Your busiest preparer quit in October, two clients just dropped off shoeboxes of documents with April circled on their calendars, and your review pile is already taller than your patience. This is the tax season you are walking into — short-staffed, document-heavy, and with no cavalry coming, because the accountant shortage stopped being a headline and started being your staffing plan. Into this exact pain point, practice-management vendor Canopy announced in January 2026 that it is wiring AI-powered tax preparation directly into its platform through a new integration with Filed, an AI tax prep company. The feature is called Smart Prep, and its promise is the shortest sentence in tax software marketing: finished client intake in, review-ready return out.
This post covers what was actually announced, how the intake-to-review workflow works, the time-savings numbers the vendors claim and how to test them, what a small firm should budget for tooling like this, the catches worth knowing before you rely on it, and a practical checklist for evaluating any AI prep tool this season.
What was announced
On January 15, 2026, Canopy — the Salt Lake City practice-management platform used by accounting firms for workflow, documents, billing, and client communication — announced it is adding AI-powered tax preparation to its platform through an integration with Filed. Trade coverage followed within days. Canopy brands the capability Smart Prep, powered by Filed, and describes it as dramatically accelerating the journey from completed client intake to a review-ready tax return.
Three details in the announcement matter more than the headline:
It starts at intake, not at a blank return. The workflow begins after the client has finished submitting documents through Canopy's portal. Smart Prep takes that completed intake package and produces a draft return, rather than asking staff to re-key W-2s, 1099s, and brokerage statements into tax software by hand. Document chasing stays human; data entry is what gets automated.
The initial launch covers individual returns. The pilot handles basic and advanced Form 1040 returns. Business returns, trusts, and nonprofits are the stated direction of travel — Filed's own platform already advertises 1065, 1120, 1120-S, 1041, and 990 support — but a firm evaluating Smart Prep for this season should assume 1040s only and treat everything else as roadmap.
It launched as a pilot. The integration was in pilot at announcement, with broader availability promised after the 2026 tax season. That timing is deliberate: no vendor wants its first production run to be your March. But it also means early adopters are buying into a controlled rollout, with all the feature and capacity limits that implies.
How the workflow actually works
Smart Prep is best understood as a bridge between two products that already exist, not as a new tax engine. On one side sits Canopy: the client portal where documents land, the workflow engine that tracks each return's status, and the delivery system that sends finished returns back for e-signature. On the other side sits Filed's AI preparation pipeline, which reads source documents, extracts the figures, checks them for inconsistencies, and drafts the return.
The handoff runs like this:
- Collect. The client uploads tax documents through Canopy's portal — the Smart Intake side of the house. Staff confirm the package is complete, the same completeness check they do today.
- Prepare. With intake finished, the firm sends the file package to Filed from inside Canopy. Filed's AI reads the documents and prepares the draft return inside the firm's existing tax software. Canopy names Drake, UltraTax, ProConnect, and CCH Axcess as supported destinations, matching the programs most small firms already run.
- Review. Filed's Reviewer component checks the draft against the source documents and produces a cited list of review points. A licensed preparer at the firm then reviews, corrects, and signs off — the professional-in-the-loop step the whole design depends on.
- Deliver. The reviewed return flows back into Canopy's workflow for assembly, client delivery, and e-signature, with status updates syncing automatically so nobody retypes a status or hunts through email for where a return stands.
Notice what this changes about staffing. The scarce resource in most small firms is not review capacity — partners and senior reviewers exist — it is preparation hours: the junior and seasonal staff who turn documents into first drafts. Smart Prep attacks exactly that bottleneck. The reviewer role does not disappear; if anything it becomes more important, because someone with judgment must verify work they did not personally key in.
Smart Prep grew into a partner framework
One later development is worth knowing. Canopy's Smart Prep page now describes the capability as a partner framework rather than a single-vendor feature: Smart Prep connects Canopy firms to third-party AI prep providers, with Filed joined by additional partners such as Juno and Byron. The practical consequences are twofold. First, the AI prep subscription is separate from Canopy itself — each partner prices its own product, so Smart Prep is a second invoice, not a checkbox on your Canopy renewal. Second, the framework framing means a firm can switch AI prep providers without switching practice management, which is genuinely good for buyers. Evaluate the partner, not just the portal button.
The time-savings claims, and how to test them
Filed publishes the most specific numbers in this corner of the market: firms using its platform report 64% shorter review cycles and about 1.5 hours saved per return, with data-entry review of a standard return taking under three minutes before the preparer approves the entry plan. Independent roundups of 2026 accounting AI tools cite broader industry claims of 30 to 50% total time savings per return and 25 to 40% more clients per season with the same team.
Treat every one of these figures as a hypothesis about your firm, not a fact about the software. Savings scale with return simplicity: a W-2 plus standard-deduction 1040 automates beautifully, while a rental-property return with three K-1s, a home-office worksheet, and a prior-year carryforward still needs substantial human prep. Three ways to make the numbers honest before you commit:
- Time the review, not the prep. The vendor times how fast its AI drafts. You should time how long your reviewer takes to verify the draft against source documents, because that is the labor you still pay for. If review takes nearly as long as your old prep-plus-review, the savings are cosmetic.
