An accounting or bookkeeping firm runs into a constraint that most operating businesses recognize: there is more work available than there are people to do it. A Wall Street Journal analysis counted more than 300,000 U.S. accountants and auditors leaving their jobs in the two years to late 2022, a decline of seventeen percent, and the pipeline behind them is thin. The AICPA's 2025 Trends report counts 55,152 accounting bachelor's and master's degrees awarded in the 2023-24 school year, down 6.6 percent from the year before, and the profession's National Pipeline Advisory Group reports that among undergraduate business students, only one in nine chooses an accounting major. The Bureau of Labor Statistics projects on the order of 130,000 openings a year for accountants and auditors. When a firm wants to grow, the question is not whether it can afford more staff. It is whether the staff exist to be hired.
That distinction changes what automation is for. In most cost-reduction accounts, software replaces a person to make a line item smaller. In a firm that cannot fill its open seats, software does something different. It lets the people already on payroll carry more accounts each, which is the only growth path available when hiring has stalled. This is one instance of a pattern that holds across operating businesses where labor, not budget, is the limit. We have written about the same dynamic in a staffing firm, where the constraint is recruiter hours rather than recruiter cost.
Where the hours actually go
If you sit behind a bookkeeper closing out a client's month, the work sorts into three layers. The first is transaction categorization, where every line on the bank and credit-card feed gets assigned to an account, with the bookkeeper deciding whether a charge at an office-supply vendor is supplies, equipment, or a personal expense run through the business by mistake. The second is account reconciliation: bank to ledger, the merchant-processor deposit to the invoices it settled, the loan statement to the recorded balance, line by line, with anything that does not tie flagged for follow-up. The third is workpaper preparation, which means assembling the supporting schedules, the fixed-asset roll-forward, the accruals, and the documentation that backs each number so a reviewer or an auditor can trace it.
Across these three layers, the volume is heavily weighted toward the repetitive. The great majority of transactions in a typical month are ordinary and recur from period to period: the same rent payment, the same software subscriptions, the same handful of recurring vendors. The reconciliations that tie cleanly outnumber the ones that do not, often by a wide margin. The judgment that requires a qualified accountant is real and consequential, but it is a small fraction of the line count. It shows up in the unusual transaction, the reconciliation that will not close, and the client circumstance that changes how an item should be treated. The firm's senior people spend a large share of their time on the mechanical majority precisely because there is no one junior to hand it to.
The redesigned workflow
In the redesign, an agent takes the repetitive layer and the accountant keeps the rest. On categorization, the agent proposes an account for each transaction using the client's own history and chart of accounts, posts the high-confidence recurring items, and routes the ambiguous ones to a queue with the candidate accounts and the reason it is unsure. On reconciliation, the agent matches deposits to invoices and statement lines to ledger entries, closing the items that tie and surfacing only the breaks, each with the documents already pulled together so the accountant can adjudicate without hunting. On workpapers, it assembles the schedules and the supporting tie-outs into the firm's templates, leaving the accountant to review, sign off, and write whatever commentary the client or the file requires.
What the accountant keeps is the part that was always the point: the review, the judgment at the exceptions, and the client relationship. The agent does not call the client to ask why a large unexplained transfer hit the operating account in the third week of the month. The accountant does, because that conversation is where the value and the trust sit. The mechanism here is the same one we walked through in the month-end close. The agent does the chasing and the matching, then hands a person a short, ranked list of the few things that genuinely need a decision.
Why this is a capacity argument, not a cost one
The reason to frame this as capacity is that the firm's growth is gated by a number it cannot move by spending. The 2024 CFO Pulse Survey, reported through AACSB, found that eighty-three percent of finance leaders could not find qualified talent, up from seventy percent in 2022. A firm in that market does not respond to demand by hiring; it either turns work away or stretches the people it has. Retention makes the squeeze worse. The Institute of Management Accountants and Robert Half reported turnover of thirty-nine percent among accounting and finance professionals aged eighteen to thirty-six, so even the seats a firm fills do not stay filled.
Automating the repetitive layer is how a firm adds capacity inside that constraint. The same headcount can carry more accounts because the hours that went to categorizing recurring charges and ticking off clean reconciliations are returned to review and to clients. Firms that have already reached for outside capacity show the demand is real: the AICPA reports that roughly a quarter of firms outsource part of this work. Outsourcing and automation are answers to the same question, which is how to do more accounting work than your in-house staff can do alone. This is why the redesign adds capacity rather than cutting cost: when a firm cannot hire the people to take on more work, automating the repetitive layer is how it takes on more, a dynamic we have described in why services-as-software firms scale.
A reasonable counter
A reasonable counter is that bookkeeping data is messy, every client keeps its books differently, and an agent confident enough to post the routine cases will eventually post a wrong one. The objection has real force, and the answer is in how the work is split. The agent posts only where its confidence is high and the pattern is established, and the categories that hold the most judgment never bypass review. A misposting is caught at the same review step that already exists, because nothing in the redesign removes the accountant's sign-off. It removes the manual assembly that precedes it. We treat that worry on its own terms in the objection that the data is too messy.
There is a sharper version of the same worry: that this is a quality risk dressed up as a capacity gain, and that a firm under labor pressure will be tempted to thin out review to grow faster. That temptation is real, and it is a management decision, not a property of the tooling. The redesign does not force a firm to review less. It gives the same reviewers more time per file by taking the assembly off their desks, which is the opposite of thinning review. A firm that chooses to cut corners could do so today with offshore staff or with junior hires it does not supervise. The capacity an agent adds is most defensible when it is spent the way the labor numbers suggest it must be: keeping experienced people on the judgment and the clients, and letting the firm say yes to work it would otherwise have declined for want of anyone to do it.