proactively.
- **The real cost of AI in finance is the time people spend checking it**: Epstein's most concrete operational point is about rework. They name where AI is showing up — billing, reconciliation, forecasting, reporting, close, collections and controls — then describe the awkward position finance teams are in: asked to help lead the company's AI conversation while still working out where AI works inside finance. The questions they raise are practical: what people are actually using, what is working, what is not, where a person is still needed, and where one can finally be taken out of the process. They also say directly that there is a limit to what can be learned from vendor demos and LinkedIn posts.
- **Peer conversation is how finance leaders should evaluate AI**: Every post is built around the same in-person event, Finance Exchange in San Francisco on October 15. Epstein says finance leaders need to sit down with each other and compare what is actually happening, which demo-heavy formats cannot deliver. The posts promote the event with an application or registration link and report demand — 200+ finance leaders registered in one, oversubscribed in another. The tone is informal alongside the numbers, including the sign-off, "let's go to the moon with finance."

## Work to read

- [“What actually changed?” — moving AI from demo to finance question](https://www.linkedin.com/posts/jeffepstein1_financeexchange-cfo-ai-activity-7498787696719495169-JOPH)
- [A changed mind on the CFO’s role in AI](https://www.linkedin.com/posts/jeffepstein1_zenskar-finance-exchange-for-finance-leaders-activity-7508201055298441216-Oy5S)
- [Who is accountable when an AI agent makes a mistake overnight](https://www.linkedin.com/posts/jeffepstein1_zenskar-finance-exchange-for-finance-leaders-activity-7503127608872607744-4Qxi)
- ["We've just made the old one faster" — the limits of a quicker close](https://www.linkedin.com/posts/jeffepstein1_financeexchange-cfo-ai-activity-7511105471110451201-6OXo)
- [What finance teams are actually using AI for, and what it costs them](https://www.linkedin.com/posts/jeffepstein1_zenskar-finance-exchange-for-finance-leaders-activity-7513657534595997696-x35R)

## Brand relationships

- **Oracle** (employment)
- **BVP** (employment)
- **Zenskar** (disclosed_partnership)

## Partnerships

The posts show Jeff Epstein appearing at the Zenskar Finance Exchange in San Francisco on 15 October, a finance-leader event they promote alongside co-presenter Priya Rajan. Access is by application or invite rather than open purchase: by early October, the organizers said the event was oversubscribed, and the posts point readers to apply to join or request an invite. The audience is finance leaders — one post cites 200+ registered — and the stated aim is peer comparison rather than vendor pitches. The posts contain no sponsorship policy, rate card, media kit, or partnership inquiry page; ask the organizers for those details.

## Alternatives

- Ahikam Kaufman: Choose Kaufman when a campaign needs someone who speaks from inside a finance-AI platform rather than from the buyer's chair. Both posts take the vendor's side of the same question Epstein asks, and go deeper on operating economics than Epstein does: what a prompt, an agent action and a context window cost; how usage scales when autonomous agents run thousands of actions; and which model should run each process. The second post pushes further into switching risk: whether deployment forces a system replacement, where the financial rules and process knowledge live, whether supporting documentation can be accessed independently, and what happens when an entity is added or an ERP changes. It cites a Deloitte finding that 84% of finance leaders expect technology sovereignty to affect investment decisions through FY2027. For a sponsor whose product is judged on integration, exit cost, permissions, approvals, segregation of duties, transaction-level traceability, audit trails and model governance, Kaufman covers ground Epstein's posts leave open.
- Anooshay Arif: Choose Arif when the buyer's decision is about how AI changes the work of accounting and finance-operations teams, not how CFOs govern it. The posts are concrete about where AI fits alongside people: in bank reconciliations, it flags anomalies while a human still reviews and approves each one; in AP and AR, pattern matching handles the obvious majority while the team keeps the messy remainder that needs judgment; in variance commentary, a machine-assisted first draft is refined by someone who knows the client's business. The stated goal is to flip the ratio most finance teams live with — roughly 80% of time producing numbers and 20% explaining them — so senior people can spend their hours on why revenue fell in a segment, what cash looks like in 90 days, and whether the pricing model still works. A second post covers a CFO's first 90 days, arguing that the real first task is alignment and trust in the numbers, not dashboards: clean dashboards that nobody believes get ignored. That is a useful angle for a sponsor selling to finance teams mid-transition or selling onboarding and adoption rather than pure AI capability.
- Bojan Radojicic: Choose Radojicic when a campaign targets finance teams that build rather than buy, or the technical half of the finance-IT partnership. One post follows a finance leader and small team building a custom reporting app with Claude Code, a Git and GitHub workflow, and a secure login. They handle month-end work that falls between existing systems: pulling trial balances, budgets and forecasts; building management-format statements; comparing actuals against budget, forecast and prior year; and drafting variance commentary. The post also lays out control rules a large vendor proposal usually skips: code calculates and AI explains; humans review and edit every comment before it reaches the CFO; MFA and role-based access; a full audit trail; and a developer review of the code and infrastructure before real data goes in. The stated timeline is 10-to-14 weeks. The other post audits an FP&A team on 70-hour weeks, with broken spreadsheets and three-day variance comments, and ends with a set of FP&A templates. For a sponsor selling enablement, templates, tooling or training alongside software, this is closer to the audience's actual work than a CFO-governance conversation. The posts describe owned media as the distribution route.

---
AI-assisted summary of public posts. Audience demographics are not available. Engagement appears only where we checked recent posts. Method: https://b2binfluencer.marketing/methodology/