Is Xero Smart Document Capture Enough Without Dext?

Xero Smart Document Capture reads receipts and bills you upload, email or photograph, extracts the details, then attaches the document to a matching transaction or drafts a new one for you to approve. For simple supplier invoices it may be enough without Dext or Hubdoc. Itemised receipts, statements and odd formats are where you should test before dropping either.

The catch is that Xero's help pages show the steps but never show a failure, and a bookkeeper's risk lives in the failures. The sensible question is not "does it work?" but "which of my clients' documents does it get wrong, and what does a wrong one cost me to fix?" A 30-document test run in one afternoon answers it, and it usually shows that the decision turns on one or two document types rather than on capture in general.

Follow me on Instagram@sagnikteaches

What the capture does, in the order it does it

Xero's September 2026 post, announcing that Files is now called Documents, describes four moves. Capture: you upload a file, email it in, or snap a photo of a receipt or bill. Extract: Xero reads the details off the page. Match or create: it either attaches the document to a transaction that already exists, such as a bank-feed line, or starts a new one. Review: the post says every extraction, every match and every decision is still yours to review and approve.

Connect on LinkedInSagnik Bhattacharya

A few smaller facts from the same post matter for a practice. Auto-matching can be turned on or off at any time. Credit notes are handled from extraction through to drafting the transaction. Team members can be given upload or view-only access without seeing other financial data that has been uploaded. Xero warns you if you try to delete a document that is linked to a transaction. And the feature now lives under the Accounting menu rather than the organisation-name menu; existing documents were not affected by the move.

Subscribe on YouTube@sagnikteaches

Two points are easy to get wrong when you describe this to a client. First, the feature launched in July 2026 and Xero's pricing page lists it on Early, Growing and Established alike, which is why nobody is quoting a price for it. The launch post also said it was still in beta in some markets, so confirm it is fully live in each client's edition of Xero. Second, the rename on 18 September 2026 means older screenshots, client guides and your own checklists that say "Files" now point at the wrong menu. If you wrote onboarding notes this year, search them for the word.

What Xero has not published, and why that matters

The Xero material I could read does not give supported file types, upload size limits, page limits or a list of the fields extracted. One user on Xero's Product Ideas forum reports, in a July 2026 request, that capture stops at ten pages and that a long invoice came through as a single line item. That is a user's report, not a documented limit, and Xero's July launch post points the other way on lines: it lists multi-line, itemised extraction, naming rental statements and itemised receipts, as already available. When a vendor claim and a user report disagree, that is the exact spot to test.

Xero's own list of what is coming next names supplier rules and a bespoke forwarding email address for "the coming months", and, further ahead, creating bills straight into the awaiting-approval or awaiting-payment states, bank statement extraction, and notes and item codes on documents. The two posts also disagree slightly about tax: the July launch post listed tax extraction as coming, and the 18 September post describes pulling the local tax rate from uploaded documents as live. Tax number validation was described as planned for two markets only and not yet live. When sources disagree like this, run the test rather than pick a side.

For comparison, Dext's business pricing page lists line item extraction, supplier statement extraction and bank statement extraction as separate add-ons, each with a pay-as-you-go option. That tells you where Dext thinks the work gets hard. It is a useful checklist even if you never buy any of them, and it is the reason the test below includes statements and itemised receipts.

A 30-document bench test you can finish in an afternoon

Take one client, ideally one who already runs Dext, and pull thirty documents from the last two months. Do not choose the tidy ones. Choose by type, so every kind of paperwork that client produces appears three or four times. Anonymise or use documents the client has agreed you can use. Then run the same thirty through Xero's capture in a demonstration or test organisation, or in the live one if you can safely discard the drafts, and compare against what Dext produced for the same files.

Score each document on five things and nothing else, otherwise the sheet gets unreadable by lunchtime.

• Supplier: right name, and the same name you already use in the contact list, not a near-duplicate that creates a new contact.
• Date and total: both exactly right, including the invoice date versus the due date.
• Tax: the tax amount and rate right, or at least visibly wrong rather than quietly plausible.
• Lines and coding: one line or several, and whether the account suggested is one you would accept.
• Match: attached to the correct bank transaction, or left alone, not attached to the wrong one.

The last item deserves emphasis. A wrong match is worse than no match, because it looks finished. In your sheet, mark a wrong attachment as a failure and an unmatched document as a pass.

