Across the Table Build. Grow. Thrive.

AI & Technology Industry Stack 07 · Accounting firm Free Business Insights

The Accounting Firm Stack

Seven systems for a practice that holds other people’s financial data — and the one rule that decides every tool choice.

About 7 min read

Download Knowledge Brief (PDF) .pdf · 141 KB

Who this is for

You run a bookkeeping or accounting practice, or a tax shop, or you are the fractional finance person for several small businesses. Your clients hand you their most sensitive records and expect them back organized and correct.

This guide is for the practice owner weighing which parts of the work can be assisted and which cannot. The honest answer is that a great deal can — and that the data rule in this industry is stricter than in almost any other on this list.

Read the governance note before you try anything.

Start with the work, not with the apps

These are not twenty lists of AI apps. Each stack starts with the jobs a particular business performs, then matches AI and conventional software to those jobs. Treat it as a menu to evaluate, not a shopping list.

For this kind of business the work runs in one line, start to finish: Client → Documents → Accounting → Advisory → Billing

Broken into the steps a customer actually moves through:

Client → Documents → Accounting → Reconciliation → Reporting → Advisory → Tax → Billing

Read that line left to right. A client arrives, documents arrive after them, the bookkeeping happens, it reconciles, it reports, you advise on what the numbers mean, and then you bill.

The work most firms want back is the front of that line. Chasing documents and classifying them is the least valuable thing a qualified person does, and it is where most of the month goes.

The stack, piece by piece

Every tool below is matched to one of those steps. You do not need all of them, and you should not buy all of them at once. Read the “What it is” column first and find the step where your business is losing the most time or money.

ToolThe step it coversWhat it isPrice
QuickBooksClient accountingThe small-business accounting standard — books, invoicing, reporting.Per month by plan.
XeroAlternative client accountingCloud accounting with strong bank feeds and app ecosystem.Per month by plan.
DextDocument capturePulls receipts and invoices into your accounting system.Per client/month.
Microsoft 365ProductivityOffice apps, Teams, SharePoint and admin controls.Per user/month.
ChatGPT / ClaudeExplanations and internal assistanceOpenAI’s general assistant — writing, analysis, research and, with Agent mode, multi-step work.Free tier; paid plans per user/month.
CalendlyClient meetingsShared booking links tied to your calendar.Free tier; paid per user/month.
HubSpotCRMCustomer records, marketing, sales and service in one platform.Free CRM; paid hubs per month.

Where two names share a row with a slash between them, they do the same job and you pick one, not both. The description shows the first of the two.

If you are a one-person bookkeeping practice, the first three rows are the practice. The CRM row matters when you have more prospects than you can remember.

What to watch out for in each one

None of these are reasons not to buy. They are the things owners tell us they wish they had known first.

QuickBooks — Only as clean as your categorization habits.
Xero — Fewer US-specific integrations.
Dext — Client habits decide the quality.
Microsoft 365 — Licensing tiers get confusing.
ChatGPT — It sounds confident when it’s wrong. Check anything that touches money, law or a customer.
Claude — Same rule as any assistant: verify facts and never paste data you wouldn’t email.
Calendly — Set buffers or your day disappears.
HubSpot — Costs step up sharply by tier and contact count.

Where AI actually fits in this business

Document classification, financial explanations, anomaly investigation, report narratives, client education.

In plain terms, that means:

Sorting and coding the receipt pile, with a human reviewing the exceptions rather than every line.
Writing the plain-English explanation that goes on top of a management report, so the client reads it instead of filing it.
Investigating an odd variance by asking questions of the data, then checking the answer yourself.
Drafting the client education material — what a balance sheet is, why cash and profit differ — that you have explained a hundred times out loud and never written down.
Turning a messy client email into a clear list of what you still need from them.

Notice that none of those involve pasting a client’s books into a chatbot. The pattern that works is AI inside tools built for accounting, with proper data terms, plus general AI for work that contains no client data at all.

Governance check

Never feed confidential financial or taxpayer data into an unapproved AI system. Approved data controls first, convenience second.

In practice: approved tools with signed terms for anything containing client financial information. Anonymize or generalize before using anything else. "Explain depreciation in simple terms" is safe. "Here is my client’s trial balance" is not.

What to set up first

Do these in order. Finish one before you start the next. Most owners who try to do all of it in one month end up using none of it.

