The Agent Is the New Service Provider in Finance

Why the next great finance company won't sell software. It will sell the work.

Sequoia recently published a piece arguing that the next trillion-dollar company will look like a services firm powered by AI, not a software company. Their core insight: for every dollar spent on software, six are spent on the people making that software useful. The real opportunity is not building better tools. It is replacing the service layer entirely.

They are right. And finance is where this plays out first.

The Six-to-One Ratio Lives in Every Finance Department

Every business that manages cash flow knows this ratio intuitively, even if they have never quantified it.

They pay for accounting software. Then they pay a bookkeeper to operate it. They pay for a forecasting tool. Then they pay a fractional CFO to build the model. They pay for a banking platform. Then they pay someone to log in every morning, pull the balances, and paste them into a spreadsheet.

The software was never the expensive part. The person sitting between the data and the decision was. And that person was not doing strategy. They were doing execution: pulling, formatting, reconciling, rebuilding. Week after week. The same workflows. The same manual steps. The same stale outputs.

That execution layer is the actual product most finance teams are buying. The software is just the thing that makes the execution possible.

Copilots Had a Ceiling. Autopilots Don't.

The first generation of AI in finance followed the copilot playbook. Put the tool in the hands of the CFO. Let them prompt it. Let them decide what to do with the output. The professional remains in the seat. The AI is leverage.

That model helped. But it did not transform. A better copilot still requires someone to drive it. The CFO still needs to know what to ask for, how to interpret the result, and when the output is wrong. The bottleneck moved from the spreadsheet to the operator.

The shift happening now is different. Businesses are starting to buy outcomes, not tools. Not a dashboard that shows cash position, but the cash position itself, updated, reconciled, and ready. Not a reporting template, but the report, built from live data, formatted, accurate, delivered before anyone asked for it.

This is the autopilot model. The AI does not assist with the work. It absorbs the work. The 13-week forecast that a fractional CFO rebuilds every month. The cash positioning view that requires logging into three bank portals. The variance analysis that takes half a day to produce. All of it becomes something the system handles, continuously and automatically.

Why Finance Outsourcing Makes This Inevitable

Autopilots win fastest in domains where the work is already outsourced. When a business is already paying an external firm for the outcome, the substitution is clean. No organizational change. No workflow redesign. Just a better vendor.

Finance execution is one of the most outsourced functions in business. Bookkeeping. Reporting packages. Reconciliation. Audit prep. Tax filing. These are well-scoped, repeatable deliverables that businesses already purchase externally. The budget already exists. The buyer already accepts that this work can be done outside the team. And the performance bar is already defined by the current provider.

That is why the agent model lands so naturally here. Bringing an AI-native platform into a role that an outsourced service was filling is not a transformation project. It is a procurement decision. And unlike a consulting engagement, the agent compounds. Every cycle it runs makes the next forecast sharper. Every pattern it observes deepens its understanding of how that specific business moves money.

The Intelligence Ratio in Finance Is Higher Than People Admit

Sequoia distinguishes between intelligence work and judgment work. Intelligence is rule-based, structured, and repeatable. Judgment requires experience, taste, and instinct. AI has crossed the threshold for intelligence. Judgment is still human.

Finance teams believe most of what they do is judgment. They are wrong.

Reconciling actuals is intelligence. Updating a forecast model is intelligence. Monitoring cash positions across accounts is intelligence. Flagging a deficit before it arrives is intelligence. Building a board-ready report from raw transaction data is intelligence.

The judgment calls, whether to raise capital, whether a cash gap is structural, whether the business can absorb the next hire, represent a small fraction of the actual hours. Everything else is a workflow. And cash flow, more than any other finance function, is almost entirely workflows.

That is why agents win here first. Not because the stakes are low. Because the intelligence ratio is high.

What Changes When the Service Layer Becomes Software

When agents absorb the execution layer of finance, three things happen that were not possible before.

First, the cost of financial intelligence drops to near zero. A 20-person company gets the same reporting depth as a Fortune 500 treasury department. Not because they hired better, but because the work itself became infrastructure.

Second, finance becomes real-time. The monthly reporting cycle existed because humans needed time to build it. When the agent builds continuously, the concept of a reporting period starts to dissolve. The numbers are always current. The forecast is always updated. The position is always known.

Third, the CFO's role fundamentally changes. Not eliminated. Elevated. The CFO who used to spend Monday mornings pulling data now spends them making decisions. The strategic work that was always supposed to be the job finally becomes the job.

This Is Not About Technology. It Is About Who Does the Work.

The conversation around AI in finance has been dominated by features. Better dashboards. Smarter alerts. Natural language queries. These are copilot improvements. They matter, but they are incremental.

The real shift is structural. It is about whether the work gets done by a person or by a system. Whether the business buys a tool and hires someone to operate it, or buys the outcome directly.

That distinction changes what every finance professional gets to do with their time. The controller who spent half their week reconciling actuals can spend it reviewing them. The bookkeeper who rebuilt the same report every month can focus on what the numbers actually mean. The CFO who produced the forecast can spend their Monday acting on it instead.

When the execution layer becomes infrastructure, financial intelligence stops being a function and starts being a capability available to every business, at every size, continuously.

That is the version of finance work that was always supposed to exist. We are building the infrastructure that finally makes it possible.