AI can do the number-crunching here, but the roadmap needs a human hand.

The analytical structure — trend analysis, margin progression, top line items — is repeatable. But the roadmap component requires fresh contextual judgment each time based on business-specific constraints, making it only partially templatable.

Identifying spending trends and top margin consumers has reasonably crisp success criteria. 'Cost optimization without cutting staff or corners' is underspecified — what counts as a corner? What's the acceptable trade-off? These gaps leave room for the agent to produce a plausible-sounding but wrong roadmap.

The agent needs three years of structured P&L data, which must be explicitly provided — it won't have access to internal accounting systems by default. If the data is clean and formatted, analysis is tractable; if it's messy or requires reconciliation, the agent will struggle.

Misidentifying the true margin drivers or recommending cost cuts in the wrong areas could lead to real operational or financial harm. At $1.2M–$1.8M revenue, a bad 90-day cost plan could meaningfully damage the business, and errors may not surface until damage is done.

The roadmap requires understanding vendor leverage, team morale, delivery quality thresholds, and strategic priorities — none of which the agent can infer from P&L data alone. A finance professional or operator needs to validate and own the recommendations.