The phones flood with routine calls and real work stops getting done

It is two in the afternoon on a Wednesday in March, and the phone at your accounting firm has not stopped since lunch. One caller wants your hours. The next wants a fee range before committing to anything. The actual work on your desk has not moved in an hour because the phone keeps pulling someone off it. By six the office goes quiet, and three after hours voicemails wait for tomorrow, from people who may have already called someone else.

When I sit down with owners of small accounting and bookkeeping firms, the phones come up first, almost word for word: the office floods with routine calls in season and it breaks everyone’s concentration, then after hours calls get missed entirely. Then the rest follows. Can something chase clients for missing documents without me asking twice. Can I stop losing hours on hold with the IRS. Is it safe to put a Social Security number near an AI tool.

I am Isaac Krupp, an AI consultant for small businesses. I spend a full working day inside firms like yours, watching how the work actually gets done, then hand over a written plan. I sell no software, build nothing custom, and take no referral fees.

What AI for accountants can actually do right now

Phones ring nonstop in season

During filing season the office floods with routine calls, hours, pricing, where is my refund, and those calls break concentration while after hours calls go missed. One owner put it plainly: her firm stopped offering IRS call ins because “we cannot maintain productivity staying on the phone all day”.

The fix is an AI phone or chat receptionist that answers hours, fee ranges, document status, and refund timing, books appointments, and rings a person for anything needing real judgment, usually $100 to $300 a month for a small firm.

The boundary: never tax advice, never a client’s numbers or return status in detail, never a refund amount or deadline extension over the phone. Routine information and scheduling only, everything else goes to a person.

Clients send documents late and piecemeal

Clients drip in incomplete files, email sensitive documents instead of using the portal, and staff spend hours chasing paperwork instead of doing the work. A 2020 survey of US taxpayers found 34 percent procrastinate on preparing their return and on average do not start thinking about it until February.

The fix is an AI intake assistant that sends reminders for missing documents and checks uploads against a checklist the preparer set, often already a feature inside a portal running $20 to $50 a month, or a small add on to one already in use.

The boundary: never reading or interpreting a document’s financial content, never deciding on its own what is missing instead of following the checklist, never accepting documents outside an encrypted, access controlled portal.

Getting paid late hurts cash flow

Clients sit on invoices that stretch weeks past due, and one slow payment can make it hard to cover payroll. Intuit’s own data found 59 percent of US small businesses report invoices overdue thirty plus days, averaging $17,700 outstanding.

The fix is AI assisted invoicing already inside most accounting software: automatic payment reminders, flags on aging invoices, and a card on file for retainer clients. Often already built in, or a small add on running $20 to $40 a month.

The boundary: never writing off a balance, never threatening collections, never handling a payment dispute conversation. Payment terms and hard conversations stay with a person.

Too many disconnected software tools

We run a pile of different apps that do not talk to each other, re key the same client into half of them, and feel overwhelmed by the tech stack every week. Intuit’s survey of accounting professionals found the average firm runs eight apps for core operations and 66 percent feel overwhelmed by technology complexity at least weekly.

The fix is AI powered automation that connects the practice management, accounting, and e signature systems already in use, so data moves once instead of by hand, typically tens of dollars per seat a month.

The boundary: never moving money, never filing anything, never overwriting a client record without a human reviewing the sync first. It only moves data between systems the firm already trusts.

Hours lost on hold with the IRS

It takes attempt after attempt just to reach a hold queue, and staff cannot maintain productivity staying on the phone all day. Clarity Tax Relief reports that only 26 percent of Accounts Management callers reached a person in FY2025, billable time spent listening to hold music.

The fix is a narrow callback or queue monitoring tool. It dials in, waits on hold, and pulls in a staff member once a live person answers, usually tens of dollars a month per line, not a broad AI platform.

The boundary: never speaking to the IRS on the firm’s behalf, never sharing taxpayer information, never making representations. It only holds the place in line, a licensed preparer handles the conversation.

Everything lands in the same few weeks

The workload simply exceeds the hours on the team during the crunch, and owners quietly overwork and underprice just to get through it. One worked example shows five accountants working forty five hour weeks, 225 hours total, against a 300 hour peak season workload, a 75 hour weekly gap before review even starts, and Accounting Today describes owners overworking and underpricing to get through the crunch.

The fix is an AI intake and scheduling assistant that pre sorts incoming work by deadline and complexity, so simple returns get routed and batched automatically, usually already inside practice management software.

The boundary: never deciding how a return is prepared, never skipping a review step, never reassigning a client’s work without a partner’s sign off. Triage and scheduling only, never technical judgment.

Can’t find or keep qualified staff

Finding qualified people is the number one problem for every firm size except solo practitioners, and most firms say hiring an experienced preparer is hard. Intuit found 80 percent of accounting professionals report difficulty hiring experienced staff, and the AICPA’s survey found it ranks as the top issue for firms of every size except solo practitioners.

AI does not hire anyone. It hands a team back the hours that need no license: intake reminders, a receptionist, extraction, connected systems, so a three person firm behaves like a five person one, usually a few hundred dollars a month, well below a part time hire.

