Founders ask Collective 54 this 2 times in our records, 2 of them in 2026. The financial forecast and financial metrics answers on this site cover a forecast you can trust and what to track; this page covers using AI to build the model underneath them and keep it current.
The finance essay in the newer book argues that finance in boutique firms has mostly existed to record the business, not to run it. In the first era it was slow, expensive and built on manual work, so it was backward-looking by default. In the second era, software made closes faster and costs lower, but the essay says it made almost no difference to what finance actually delivered. Dashboards multiplied, and the presence of data was mistaken for understanding.
It names a deeper cost: founder financial illiteracy. Founders were rarely taught how gross margin should be calculated in a services business, what belongs in overhead versus sales and marketing, what EBITDA means, or how to work backward from the profit and loss statement and the general ledger to improve margins on purpose. As an inference, a model built by someone else and opened once a quarter does not fix that. A model the founder can question at any time, and that answers in plain terms, can.
The essay says most boutiques managed in units of effort rather than units of economics. They tracked hours, utilization and capacity, but not what those hours cost or returned, because attaching dollars to every unit of work was too slow and manual to be practical. AI removes that barrier. Its example: an analyst spending 25 hours on a task is not just a utilization statistic but a 2,500 dollar delivery cost, and once that cost is visible the questions follow. Should the task be automated, shifted to AI, moved offshore, handled by a more junior person, or is it exactly where senior expertise belongs?
The essay calls the capability behind this activity-based costing and says it informs how services are designed and priced, how deals are scoped, how teams are staffed, how capacity is planned and how cash is managed. The account management essay applies the same idea to clients, with account-level costing that shows which clients create or destroy EBITDA.
As an inference from that material, structure the model around the few drivers that actually move a boutique, so AI can update it from source data rather than someone retyping totals. The yield chapter of the 2020 book supplies the core relationship: yield is the average fee per hour times the average utilization rate, so a 400 dollar average fee at 75 percent utilization is a yield of 300 dollars per hour. Add fully burdened cost per hour by role, so every engagement carries its delivery cost. Add revenue by type, so contracted, recurring and new work stay separate. Add overhead by function. And add collections, because the cash flow chapter says boutiques run on cash rather than on net income or EBITDA, and offers a way to view it: cash flow per partner, equal to cash flow over fees, times fees per staff member, times staff per partner, which in one firm reviewed by Capital 54 came to a healthy 650,000 dollars.
The essay assigns roughly 80 percent of the finance function to AI: data ingestion and reconciliation, continuous reporting, variance detection, forward-looking projections, and pattern recognition across financial, operational and commercial data. It describes the AI living inside the business, continuously ingesting data from sales, delivery, payroll and cash, and producing standardized, decision-ready outputs rather than only historical summaries.
As an inference, in practice that means connecting the model to your accounting system, time records, payroll and CRM; having the AI draft the model structure and write down every assumption it makes; updating it each time new data arrives rather than once a quarter; and asking it plain questions, such as which clients lost money last quarter, or what happens to cash if a large client pays 30 days later.
The essay keeps roughly 20 percent with people: judgment, interpretation, benchmarking, strategic tradeoffs and accountability. It says that human layer is best delivered by an outsourced fractional finance firm, but only one specialized in boutique professional services, which quality-controls the AI output, applies benchmarks from comparable firms and adds judgment where tradeoffs and accountability matter. It says traditional fractional finance models are obsolete unless paired with AI, and that the question when choosing a provider is no longer whether they can keep the books but whether they specialize in firms like yours and can add strategic value on top of AI.
The essay lists beliefs founders held because they had nothing to compare against. Being paid in 45 days seemed acceptable, while peers were paid in advance and ran negative working capital. Compensation bands seemed market aligned while they were above peers. Closing the books in two weeks seemed efficient when it should take 24 hours. A 50 percent gross margin seemed strong when the best firms operate far above it, and a 25 percent EBITDA margin seemed impressive while it left money on the table.
As an inference, AI can tell you what your numbers are; it cannot tell you whether they are good unless someone supplies comparisons from firms like yours. That is the main reason the human layer matters.
As an inference, a model built by AI can be confidently wrong. Before relying on it, rebuild the last twelve months from the model and compare the result with your actual statements; differences usually point to a missing data source or a wrong assumption. Keep the assumptions on one page where a person signs off on them. Keep client and payroll data inside tools your firm has approved; the data protection answer on this site covers that.