- Run a shadow set. Pick twenty already-filed returns spanning your real mix — simple, moderate, messy — and run them through the tool. Compare the AI draft to the filed return line by line, and log every correction by category: extraction error, judgment call, missing document. Your correction log is your true savings estimate.
- Price the subscription per return. Filed charges a flat monthly platform fee scoped to taxpayer count rather than per-return pricing. Divide the annual cost by the returns you will actually route through it — not your total return count — to get your real per-return cost. A flat fee is a bargain at volume and expensive for a firm that routes only its simplest fifty returns.
What it costs to run, and how to book it
AI prep tooling lands in your books the same way every other firm subscription does: as software expense, ideally in its own subaccount so you can watch it grow against the labor it replaces. If you track preparation cost per return — and a season with AI tooling is the right time to start — the subscription fee plus the reviewer's time on AI-drafted returns is your new unit cost. Compare it against last season's fully loaded prep cost per return, not against zero.
A firm keeping its own ledger in plain text might record the monthly subscription like this:
2026-01-15 open Assets:Checking:Business USD
2026-01-15 open Expenses:Software:TaxPrep USD
2026-02-01 * "Filed" "Smart Prep subscription - February"
Expenses:Software:TaxPrep 299.00 USD
Assets:Checking:BusinessThe deeper bookkeeping point is about your clients' books, not yours. Faster prep cycles compress the time between document collection and delivered return, which means estimated-payment advice, extension decisions, and balance-due conversations all happen earlier — while the client can still act on them. If you keep client workpapers in Beancount, the same discipline applies: machine-readable source documents plus a human review step produce a file you can defend. The firms that gain the most from AI prep are the ones whose intake and review checklists were already rigorous, because the tool multiplies the process you have. A sloppy intake process automated is just sloppy returns delivered faster.
The catches worth knowing
No honest evaluation stops at the feature list. Five cautions apply to Smart Prep specifically and AI prep generally:
Pilot means pilot. Broader availability was promised only after the 2026 tax season. A firm planning its season around a tool still in controlled rollout needs a fallback: confirmed access dates in writing, a cap on how many returns can go through it, and a plan for what staff do if the pilot pauses mid-March.
1040s first. If your firm's bottleneck is business returns — 1065s and 1120-S returns with messy books — Smart Prep's launch scope does not touch your pain. Match the tool's coverage to your actual backlog, not to your total return count.
The reviewer still signs. Every return needs a named professional's sign-off, and professional standards do not discount for AI assistance. Reviewer liability is unchanged; only the draft's author changed. Your engagement letters and review checklists should already reflect that the firm, not the software, stands behind the return.
Data boundaries matter. Client tax documents now flow from your portal to a third-party AI provider and back. That is a second data processor in your chain, which means reviewing the partner's security posture, data-retention terms, and whether client data trains shared models — and updating your own engagement letters and privacy notices to match. Canopy publishes AI governance documentation; read the equivalent from whichever prep partner you choose before routing a single real return.
Two invoices, two renewals. Because Smart Prep partners bill separately, your Canopy renewal and your AI prep subscription will have different terms, different renewal dates, and different cancellation windows. Calendar both. Nothing sours a vendor relationship like discovering in February that the tool your workflow depends on renewed at a higher tier in December.
How competitors frame the same bet
Canopy and Filed are not alone, and the competitive shape tells you this category is real rather than a single vendor's experiment. Thomson Reuters ships Ready to Review, an agent-based flow that automates the gather-and-prepare stages for 1040 returns inside its own ecosystem. Wolters Kluwer added agentic preparation to CCH Axcess Tax for individual returns. Independent AI prep vendors — Black Ore, Juno, and Filed itself outside the Canopy integration — sell direct to firms on various tax platforms. The common architecture is identical everywhere: AI drafts, human reviews, existing tax software remains the system of record.
That convergence is useful information for a buyer. It means the switching cost sits in the review workflow and staff habits, not in the AI vendor — and it means pricing and accuracy, not feature checklists, will decide winners. Run your shadow set against two vendors if you can; a standardized accuracy test is the only comparison that matters.
A practical checklist for this season
If you are considering Smart Prep or any rival before the filing rush, work through these steps in order:
- Confirm access in writing. Pilot seat, start date, return-type coverage, volume cap. Verbal promises from a sales call do not survive March.
- Run the shadow set. Twenty filed returns across your real complexity mix, corrected and categorized, timed at review.
- Compute unit cost honestly. Annual subscription divided by returns actually routed, plus reviewer minutes at loaded cost, versus last season's prep cost per return.
- Read the data terms. Retention, model training, subprocessors, breach notification — for the prep partner, not just Canopy.
- Update engagement letters. Disclose AI-assisted preparation and confirm the firm's review responsibility in language your liability carrier accepts.
- Train reviewers on the new failure modes. AI drafts fail differently than junior preparers: confident extraction errors, misread statements, and edge cases nobody flagged because no human puzzled over them. Your review checklist needs new lines, not just faster old ones.
- Keep the fallback warm. Staff who can still prep by hand, software licenses current, and a triage rule for which returns never touch the AI path — the complex, the adversarial, and the ones where the client relationship demands a human touch.
Work that list and you will know by February whether AI prep earns its place in your season or waits another year. Either answer is fine; the expensive outcome is discovering the answer in April, with a review pile that has opinions of its own.