The boundary: none of it replaces licensed review, signing authority, or seasoned judgment, and it is never presented to a client as their accountant.

AI pressure with no clear plan

Everyone says AI will change the firm and clients expect it, but nobody actually knows where it fits. The AICPA’s 2026 survey found change management around technology, including the rise of AI, is the leading five year concern across every firm size, while Intuit found 64 percent of accounting professionals plan to invest in or upgrade AI this year with no shared sense of what to buy first.

This is the gap the free day is built to close: a walk through of the firm’s actual workflow ending in a written plan naming three to five moves with honest costs, never a platform pitch.

The boundary: never touching a tax position, never filing anything, never replacing a preparer’s judgment. Every suggestion stays in the category of taking work off desks, never deciding for people.

Protecting client data from breaches

We hold everyone’s Social Security numbers and bank details, and phishing, shared logins, and remote access keep owners up at night. The AICPA’s 2026 survey ranks data security as a top four issue for solo practitioners and firms up to ten people, and Accounting Today named it one of small firms’ strongest concerns heading into 2025.

The fix here is mostly hygiene rather than AI: a password manager for the firm, multi factor login on every app, an email filter with phishing detection, and a written policy naming which tools may see client data, usually a few dollars per user.

The boundary: never the sole holder of encryption keys or credentials with no human backup, and any flagged incident still goes to a person. This is detection support, not a compliance decision maker.

A day in an accounting firm with AI turned on

Priya runs a five person tax and bookkeeping practice in Glendale. Here is a Wednesday in March, before and after.

Before. She is in by seven thirty to clear email before the phones start. Her office manager, who bills out at, say, $40 an hour, spends the morning fielding calls about missing documents and typing two new clients into the tax software, portal, billing, and calendar, one system at a time. Priya, at, say, $180 an hour, takes three did you get my documents calls herself, spends forty five minutes redialing the IRS line and gives up, then keys numbers off a brokerage statement by hand before leaving at seven to review returns at home. Nothing that needed her license took more than two hours of her day.

After. Priya is in by eight thirty. Overnight the intake tool nudged the nine clients still missing a document and marked four returns ready to start. The receptionist handled the routine calls, booked an appointment, sent two refund callers to the IRS tracker, and texted one summary, a client received a notice and wants a callback, which she takes herself. A hold service waits in the IRS queue and rings her the moment a person picks up. The brokerage figures were already extracted, so she spends fifteen minutes verifying instead of an hour keying, and leaves at six with review done.

Every piece of that exists today, off the shelf. The hours that needed her license stopped going to typing and waiting.

What it costs, honestly

I sell no software and take no referral fees, so here is the version without a sales deck.

Intake tools usually run tens of dollars per seat a month. An AI receptionist is usually around a few hundred dollars a month, often less than the answering service it replaces. Invoicing add ons, connectors, and IRS hold services typically run tens of dollars a month, security basics a few dollars per user, and research libraries are the outlier, a few hundred to a couple thousand a year.

A typical small firm lands at a few hundred dollars a month across the whole practice, well under one part time hire.

The expensive mistake is custom software, paying for something built just for your firm before anyone has used the off the shelf version for six months. Try the boring tools first. Most firms never end up needing anything custom.

What NOT to hand to AI in an accounting firm

This is a regulated profession, and these tools are good at the mechanical layer only. Your own licensing and regulatory obligations govern everything below, and your advisor or state board has the last word on anything unclear.

Anything that decides a tax position. Elections, deductibility, entity choice, stays with a licensed preparer who reviews and signs.

Client confidentiality. Clients gave your firm their Social Security numbers and bank details under duties you cannot hand off to a vendor. No client data goes into a consumer chatbot or any tool without a written no train, no retain agreement, shared logins end, and third party handling is disclosed where the rules require it.

Representation. No AI tool talks to the IRS or a state agency on the firm’s behalf. That takes a signed power of attorney and a licensed person who reviews and signs every letter.

How the free AI day works for an accountant

I spend one full working day inside your firm, free. In person across Los Angeles and Southern California, by video anywhere in the United States. No card, no contract, no catch. The full process is on the how it works page.

In the morning I watch. I sit with whoever answers the phones and tag call logs as routine or judgment, sit with whoever chases documents and count the touches to finish last season’s slowest returns, then follow a return from intake through signature, checking every login for where client data gets typed by hand.

In the afternoon we total last quarter’s IRS hold time against a hold service, check quoted hours against actual hours on recent fixed fee jobs, and list every tool that can see client data along with its training terms.

You end the day with a written plan naming three to five moves, each with an honest cost and what it must never touch, plus a one page AI use policy if your firm lacked one. The same boundaries carry over to the AI for lawyers and AI for financial advisors pages.

I will not tell you how to use AI in accounting from a slide deck. I will watch your firm work for a day, then hand you a plan for Monday. What you’re really buying is a season where you leave the office at a normal hour and trust the returns are still moving with the same care you’d give them. If that sounds useful, book your free AI day.