Between February 1 and April 15, the average solo CPA works 55-70 hours a week and still leaves money on the table. A Karbon 2024 study of 1,300+ accounting professionals pegged the average during-season workweek at 57 hours, with roughly 15-20 hours per week spent on work that is not actually tax prep or advisory — bank feed cleanup, document chasing, re-keying numbers between systems, client status updates, IRS notice triage. (Karbon 2024 Accounting Industry Workforce Report)
A CPA in Framingham we talked to in March rebuilt one piece of that: bank feed cleanup across her 34 monthly bookkeeping clients. Before, an admin spent roughly 4 hours on Monday mornings categorizing uncategorized QuickBooks transactions. After pointing Dext Prepare at the same feeds and training its rules on 90 days of her categorization history, Monday mornings are back to 40 minutes of review. That is one automation. It recovered 3+ hours a week during the busiest stretch of the year.
This is the shape of the real opportunity for solo and small-firm CPAs. QuickBooks sits at the center. The leverage layer is everything built around it.
Inside QuickBooks: what AI actually does vs. what the marketing says
Intuit spent most of 2024 and 2025 marketing "Intuit Assist," its embedded AI layer across QuickBooks Online, TurboTax, and Mailchimp. In QBO specifically, the features that actually shipped as of early 2026:
- Transaction categorization suggestions in bank feeds, based on prior categorization
- Invoice drafting from a natural-language prompt
- Cash flow forecasting charts based on historical data
- Receipt capture with OCR extraction (this has existed since the QuickBooks Receipts launch in 2019; the "AI" labeling is newer than the capability)
What Intuit Assist does not do, despite the marketing framing: reconcile with any reliability across multiple bank accounts, handle complex chart-of-accounts rules, produce client-facing advisory narratives, or meaningfully reduce month-end close time beyond the transaction-categorization piece. Independent accounting-tech reviews from Insightful Accountant's 2024 roundup and CPA Practice Advisor's QBO feature scorecards put Intuit Assist's real time savings for a typical multi-client firm at 15-25 minutes per client per month. (Insightful Accountant, Intuit Assist Review 2024)
Useful. Not transformative. The transformative work lives outside QB.
Context for why. Intuit's roadmap is correctly pointed at the QuickBooks Online small-business user, not at the CPA who logs into 20-40 client files a week. The AI inside QB is optimized for a self-serve bookkeeper, not a multi-client practice. That is a product decision, not a capability gap, and no amount of waiting for the next release will change it.
The leverage layer: what actually works around QuickBooks
Four categories of tools, with pricing and honest time estimates.
Bank feed and document classification
Dext Prepare (formerly Receipt Bank) runs $31-$50/user/month depending on volume and integrations. (Dext pricing, 2026) It sits between bank feeds, supplier emails, and QuickBooks, auto-categorizing transactions against rules it learns from the practice's actual history. For a solo CPA with 20-40 bookkeeping clients, Dext typically saves 3-6 hours per week during normal months and 8-12 hours per week in tax season, according to its own case studies and what we have measured at firms we work with.
Keeper ($125-$185/month for practice-wide usage depending on client count) is built specifically for the review step: flagging uncategorized transactions, missing receipts, unreconciled balances, and open client questions across a whole book of business in one dashboard. (Keeper pricing, 2026) The 2024 Keeper State of Bookkeeping report found firms using it cut month-end close time by 45-60% on average. (Keeper State of Bookkeeping 2024)
Hubdoc is bundled with Xero, useful if a CPA runs a mixed QB/Xero book. Free inside Xero plans, otherwise $20/month standalone.
Client-portal intake and workflow
Karbon ($79-$99/user/month), Canopy ($45-$99/user/month), and TaxDome ($66-$99/user/month) are the three major practice-management platforms for tax and accounting firms, and all three shipped meaningful AI features in 2025-2026. (Karbon pricing, Canopy pricing, TaxDome pricing)
- Karbon's "Karbon AI" auto-drafts client email replies based on the firm's prior tone, summarizes long threads, and extracts action items. The 2024 Karbon study cited above found users save an average of 16.5 hours per week across the firm once rolled out.
- Canopy's document AI extracts data from uploaded client documents (W-2s, 1099s, K-1s) and pre-populates the client's profile.