The cash flow chapter of the 2020 book asks whether doubling the firm would make you run out of working capital, need a lot of short-term debt, develop a collections problem, see cash payments exceed cash income, or see payroll grow faster than receivables, and whether cash flow problems would be hidden by a lack of forward visibility. As an inference, those questions make a good first scenario: ask the model what doubling looks like, month by month, and see where it breaks.
Collective 54 names no finance software or AI tool, publishes no model template, and gives no accounting or tax advice. The published positions are finance reporting the past, speed and cost improving without quality, dashboards mistaken for understanding, founder financial illiteracy, effort rather than economics, the hours to dollars example and activity-based costing, account-level costing by client, yield as fee times utilization, cash flow per partner and the doubling questions, the roughly 80 and 20 division of finance work, a specialized fractional partner on top of AI, and the unbenchmarked beliefs.
If your books are not yet clean and current, as an inference, fix that first; a model built on unreliable records will be fast and wrong.
If you are preparing for a sale or raising capital, the model will be examined by people with their own analysts, so have a qualified adviser review it.
And if your firm is small and simple, a short driver model updated monthly may be all you need.
Use AI to build and keep the model current from your real data: accounting, time, payroll and pipeline. Structure it on the drivers of a boutique, including fees, utilization, fully burdened cost per hour, revenue by type, overhead and collections, so hours become dollars, as the finance essay describes. Let AI reconcile, report, flag variances and project forward, and ask it plain questions. Keep the assumptions, benchmarks and decisions with people, ideally a fractional finance partner who specializes in firms like yours. Check the model against the last twelve months before trusting it, and test it with the doubling questions from the 2020 book.
The finance essay says AI can handle ingestion, reconciliation, continuous reporting, variance detection and projections from your sales, delivery, payroll and cash data. As an inference, check its output against your actual books before relying on it.
Collective 54 publishes no template. As an inference, build it on fees, utilization, fully burdened cost per hour, revenue by type, overhead and collections, so hours translate into dollars.
The finance essay says the human layer still matters for judgment, benchmarks and accountability, ideally from a fractional firm specialized in boutique professional services and working on top of AI.
The finance essay describes attaching a fully burdened cost to every hour and task, so 25 analyst hours become a 2,500 dollar cost, which then informs pricing, scoping, staffing and whether to automate.
Sources: Greg Alexander, The AI-Native Boutique Firm (Advantage Books, January 2027), specifically The AI Finance Manager for finance reporting the past, the first era as slow and manual and the second as faster and cheaper but not better, dashboards mistaken for understanding, founder financial illiteracy, managing in units of effort rather than economics, the 25 hour and 2,500 dollar example, activity-based costing informing design, pricing, scoping, staffing, capacity and cash, the roughly 80 percent AI and 20 percent human division of finance work, the AI inside the firm ingesting sales, delivery, payroll and cash data, the specialized fractional finance partner quality-controlling output and adding benchmarks, traditional fractional models becoming obsolete without AI, and the unbenchmarked beliefs on payment terms, compensation, close speed, gross margin and EBITDA; The AI Account Manager for account-level costing showing which clients create or destroy EBITDA. Greg Alexander, The Boutique: How to Start, Scale, and Sell a Professional Services Firm (Advantage, 2020), chapter 14 for yield as average fee times utilization and the 400 dollar and 75 percent example; chapter 12 for boutiques running on cash, cash flow per partner and its formula, the 650,000 dollar example from Capital 54, and the doubling questions. Related Collective 54 answers on this site: how do I build a financial forecast I can actually trust; what financial metrics and priorities should I be tracking to grow; how do I calculate the true cost of delivering a service; how do I protect client and sensitive data when using AI tools; should I hire or outsource help to run my books. Note on scope: Collective 54 names no software, publishes no template and gives no accounting or tax advice. The driver-based structure, connecting sources and asking plain questions, benchmarks as the reason for the human layer, rebuilding the last twelve months as a check, the doubling scenario, and the flips are inferences used here to organize the source material rather than published Collective 54 positions.
Collective 54 is the private community for founders and executives of boutique professional services firms between $5M and $50M in revenue. Members work these answers against their own numbers.