Doc # | Type            | Supplier OK | Date+total OK | Tax OK | Lines OK | Match OK | Minutes to fix
01    | Supplier PDF    | Y           | Y             | Y      | n/a      | Y        | 0
02    | Till receipt    | Y           | N (total)     | N      | n/a      | -        | 2
03    | Statement       | N           | N             | N      | N        | -        | 4
Pass = all Y or n/a. Fix time = minutes to correct, not to review.

The main worked example: a bookkeeper's florist client

This is an illustration with invented figures. Say a bookkeeper looks after eleven small clients and wants to know whether the florist can come off a paid capture tool. The florist generates about 85 documents a month: wholesale flower invoices, van fuel receipts, packaging orders, card-machine statements, and the occasional handwritten market-stall receipt.

She samples thirty documents: ten wholesaler invoices, six fuel and till receipts, five packaging invoices, four supplier statements, three credit notes and two handwritten receipts. After running both tools, her illustrative scores look like this.

On those invented numbers Xero passes 23 of 30 and Dext 27 of 30, which is a 77% and a 90% pass rate. The headline gap is small. The useful part is where it sits: nearly all of Xero's lost ground is in the four supplier statements and one supplier-name mismatch. For a florist with 85 documents a month, the invoices and receipts are roughly 75 of them, and Xero handled those almost as well as Dext.

Now the arithmetic. Say each failed document costs four minutes to fix, and Xero fails 25% of the time on the ones that matter against 10% for Dext. On 85 documents that is about 21 against 9 documents to fix, a difference of 12 documents, or roughly 48 minutes a month. At an illustrative internal rate of $40 an hour, that is about $32 a month of your time. Dext's Business plan is listed at $302.50 a year for 250 documents a month and five users, which Dext shows as $25.21 a month, but that excludes add-ons such as line item extraction from $20.50 a month, and a practice plan prices each client differently. On these made-up numbers the florist is roughly a coin toss, which is exactly the answer worth having: it tells you the decision is about statements, not about capture in general.

Where capture slips: six document types to include on purpose

These are the places I would aim the test, drawn from what vendors list as separate add-ons and what the forum reports. Treat each as a hypothesis to check, not a finding.

Itemised receipts that need splitting

A hardware-store receipt with paint, brushes and a drill is one total, three accounts. If capture produces a single line, you are coding the whole thing to one account or splitting it by hand. Xero's launch post says multi-line, itemised extraction is built in; Dext sells line item extraction as an add-on beyond a small allowance of free credits on its Business plan. Either way it is a claim to test: in your sheet, count lines extracted against lines on the paper.

Supplier statements

A statement lists many invoices and is not itself a bill. A tool that reads it as a bill creates a bogus liability. Dext lists supplier statement extraction to check invoices against the statement and flag gaps and duplicates. Xero's own pages do not claim that job, so score whether Xero leaves statements alone or drafts something wrong.

Multi-page invoices

Run at least one invoice with a long schedule or several pages. The user report of a ten-page ceiling is the reason; if a client's suppliers send 20-page invoices, put two of them in the sample. Check whether the total comes from the last page, not the first subtotal.

Foreign-currency documents

Include a document in another currency if the client has any. Check the currency code, the symbol and whether the amount is posted in the right currency or silently converted.

Photos taken in poor conditions

Snap a receipt at an angle, in shadow, folded once. Clients do not photograph documents the way you would. If the extraction is confidently wrong, rather than blank, that is a bigger risk than failing outright.

Emails with the invoice in the body

A long-running request on the same forum asks for Documents to accept emails with no attachment, and a user comment there says the email body is ignored, which matters for suppliers that send receipts only as HTML emails. Forward one such email and see what lands in Documents.

A second client, to show the rule giving the opposite answer

Take a driving school with four instructors, again with invented figures. It produces about 60 documents a month: fuel-card receipts, a monthly lease invoice for each car, insurance instalments and a few repair bills. There are no supplier statements to speak of, and almost every document is a clean PDF from a large company that formats invoices identically every month.

The bookkeeper samples twenty documents instead of thirty, because the variety is low, and Xero passes 19 of them, a 95% rate. The one failure is a garage bill with parts and labour on separate lines, captured as a single line. Fixing it takes about three minutes. Fixing time across the month is perhaps ten minutes, against a paid tool that would cost more than that in fees. Here the rule says Xero alone, and the saving is the whole subscription share for this client.

