Operating guides for CPA and fractional CFO firms, drawn from practitioner conversations. Every piece is dated and corrected in the open.
Several standing pages sit alongside these: where client advisory delivery hours go for firms, and what to ask before signing a firm for the owners they serve. Then the four on the work itself: controller services and how to package them, the 13-week cash flow and the budget vs actual report.
Which clients go on (all of them, and why the usual answer is wrong), how it connects, whose name is on it, what changes in the week, and what the firm still owns.
October 15, 2026
Hire for the work only a person can do, not for daily controller work software can carry. Which kinds of controller work software performs, which still need a person, and four questions that tell you which situation you’re in.
October 15, 2026
An AI CFO can produce the forecast, not the accountability. What the software does well, what stays with the owner, and the questions to ask before you buy one.
October 12, 2026
Two places: the firm using AI on its own analysis, and clients asking it directly. A shared skill library for the first, testing and handoff rules for the second.
October 5, 2026
Three groups sorted by when you can check the output: against the ledger today, only when the future arrives, or never without the firm. Plus what checking costs.
October 3, 2026
Where a language model belongs, which parts should be plain deterministic code, the one place a person is actually needed, and four questions to put to any vendor.
October 2, 2026
The role splits into judgement and attention, and attention looks both backward and forward. Three ways to staff it before the hire, and where each one breaks.
September 23, 2026
The same ten tools read through a partner ledger: capacity, margin, new revenue, what rollout costs across the book, and who owns the client relationship.
September 23, 2026
Bookkeeper, CPA, controller and CFO are four different jobs. Ten tools, each described by who it fits, the problem it solves and what its AI actually does.
September 23, 2026
A bookkeeper produces the records. A controller is accountable for whether they can be relied on, and for the forward view between closes.
September 16, 2026
Price the recurring scope as a fixed monthly fee, set from what the engagement consumes every month rather than the hours a quiet month takes.
September 16, 2026
An eight-step preparation sequence, from prior decisions and reconciliation status through verified outcomes, current results, and the questions likely to come up.
September 3, 2026
Five structural failure modes in advisory delivery: lost decisions, stranded records, live reconstruction, deferred models, and cross-domain questions.
September 3, 2026
Practitioners described two hours to half a day per client per cycle. Where the time goes, who actually decides when it gets spent, and which hours are not yours to control.
September 3, 2026
A working commitment record: what was decided, what observable result should follow, when it was promised, when to check next, and whether it happened.
September 3, 2026
Why clients can know every commitment and still miss its combined cash effect, and how the firm closes the gap: read the forecast for decisions, invert who initiates, keep it in one place.
September 6, 2026
Bookkeeping strain forms a visible queue; controller-level attention never does. Why the gap stays unnamed, what the work actually is, and who should hold it.
September 7, 2026
A daily rebuild keeps the data fresh and the assumptions stale. Why information shared in conversation never reaches the model, and what capture has to look like.
September 8, 2026
Comparison happens; learning does not. Why forecast errors repeat, what scoring actually requires, and how verified outcomes should change the next forecast.
September 9, 2026
Why advisory impact cannot be measured from the books, the five fields a measurable advisory record requires, and why the record protects the adviser as much as it holds the client to account.
September 9, 2026
Three tests before the client acts: which obligations rest on a single event, what the assumptions add to the baseline, and what the track record says.
September 10, 2026
A budget serves control, coordination and feasibility. In owner-led businesses the purpose that most needs protecting is feasibility, and it needs a maintained reference point, not a document.
September 9, 2026
The budget lives in the planning workflow; the decisions happen in the operating workflow. Four structural failure modes that follow, and how to close the gap by hand.
September 10, 2026
Review budget-to-actual monthly, move the forecast as facts change, and revise the approved plan only deliberately. Continuous feasibility, not continuous re-planning.
September 10, 2026
The cost is assembly, not lookup: position, forecast, plan, and the decisions in flight. Why per-question billing backfires, and what actually reduces the cost of answering.
September 11, 2026
The budget records the approved plan, the forecast estimates what will happen, and a scenario tests a proposed decision. Why an affordability question needs all three.
September 12, 2026
Capacity is set by what it costs to answer a question, not by how the firm prices its time. Why changing the fee model leaves the ceiling where it was, and what moves it.
September 12, 2026
Why the published client-count ranges do not travel, what makes one engagement heavier than another, and a four-line worksheet for estimating your own capacity.
September 14, 2026
The package records a period; the questions arrive on the fourteenth. A three-way split between what is delivered on a cadence, what a client can check any time, and what needs judgment.
September 14, 2026