What Clients Actually Ask Their Bookkeeper: The 18 Questions, Sorted
253 questions typed into a live bookkeeping demo, grouped into 18 themes, and sorted into what a lookup answers and what needs an accountant's judgment.
On this page +
See it live
Ask your finance system anything.
Invoices, AR aging, unpaid bills, payments.
Schedule free onboardingNo signup. Runs in your browser.
Most people asked for status, not accounting advice: who owes us money, which invoices are late, whether a customer paid, what revenue looks like for a period, and which bills or vendors need attention.
The finding comes from 253 typed rows in Kipper’s public bookkeeping demo between June 8 and September 15, 2026. Those rows contained 232 distinct wordings, but once grouped by what the person was actually asking, they collapsed into 18 questions.
Data verified: September 16, 2026. Source was an internal export of the public interactive demo question log. The export had 487 total rows and 258 exact unique strings.

Method
This is a public demo that uses Intuit’s QuickBooks sample company. The people asking were prospects trying the product, not a partner firm’s private clients. The post treats the log as evidence of what people ask when handed a bookkeeping question box, not as a literal transcript of client emails to a firm.
The demo includes starter prompts and suggested follow-ups. Those clicks are useful product telemetry, but they are not typed questions. We excluded them before counting the data set.
| Bucket | Rows | Unique strings | Treatment |
|---|---|---|---|
| Starter chips | 187 | 5 | Excluded |
| Likely suggested follow-ups | 47 | 21 | Excluded |
| Typed questions | 253 | 232 | Counted |
The rule was conservative: repeated, formally punctuated strings were treated as chips. Informal variants stayed in. For example, How many customers do I have? appeared 9 times and looks like a button. how many customers do I have? appeared once and stayed in the typed set.
The strongest signal was bank balance. People asked for it 29 times across 15 exact strings, which reduce to 8 patterns once case and punctuation are stripped: what is our bank balance, what is the bank balance, what is my bank balance, what is my total bank balance, what's my bank balance, what is our bank account balance, what is my bank account balance, and show me our bank account balance. A thirtieth row asked for the checking account balance instead, which is why the table below counts 30. No suggestion chip produces that spread.
The 18 questions below are an editorial grouping over those 253 typed rows. The hard counts are 253 rows and 232 exact strings. The grouping asks a different question: when two prompts use different wording but want the same thing, should a firm experience them as the same interruption? Reasonable reviewers might group a few edge cases differently, so the table shows row counts and keeps the arithmetic visible.
Before display, questions were lightly normalized: sample-company names became [customer] or [vendor], obvious identifying text was removed, casing and punctuation were cleaned up, and the 16 rows the export truncated at roughly 80 characters were not used as examples. Do not read the displayed questions as verbatim prompts.
The 18 questions
The 18 grouped questions cover 215 of the 253 typed rows. Another 21 rows were product probing, greetings, and test messages. The remaining 17 rows were one-off fragments or follow-on continuations that did not group cleanly. Together: 215 plus 21 plus 17 equals 253.
| # | Question | Rows | Class |
|---|---|---|---|
| 1 | Show me invoices for a period | 36 | Answerable |
| 2 | Show me everything for [customer] | 20 | Answerable |
| 3 | Which invoices are overdue? | 14 | Answerable |
| 4 | Has [customer] paid, are they current? | 14 | Answerable |
| 5 | What is our revenue for a period? | 14 | Answerable with scope |
| 6 | How many customers, vendors, or products do we have? | 15 | Answerable |
| 7 | What are our inventory levels? | 15 | Answerable |
| 8 | What bills are open or due? | 10 | Answerable |
| 9 | Who are our top customers? | 9 | Answerable with scope |
| 10 | What is our total AR, and how much is past due? | 8 | Answerable |
| 11 | Has this specific invoice been paid? | 6 | Answerable |
| 12 | Which items sell the most? | 5 | Answerable |
| 13 | Who owes us the most? | 4 | Answerable |
| 14 | What did we spend with [vendor]? | 3 | Answerable with scope |
| 15 | What is our bank balance? | 30 | Answerable |
| 16 | Show me the P&L, net income, or a GL extract | 4 | Partly answerable |
| 17 | What needs reconciling or cleaning up? | 4 | Out of scope |
| 18 | Change something in the books | 4 | Write request |
Fifteen of the grouped questions are answerable lookups or answerable with clearer scope, one is partly answerable, and two are not. The largest single group, bank balance, is one Kipper answers.
That is the post in one table. A client-facing answer layer is valuable because many interruptions are simple record lookups. It is trustworthy only if it also knows when the question belongs back with the firm.
AR and customer status
Invoices, AR, and customer status dominate the useful part of the log. People asked for invoices by period, overdue invoices, named customer history, whether a customer was current, who owed the most, and whether a specific invoice had been paid.
The wording varied, but the work underneath was usually plain: filter invoice records, join payment status, sum open balances, and show the supporting rows. These are the questions that do not need a senior bookkeeper in the loop every time.