Notice what the two examples have in common: neither needed a view on which tool is better. The florist's statements and the driving school's clean PDFs pushed the answers in different directions. A before-and-after for the garage bill shows the kind of slip to expect and the fix:

Paper: Brake pads 62.00 | Fitting labour 45.00 | Shop supplies 6.50 | Tax 22.60 | Total 136.10
Captured (illustrative): 1 line, "Garage repairs", 113.50, tax 22.60, total 136.10
Fix: split into parts, labour, supplies; keep the total unchanged; save supplier rule if available

The total and tax were right, so a quick glance would pass it. Only the line count gives it away, which is why the scoring sheet has a separate column for lines.

Writing the decision down so the client sees it

After scoring, set a rule before you look at who "won". A reasonable one: Xero alone for a client when at least 90% of documents pass and the failures are cheap to fix, a paid tool for a client when statements, itemised receipts or high volumes cause more than about an hour of fixing a month, and a split approach otherwise. Use the sheet's last column, minutes to fix, as the tiebreaker because it prices the failure rather than counting it.

It also helps to separate "can capture the document" from "should be trusted to post". Keep your existing review step either way. On a client's plan, consider a simple threshold: any bill over a figure you choose gets a second look before approval, whichever tool created it. That rule costs nothing and survives a change of software.

Client note (illustrative):
Hi [first name], from [date] your receipts and bills will go through Xero's built-in capture
instead of a separate app. What changes for you: forward or photograph documents as before,
and use Documents (it used to be called Files) in Xero if you want to look one up.
What doesn't change: I review everything before it posts. Statements from suppliers
should still come to me directly, not to the capture inbox.
[your name]

Cost sums to run before you cancel anything

Capture inside Xero is included, so the saving is whatever you pay for the separate tool minus the extra fixing time. One catch on the cheapest plan: Xero's pricing page limits Early to entering five bills (it does not state the period), so a client on Early whose capture would draft dozens of bills a month needs Growing anyway, and that upgrade belongs in the sum. List what each client currently costs. A Dext Business subscription is $302.50 a year for 250 documents a month and five users, with a few free credits for line item, supplier statement and bank statement extraction, and the page lists add-ons: line item extraction and supplier statement extraction from $20.50 a month each or $0.50 a document, bank statement extraction from $13.00 a month or $0.32 a document, and an AI Assist add-on at $10.50 a month until 31 January 2027. Practice plans are priced differently, so see the Dext pricing breakdown for accountants before you compare. If your clients sit on a practice plan, the saving from dropping one client may be small because the fee is shared.

Two sums to do. One: the monthly fee for the clients you would move, divided by the number of documents, which gives a cost per document. Two: expected fix minutes per month times your internal rate. If the second sum is larger than the first, keep the paid tool for that client. Add the one-off cost of retraining the client, which is usually an hour of explaining where things went, and the risk that a half-migrated client sends documents to both places and you get duplicates.

Moving one client without breaking the books

Do it in a quiet month, not at year end. Switch one client first and keep the old tool running in parallel for a single cycle, with a rule that only one of the two may publish to Xero so documents are not posted twice. Check the duplicate warnings, then compare the bank-reconciled totals at month end. Xero's auto-matching can be switched off; if a client's bank feed is messy, start with it off and attach documents by hand until you trust the matches.

Remember Hubdoc here. Hubdoc retired its standard fetch connections, the ones that logged into supplier portals to collect bills, in April 2022. Its website still offers sign-ups, but Xero's new capture features are going into Documents, so do not promise a client a Hubdoc workflow you have not checked. Open the client's Xero organisation, look at what is connected, and describe what is actually there. The wider comparison of those two paid tools is in Dext versus Hubdoc, which is useful if the test says you still need one of them.

How to check it kept working after the first month

Capture quality drifts when suppliers change invoice layouts. Re-run a shortened test, ten documents, every quarter, and log two numbers per client: the share of documents posted without correction, and the minutes spent fixing. If the first falls under 85% or the second doubles, re-open the decision. Also watch the contact list: a growing pile of near-duplicate suppliers is the quiet sign that extraction is spelling names differently each time.

