Beyond the Pilot: What Happens When AI Meets HubSpot's General Ledger?

A conversation with Blake Winchester, Director of Accounting at HubSpot, on the moment AI stopped being a chatbot and started living inside the close.

Most AI conversations in accounting still circle the same drain: "Will it replace my job?" and "Can it summarize this email?" Blake Winchester has moved past both questions. As Director of Accounting at HubSpot (NYSE: HUBS) — a company that, by his account, genuinely walks the AI walk — Blake joined Roshan Chekuri at FloQast's Accounting Next to describe what it actually looks like when AI agents stop being experiments and start running accruals.

The answer is less futuristic than you might think, and far more practical.

Blake’s accounting team got serious about their AI journey in 2025 and began considering their options, including FloQast Transform. Transform enables accounting teams to build auditable AI agents that automate routine tasks such as accruals, journal entries, and reconciliations — all without IT support. After the HubSpot team saw a proof of concept for what Transform could do, they bet on FloQast. 

From Chatbot to Agent: The Real Shift

The popular mental model of "AI in accounting" still defaults to a chat window — ask a question, get a paragraph. Blake draws a much sharper line. The real difference is that FloQast AI Agents aren’t producing paragraphs; they’re putting out repeatable Python code. And because the code is auditable, every step of the process can be reviewed, challenged, and improved by the accountants who own the work.

"When you get under the hood, FloQast Transform AI agents are writing Python code that's repeatable. It's auditable."

Blake Winchester, Director of Accounting, HubSpot

This distinction matters enormously for enterprise-grade governance. A chatbot produces text. An agent produces a workflow — one that can be versioned, tested, and handed off. This is the infrastructure accounting teams have always needed but never had easy access to.

Watch the On-Demand Replay

Accounting AI in Practice: The HubSpot Story

HubSpot's accounting department got serious about its AI journey in earnest last year. The first thing Blake's team did was build a shared understanding of what the work actually involved. Before any agent could be deployed, the team spent significant time inside their own reconciliation workbooks asking a deceptively hard question: exactly what do you need the agent to give you?

The first agent to go live tackled accruals — a proof of concept that forced the team to break down a process they'd done manually for years into precise, transferable steps. That exercise alone, Blake notes, was transformative for the team's understanding of their own work.

The team also ran a hackathon — not an IT exercise, but an accounting exercise. People showed up with build ideas rooted in their own day-to-day challenges. The team left with working prototypes and, more importantly, the conviction that building agents was something they could do.

"The great ideas are coming from the [accounting] team. We asked them, ‘What can AI do to move the needle from day to day?’ We've enabled the team to become subject matter experts. They know how to build Agents that improve their workflows and meet their needs."

Blake Winchester

HubSpot's model is deliberately hybrid: FloQast’s Forward-Deployed Accountants build some agents, Blake's team builds others. The rationale is simple — the great ideas are coming from the people closest to the work. Centralizing all agent development in a vendor or an IT team would miss most of the value.

The Governance Question: Accuracy, Repeatability, Auditability

For finance and accounting leaders who've watched AI hype cycles come and go, the skepticism often boils down to one concern: can I trust it? Blake's answer is structural. Because agents produce code rather than narrative, every decision in the workflow is traceable. The accruals agent doesn't "figure out" the right answer — it runs the same deterministic logic every month. If the logic is wrong, you can find and fix it. If regulators or auditors ask, you can show your work.

This is precisely why the "build vs. buy" conversation at HubSpot resolved in favor of FloQast's platform. Blake's transformation team could build the tooling from scratch, but doing so would consume months of accountant time and require IT to manage a custom infrastructure. The opportunity cost — of that time, of that distraction — isn't worth it when a purpose-built platform already exists, HubSpot found.

Watch the On-Demand Replay

What the Accounting Department Looks Like in a Few Years

Blake is careful not to promise a revolution. The specialties aren't going away. Tax accountants will still do taxes. Revenue accountants will still do revenue. What changes is the texture of the work — less assembly, more analysis; fewer manual steps, more judgment calls that actually require human expertise.

His forecast: quicker closes, deeper analyses, and a generation of accountants who list agent-building on their résumés the same way their predecessors listed Excel modeling. "AI is the new preparer," he says — meaning proficiency with AI workflows is no longer a differentiator, it's a baseline expectation.

Three Tips From the Trenches

When asked what advice Blake has for accounting decision makers considering AI, he gave three pieces of wisdom:

1. Start now — nobody is behind. Every CFO is about to ask about your AI strategy. Get your feet wet before they do. "No one is behind; everyone is figuring it out."
2. Interrogate your own processes before you automate them. Spend time in your workbooks. "What exactly do you need the agent to give you?" That question cuts more steps than any technology will.
3. Leverage your network to find out how they are using AI. "What agents did you build?" is a more useful question than "Which vendor did you choose?"

The clearest signal that HubSpot's AI journey has moved beyond pilot stage? Blake's hiring filter has changed. He no longer asks candidates if they use AI. He asks whether they've built an agent to automate a process — and he expects an answer.

Want to see AI agents in your close? Explore FloQast Transform.