For a firm, this is the low-friction win. A client asking “is [customer] current?” is not trying to run a report. They want a yes, a balance, and the invoices behind it.
Revenue, counts, AP, and inventory
The next answerable cluster is a mix of revenue, counts, AP, vendors, and inventory. People asked how many customers or products existed, who the top customers were, what bills were open, what they spent with a vendor, which items sold most, and what inventory was available.
Some of those questions need scope. “What is our revenue for a period?” is answerable if the system and firm agree whether that means invoiced sales, recognized revenue, or a report-level P&L number. “What did we spend with [vendor]?” is usually a vendor-bill lookup. “What did we spend by category?” depends on whether the connected system carries that category as a Class, Tracking Category, or Department.
Inventory also has a clean boundary. “What are our inventory levels?” and “which items sold the most?” are lookups and sums over item, invoice line, and inventory records. “What are we going to run out of?” is prediction and belongs with a person.
Bank, P&L, cleanup, and writes
Thirty typed rows asked for bank or cash balances, which makes it the single largest group in the log. Others asked for P&L, net income, a GL extract, reconciliation, cleanup, or a change to the books.

Bank and cash balances are lookups, and Kipper answers them: it reads the balance the accounting system itself reports. So is the P&L, and so is net income, which comes off it. A GL extract is not, and neither are the individual bank transactions underneath the balance, imported statements, or reconciliation status.
The boundary has moved but it has not disappeared, and where it now sits is the more useful line. Reconciliation, cleanup and close work stay with the firm, because they are jobs someone has to be accountable for rather than numbers someone needs to read. Cash planning stays there too: Kipper can tell you the balance and what you spent last month, but the judgment about what that means for next month is the firm’s.
Write requests are a separate boundary. The log included requests to send a chasing email, create a journal entry, mark an overdue invoice paid, and delete a record. A read-only answer layer should not carry those requests into the books at all.
That boundary keeps broad client access sane. A client can safely ask about open invoices or unpaid bills because those are record lookups. Bank balances, reporting, reconciliation, cleanup, and writes are different work.
Product probing and residual rows
The log also had test messages and off-topic prompts. Public demos always do: hi, yes, okay, what tools are available, and broader product-comparison questions.
Those rows are useful evidence of demo behavior, but they are not bookkeeping questions. The 18-question table excludes them from the displayed set and discloses the count instead.
The residual 17 rows were one-off fragments, continuations, or prompts that did not group cleanly without over-interpreting them. Keeping them out of the table is part of the same rule: the headline counts what the page can defend.
What a lookup layer can answer instantly
A bookkeeping answer layer is at its best when the question has four traits:
- It asks about existing records.
- It gives enough scope, or the system can ask for the missing scope.
- The answer can cite the invoices, bills, payments, customers, vendors, purchase orders, items, or inventory records behind it.
- Nothing in the accounting system changes.
Most of the useful typed questions in the log fit that shape. So does the strongest use case for Kipper for Firms: clients and their staff ask plain questions, the firm controls access, and Kipper answers only from the supported read-only record layer. Kipper holds a read-only connection to QuickBooks, Xero, or NetSuite and answers in Slack, Microsoft Teams, SMS, or an MCP client. It does not replace the firm.
If you want a curated list of clean prompts, start with 30 questions you can ask AI about your QuickBooks data. This post is doing the opposite job. It keeps the rough edges because the rough edges are the evidence.
What this means for client accounting services
For a firm selling client accounting services, the log points to a simple split.
The easy half is interruption load: invoice status, customer balances, overdue bills, vendor lists, item lookups, and period comparisons. These questions are short, repetitive, and urgent to the person asking. They do not need a partner or senior bookkeeper in the loop every time.
The hard half is still the firm’s work: judgment, cleanup, reconciliation, advice, cash planning, reporting, and anything that changes the books. Those are the moments where a person should stay accountable.
The product shape is straightforward: answer the lookup questions quickly, give the firm visibility into what was asked, and hand off the questions that need judgment instead of making a chatbot improvise.
FAQ
What questions do clients usually ask their bookkeeper? In this demo log, most questions collapsed into a small set of status checks: which invoices are overdue, whether a customer has paid, what revenue looks like for a period, which bills are due, and what vendors or customers have the largest balances.
Which bookkeeping questions can software answer? Software can answer direct lookups, filters, sums, and date-range comparisons when the underlying records support the question. Many invoice, payment, bill, vendor, customer, purchase order, item, and inventory questions fit that pattern.
Which questions should still go to an accountant? Questions about judgment, advice, cleanup, reconciliation, cash position, P&L, the general ledger, or changing records should stay with the firm. A read-only lookup layer should not pretend to be an accountant or write back to the books.
Can clients ask these questions without a QuickBooks, Xero, or NetSuite login? Yes, if the firm gives them a controlled answer layer such as Kipper. The accounting system is connected once, permissions are set for each person or workspace, and clients ask in plain language instead of logging into the finance system.