Pair this with your reconciliation checks. If you use Xero's automatic matching on the bank side, the JAX bank reconciliation setup and checks tutorial covers how to spot matches that look right and are not, and the receipt capture and bank categorisation setup shows the same logic from the coding side.

When the answer is "keep both"

Plenty of practices will end up with both. A client with 15 invoices a month and one itemised fuel card does not need a paid capture tool. A building firm with hundreds of supplier invoices and monthly statements probably does. The test gives you a documented reason either way, and the reason is what you tell the client. If a client asks why their fee changed, "I tested 30 of your documents and here is what failed" is a better conversation than "the software got cheaper".

Smart Document Capture questions bookkeepers ask

Does Smart Document Capture cost extra?

Xero launched it in July 2026, and its pricing page lists it on all three plans with no separate charge. What can still cost you is time spent correcting documents, the Early plan's limit of five bills, and add-ons you keep buying elsewhere. Check each client's plan before telling them it is free.

Will it create bills without me approving them?

Xero's own wording is that every extraction, match and decision stays yours to review and approve. Treat that as the design: drafts wait for you. Test it with one client before relying on it, and keep your usual review step for anything over a threshold you set.

Is Hubdoc still an option?

Hubdoc's own website still offers sign-ups and still describes capture into Xero and QuickBooks Online, but Xero's new capture work is going into Documents. Hubdoc retired its standard fetch connections, which logged into supplier portals, in April 2022. If a client still depends on Hubdoc, confirm what is connected in their Xero organisation before changing anything.

How many documents should the bench test use?

Thirty is a practical minimum: enough for each document type to appear three or four times, small enough to finish in an afternoon. Use real, anonymised or consented client paperwork, and include the awkward ones on purpose, because the tidy ones will pass in any tool.

Further reads

• Dext vs AutoEntry for a Small Practice: Costs, Accuracy Claims and How to Check Them — If Xero's capture falls short, the next comparison to run.
• Best AI Accounting Agents: QuickBooks, Xero JAX and Sage Copilot — Where capture sits among the wider accounting AI agents.
• Xero vs QuickBooks AI: Which Saves More Bookkeeping Time? — The same capture question, asked across the two big platforms.
• Is AI Worth It for a One-Person Bookkeeping Business? — Whether the saved minutes matter for a one-person practice.
• ChatGPT Prompts for Bookkeepers: Client Queries, Chasers, Notes — Client chasers and queries that follow once capture is settled.
• How to Use Xero's JAX for Invoices and Cash-Flow Questions — Another Xero AI feature worth testing the same way.
• Can AI Do a Restaurant's Bookkeeping? What a Person Still Checks — One restaurant month worked through: what AI bookkeeping handles for receipts, deposits and sales, and what a person still checks on tips, tax and stock.
• Best AI Receipt Scanning App for UK Tradesmen: Xero, Tradify, Powered Now — A head-to-head of three receipt scanners for sole-trader tradespeople: what each reads, which plan you need and a ten-receipt test.
• Handyman Accounting Software: Receipt Capture and Invoicing AI Compared — QuickBooks, Xero and FreshBooks compared for one-person handyman work: invoice caps, receipt capture, mileage and what the AI features actually do.
• Best AI for Accountants in Australia: Tools, Privacy and ATO Checks — Which AI tools suit an Australian accounting practice, what the Privacy Act and TPB guidance expect, and the human checks before anything reaches the ATO.
• Sage Copilot Pricing: Which Sage Plans Include It, by Region — Sage Copilot costs nothing extra on the UK Accounting plans, but receipt capture has limits and the US page only offers a price request.
• AI Bookkeeping in the UK: What Works With MTD and VAT — Four MTD and VAT tests for any AI bookkeeping tool, applied to Xero, QuickBooks, Sage and Dext, plus a test pack that catches silent coding errors.
• AI for Australian Bookkeepers: BAS Prep, GST Checks and Xero — A quarter-end BAS review that pairs Xero's automation with invoice checks, reconciliations and a documented sign-off.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Xero product updates (the July 2026 smart document capture launch post and the 18 September 2026 'Files is now called Documents' post), Xero pricing plans page, Xero Product Ideas forum, Xero's Hubdoc fetch-connections post, Hubdoc website, Dext business pricing page.