  1. Standardize the ledger. Pick one accounting platform as your default for new clients. A practice running four different systems is four practices, and nothing you automate will carry across them.
  2. Fix document collection. Automated receipt and invoice capture is the highest-return change available to a bookkeeping practice. It removes the chase and the keying in one step.
  3. Put client communication on rails. A booking link and a standard request list beats individually written emails asking for the same five things every month.
  4. Then add reporting and advisory. Once the data is clean and timely, monthly reporting with a written narrative is what turns a bookkeeping fee into an advisory fee.
  5. Add a CRM last. Useful when you are managing a pipeline of prospects rather than a waiting list.

What it costs to start

Most tools here are per user or per client file per month. Document capture is often priced per client, which means it scales with your book — budget it as a cost of service, not an overhead, and price it into the engagement.

The arithmetic that matters: what does an hour of your qualified time cost, and how many hours a month does the tool remove? Document capture usually pays for itself on the first client. Reporting tools pay off when they let you raise a fee, not when they save a few minutes.

Watch for tools priced per client file that you keep paying for after a client leaves. Review the list quarterly.

The price column in the table above tells you how each one charges — per user, per location, per contact, or a percentage of every sale — not what you will pay. That is deliberate. The charging model is what decides whether a tool stays affordable as you grow, and it is the part that does not change every quarter. Get the current dollar figure from the vendor’s own pricing page before you budget.

The rules underneath all of it

Legal · Privacy · Data protection · Security · Compliance · Records · Audit · Risk · Human oversight — underneath every one of the twenty stacks, not an afterthought.

Client financial data and taxpayer information carry duties that vary by jurisdiction and by credential. Never put confidential client financial data into an AI system that has not been approved and contracted for it.

If you handle tax information in the United States, there are specific confidentiality rules about disclosure and use. Your professional body’s guidance on AI is the document to read before your vendor’s marketing page.

Know whether each vendor trains on your input, where the data is stored, and what happens to it when you cancel.

Keep an access list. In a small practice people wear several hats; that is exactly how a former contractor keeps access to a client ledger for a year.

Anything AI produces that goes into a filing or a report is checked by a qualified person. The credential is yours, not the software’s.

Mistakes that cost the most

Pasting a client trial balance into a consumer chatbot to "just ask it something".
Running every client on a different platform because each one chose their own.
Keying in receipts manually in a year when capture tools cost less than the hours.
Delivering reports with no narrative, then wondering why clients do not see the value.
Letting document access outlive the engagement.

The CEO decision

How many hours a month does your practice spend chasing and classifying documents?
Could you name, right now, every system that holds client financial data and who can get into it?
Are you billing for bookkeeping or for advice? The tools you need differ.
What is your professional body’s current position on AI use, and have you read it?
Which clients are on a platform you would not have chosen? What would it take to move them?

Your quick review

☐ I have one default accounting platform for new clients.
☐ Receipts and invoices are captured automatically, not keyed.
☐ No client financial data goes into an unapproved AI tool.
☐ Every monthly report goes out with a written explanation.
☐ Access to client systems is reviewed when people leave.

Additional resources

QuickBooks — quickbooks.intuit.com — Clean books your accountant already knows how to read.
Xero — www.xero.com — Businesses who prefer it to QuickBooks, often outside the US.
Dext — dext.com — Firms drowning in client paperwork.
Microsoft 365 — www.microsoft.com/en-us/microsoft-365 — Microsoft-centered companies running the whole business on one suite.
ChatGPT / Claude — chatgpt.com / claude.ai — Your default thinking partner when you want one tool that does most jobs well.
Calendly — calendly.com — Ending the back-and-forth on meeting times.
HubSpot — www.hubspot.com — Small businesses that want one system of record.
Across the Table AI Assessment — acrossthetable.biz/#assessment

Go deeper on the pieces

Each of these is a full guide comparing the three tools that do that job:

Ask and reason — covers ChatGPT, Claude
Manage the money — covers QuickBooks, Xero
Fill the pipeline — covers HubSpot
Market by email — covers HubSpot

Continue learning

The Across the Table AI Assessment takes fifteen minutes and starts from the work your practice performs.

For a single piece, the guides above compare the leading tools in plain language.

Read and connect with Across the Table

Across the Table website: https://acrossthetable.biz/

The Scaling Edge newsletter: https://acrossthetablebiz.substack.com/

All Knowledge Articles