Every accounting-firm managing partner I have spoken to in the last six months is having the same conversation internally. The renewal invoices are up. The vendors are bolting on AI features that feel like demos rather than tools. The workpaper platform still can’t do the one thing your senior associate has been asking for since 2023. And the pricing has moved from per-user to per-user-plus-AI, whatever that actually means.
The default response, for the last twenty years, has been to shrug and renew. The software is good enough. The cost is a line item. The partners have bigger things to worry about during busy season.
That calculus is breaking. The cost of building a piece of custom software that fits your firm exactly has fallen by roughly an order of magnitude since 2023. Meanwhile the cost of your SaaS stack has quietly tripled. The gap is closing from both sides, and at a certain size of firm, it has already crossed.
What “good enough” actually costs
Take a 25-person tax and assurance practice. The typical stack looks something like this:
- Tax prep suite. Per-return and per-seat fees that rise annually.
- Workpaper and audit platform. Per-user licensing, with the “AI add-on” priced separately now.
- Practice management and time tracking. Per-user, with a “collaboration” upgrade most firms end up taking.
- Document management with a portal. Per-user, plus per-GB storage that nobody reviews.
- Client communication tool. Per-user, plus per-send message fees on some tiers.
- Invoice and AR tool. Per-invoice fees or flat per-user, depending on vendor.
Add it up honestly, including the per-user add-ons that were optional two years ago and are now mandatory, and you will find most 25-person firms are paying roughly $8,000 to $15,000 a month for software. Call it $150,000 a year at the midpoint.
That is the floor cost, before the consultant you hired to configure the workpaper tool, the virtual assistant in the Philippines who spends three hours a week cleaning up records because two of these systems don’t talk to each other, and the senior associate whose actual job has quietly drifted into being a systems administrator.
Where SaaS actually fails an accounting practice
Most of these platforms are fine for the commodity work: holding documents, running the calculation engine on a return, storing engagement letters. Where they fail, consistently, is at the seams between them, and at the parts of your practice that are not generic.
- Engagement-letter intake. Your firm has engagement-letter templates that encode a partner’s twenty years of risk calibration. No SaaS tool understands those templates. Associates copy, paste, and edit manually every engagement.
- Workpaper review. Your reviewers have opinions. Those opinions are not written down anywhere. They are transmitted through review notes that junior staff read once and forget. An AI workpaper tool that does not know your firm’s review standards gives you a reviewer who cannot tie.
- Client-communication cadence. The partner who manages the Smith family trust knows that Mr. Smith wants a two-paragraph update before the 15th of each quarter. No practice management tool knows this. It lives in the partner’s head, and it dies when the partner retires.
- Cross-system reconciliation. Your time entries, your billing, your deliverables, and your engagement letters all live in different systems. Any report that needs data from three of them is a manual exercise.
These are the workflows that make your firm different from the firm down the street. They are exactly the workflows that off-the-shelf software cannot address, because the software is built for the average firm, not yours.
A concrete example
A regional assurance practice was spending roughly 15 hours per engagement on workpaper review documentation. The review partner had a very specific style: every exception note followed a format the partner had used for fifteen years, and deviations from that format caused rework during final review.
A workpaper-AI platform could not help. It tried to impose its own review standard on the firm, which produced output that was worse-than-nothing because it had to be manually rewritten by the reviewer.
A custom module, scoped to exactly this workflow, solved it in six weeks. Training data: eighteen months of the firm’s own-completed reviews. Output format: the review partner’s exact exception-note style. The module dropped per-engagement review documentation time from 15 hours to roughly 3. At 200 engagements a year, that is 2,400 reclaimed hours. The module cost less than one quarter of the firm’s combined SaaS stack to build.
The firm owns the module. It does not renew. It does not raise prices. It does not add features the firm did not ask for. It does one thing, exceptionally well, and it does it forever.
What to build first in an accounting firm
The best candidate workflows for a first bespoke module share the same profile regardless of practice area:
- The workflow is touched by every engagement or every month, not once a year.
- The work has firm-specific review standards that cannot be imported from a SaaS template.
- The time currently being spent on the workflow is by salaried staff, paralegal-equivalent, or by a partner doing work that should not require a partner.
Common first-module candidates in accounting:
- Engagement-letter drafting against a partner’s historical templates.
- Workpaper documentation and exception-note generation in a single practice area.
- Quarterly client-communication drafts tuned to each client’s preferred cadence and style.
- R&D credit substantiation write-ups pulled from engineering notes.
- Monthly-close checklist automation that reads the firm’s own historical close files.
Each of these is a 4-to-6 week build. Each one reclaims hundreds of hours annually. Each one is owned by the firm outright, with the source pushed to a GitHub repository that the firm’s partners control.
Start small. Own it. Compound.
You do not need to replace your tax prep suite or your general ledger. You will not be well served by doing so, and nobody building in this space is suggesting you should. What you should do is pick the single most expensive manual workflow in your firm, the one that eats staff time and partner patience in equal measure, and build a small, precise tool that replaces it.
Then own the tool. Then do it again with the second-most expensive workflow. That is how firms in 2026 will quietly build a structural advantage over their peers. Not by buying a better platform. By building small, specific modules that compound into a practice that runs differently.
