AI can scaffold the analysis, but a CFO still has to make the real calls here.

The structural framework—compare healthy vs. struggling cohorts across churn, CAC, payback, GTM, pricing, concentration—is repeatable. But each client's situation is unique enough that the agent must exercise fresh judgment every time, not just fill a template.

Success criteria are highly subjective: what counts as a 'good' intervention recommendation depends on client relationships, risk appetite, and strategic context the agent cannot evaluate. There is no crisp signal for when the work is done well.

The agent would need structured data exports for all 12 clients (financials, GTM details, product roadmaps, customer concentration tables) that almost certainly live in spreadsheets, CRMs, and the CFO's own notes—none of which are automatically accessible to an agent.

A wrong pricing or repositioning recommendation could accelerate churn or damage a client relationship; a misread on customer concentration could lead to bad retention bets. These are real business decisions with financial consequences that are hard to reverse.

The fractional CFO's value here is precisely the contextual judgment built over 18 months of client relationships—knowing which founders will act on advice, which markets are shifting, and which interventions are politically feasible. An agent lacks all of this.