The tax firms that win February decide in July

Conventional wisdom says nobody buys tax software in July. The season just ended, extensions are simmering, half the team is on vacation, and every vendor who went quiet in March is suddenly back in your inbox. So tooling decisions get pushed past October 15, to be dealt with "when things calm down."

Our own sales data says something different is happening. July and August are some of the busiest months at FIled, and it is not because firms suddenly enjoy vendor calls in the summer. A growing set of firms has worked out what the tax calendar actually rewards, and they are moving while their competitors wait.

Because here is what the wait-until-fall timeline produces. Demos in November. Contracts in December. Onboarding in January, on live client work, during the worst month to learn anything. Firms that adopt AI this way are betting a season on software they have never seen handle their returns. Some get lucky. The ones who do not spend February fighting their new tool instead of their workload.

There is a better read of the calendar: the off-season is not downtime. It is the only low-stakes proving window the tax year gives you. Between now and January there are three pieces of work sitting in your firm that AI can do today, each one useful on its own, and each one quietly training the tool on how your firm works before the volume hits.

Extension season is a live test bench

Your extension book is real client work at a fraction of season volume. That combination does not exist at any other point in the year, and it is exactly what evaluating AI requires: actual source documents, actual returns, actual reviewer judgment, without the March deadline pressure that turns every experiment into a risk.

Firms running Filed Reviewer on extension work get two things at once. The immediate one is review capacity: our AI compares each return against the source documents behind it and catches what software diagnostics miss, since diagnostics check whether the math is consistent, not whether the numbers match the W-2. The compounding one is calibration. Every correction your reviewers make teaches the system your firm's rules, and corrections made on a few hundred extension returns are corrections nobody has to make on a few thousand returns in March.

The returns you just filed are next year's advisory revenue

Every completed return in your files contains the raw material for a planning conversation: carryovers, entity structure, multi-state exposure, retirement contributions that stopped short of the limit. Most firms know this and mine almost none of it, because planning analysis takes hours per client and the hours do not exist.

This is the work Tax Planner automates. It reads the completed return and the documents behind it, evaluates 120+ strategies, and produces a client-ready plan with citations back to the client's actual numbers. The timing matters more than the tooling: a planning opportunity surfaced in August leaves months to act before year-end. The same opportunity surfaced in December is a nice observation. Firms that run planning scans across their book in late summer walk into fall with advisory meetings on the calendar, and those meetings are billable in a way that recovering from busy season is not.

Cleanup you can only do when nothing is on fire

The third piece of off-season work is the least glamorous and pays the longest. Standardizing workpapers. Encoding the review rules that currently live in your senior reviewer's head. Running a look-back review across the returns you filed in the spring, while amendments are still cheap and the client relationship is warm.

None of this happens in season, ever, at any firm. It is exactly the work an AI platform is suited to hold: Filed builds workpapers from source documents and applies your firm's conventions consistently, which is what standardization means in practice. A firm that spends the off-season getting its book clean and its rules encoded starts the season with a system that already works the way the firm works.

The math on timing

Put the three together and the adoption question inverts. Our AI is accurate from day one on document-first work, and it climbs as your team corrects it. A firm that starts in July does that learning on extension volume, so by January the system has a full off-season of the firm's own corrections behind it. A firm that starts in January does the same learning during the highest-stakes month it has, on the largest volume it will ever see.

The commercial risk runs the same direction. Filed comes with a 60-day money-back guarantee, and sixty days from July lands in September. The entire evaluation fits inside the off-season, with room to walk away and try something else before anyone files a February return. Sixty days from January lands somewhere in the middle of your busiest March week.

Frequently asked questions

When should a tax firm adopt AI tools?

The off-season, deliberately. Adopting between May and October lets a firm evaluate on extension returns at low volume, train the system on its own corrections, and be fully operational before season. January adoption forces the learning curve onto peak workload.

What can tax firms use AI for outside of tax season?

Three workflows carry the off-season: AI review of extension returns against source documents, planning scans across completed returns to surface advisory opportunities before year-end, and cleanup work such as standardizing workpapers and encoding review rules.

Does evaluating AI disrupt extension work?

It should not, if the tool fits your existing stack. Filed works inside CCH Axcess, UltraTax, Lacerte, Drake, and ProConnect, so extension returns move through the same software they always have, with our AI doing the comparison and data entry work underneath.

The one-sentence version

The off-season is when AI stops being a busy-season purchase and becomes part of how the firm operates: extensions prove it, planning scans pay for it, and by February it already works the way your firm works.

If you want to see it on your own extension book, book a demo. Thirty minutes, inside your actual tax software, and the 60-day guarantee means the whole experiment fits comfortably inside the off-season.