- TaxDome's AI handles client-portal message triage and generates draft status updates against a client's open engagement.
The honest framing: these tools are not magic. They replace 60-80% of the copy-paste and status-update work that otherwise eats a CPA's mornings. The remaining 20-40% still requires judgment.
Client status summaries for advisory conversations
This is where general-purpose LLMs like Claude ($20/month Pro, $30/user/month Team) and ChatGPT ($20/month Plus, $30/user/month Team) earn their keep for a solo CPA. Paste in a year-over-year P&L, a trailing-12 cash flow, and a short prompt ("draft a two-page narrative for a manufacturing client, focus on gross margin decline and working capital"), and the model produces a first-draft advisory memo in 90 seconds that used to take 45 minutes.
The constraint is data handling. QBO exports PII and financial data; both Claude and ChatGPT's Team tiers are configured not to train on customer data, but firm-level policy and client consent matter. Run it through your engagement letter.
Tax notice triage
IRS letters, state notices, and underpayment assessments pile up in April and October. Transcribing a CP2000 or a Notice of Assessment into a response draft used to be a 30-45 minute task. With Claude or ChatGPT, a CPA can photograph the notice, paste it in, and get a structured triage in under five minutes: what it is, what it is asking for, what documents to pull, and a first-draft response. This is the single highest-ROI LLM use case for a solo CPA and the one most underused.
A note on overhype
"AI for CPAs" is one of the most overclaimed categories in professional services software right now. A lot of what ships as AI is a rules engine with a chatbot glued on. The honest filter for any tool pitched to a solo or small-firm CPA:
- Does it actually reduce clicks, or just add a natural-language layer over the same clicks?
- Does it learn from the firm's historical categorizations, or apply a generic model?
- Does it integrate with QuickBooks at the API level, or via a CSV export every morning?
- Does the vendor publish measured time savings from real firms, or only customer quotes?
The tools above pass most of those filters. Plenty do not.
Workflow redesigns that actually compound
Tool-by-tool pricing is useful. The 10+ hours a week savings come from redesigning three workflows end to end.
1. Client onboarding: source-to-QB with zero re-keying
The typical onboarding: client fills a PDF, CPA copies fields into Canopy or TaxDome, then into QBO, then requests bank access, then sets up a client folder. Four hand-offs, same data typed three times.
The redesign: a Jotform or TaxDome intake, mapped via Zapier or Make, that creates the client in QBO, provisions a portal folder, and triggers a welcome email. Setup: roughly 6-10 hours of configuration. Savings: 45-75 minutes per new client, forever. For a firm onboarding 40 new clients a year, that is 30-50 recovered hours annually.
2. Month-end close automation
The pieces are all there: Dext or Keeper for feed cleanup, a scripted reconciliation check for accounts that typically match, a review-only dashboard in Keeper for exceptions, and auto-generated client reports. A 20-client bookkeeping book that used to take two full days of closing now takes a half day of review, with the rest automated.
3. Review-and-approve for recurring journal entries
Recurring rent, payroll clearing, intercompany, and depreciation entries are templated. Pull them out of the CPA's head, into a rules engine (Keeper's workflows, or a simple Airtable + Zapier flow feeding QBO), with a review step. The entries post themselves; the CPA approves. Typical time savings: 2-4 hours per month per book of 15-20 clients.
The first-week move
Before a practice rebuilds anything, one automation to set up this quarter:
Point Dext or Keeper at your three biggest monthly bookkeeping clients and let it run for 30 days. Measure the Monday-morning cleanup time before and after. If the savings match the numbers above (and they usually do), expand to the rest of the book. If they do not, you have learned something specific about your practice that no marketing deck could have told you.
This is the discipline that separates CPAs who get real leverage out of AI from the ones who pay for seven tools and still work 65-hour weeks in April. The goal is not to adopt AI. The goal is to recover 10+ hours a week during tax season, without re-keying a single number.
QuickBooks is the center. Everything else is plumbing, and the plumbing is where the time is.
If you want to see where your practice is actually losing hours:
- Take the Operational Intelligence Index Assessment, a 3-minute diagnostic scored against benchmarks for solo CPA practices.
- See how we work with solo CPAs and professional services firms specifically: the integrations, the automations, and the pricing, without the platform pitch.
The hours are recoverable. The question is whether the first automation goes live in May or next February.