Smart document capture.

Millions of documents a month since launching in 2026. I led the strategy, research and design, and handed over in March 2025.

MY ROLE

Lead product designer Hubdoc and Data-In

WHEN

2024 – Mar 2025

WITH

A PM, a researcher and two engineering leads

LAUNCHED

Mar 2026

Decided: Build document capture into Xero, not keep investing in Hubdoc.
After launch: It shipped. The team launched it a year after I handed over, and Hubdoc still runs alongside it.
Decided: Start with matching documents to bank transactions.
After launch: It worked. After launch, people acted on far more of the documents they uploaded.
Decided: Put capture inside Files.
After launch: I got this wrong. Bookkeepers work in Accounting, so they couldn’t find it. The team has since moved it there.

Why it mattered

Around half of Hubdoc documents never made it into Xero. Xero wanted to stop relying on a separate app for capture.

These include bank statements needed for reconciliation, supplier contracts, and invoices that don't need to be in Xero, plus junk or duplicates. Xero is missing an opportunity to help customers organize their books more easily.

Every receipt and bill has to end up as a transaction in the books. A bookkeeper uploads the document, checks what was read from it, categorizes it, and matches it to the bank payment. With Hubdoc, the first three steps happened in one app and the last in another.

  1. 01

    In Hubdoc

    Upload

    Receipts and bills, by photo or email

  2. 02

    In Hubdoc

    Check

    Confirm what was read from each one

  3. 03

    In Hubdoc

    Categorize

    Say what kind of expense it is (coding)

  4. About half never crossed over

    04

    In Xero

    Match

    Pair it with the bank payment (reconciling)

What was undefined

I ran 17 interviews to find out what “document management” meant to the small businesses and bookkeepers using it.

What we thought

Build a document-management tool and a better filing system in Xero.

We assumed

Users would match documents from the bills workflow: open a bill, find the document, attach it.

Users wanted a better place to organize and store all their documents inside Xero.

What users needed

A way to process every document fast and accurately, and turn it into the right transaction.

What research found

Matching was “a huge efficiency gain, but only for bank reconciliation.” Bills without a receipt rarely got past draft, so there was nothing to match them to.

Users wanted an empty inbox, a place where documents come in, become transactions and get cleared out.

Customers cared most about getting every document into the books, all in one place. They wanted fewer tools and a simpler workflow. But moving away from what they already knew was a big hurdle. Switching costs were especially high for bookkeepers.

The challengeHow could we provide enough value to make the switch worthwhile?

Three decisions

Decision oneKeep investing in Hubdoc, or build capture into Xero?

  • Not chosen

    Keep investing in Hubdoc

    Keeps loyal accountants happy, no migration. But the problems came from it being separate.

  • Not chosen

    Tighten the sync

    Cheaper, and keeps both products. But customers still set up two.

  • Chosen

    Build natively in Xero

    Fixes the root cause. One product that works with Xero’s own rules.

Every problem we found traced back to capture living outside Xero.

The plan was to keep investing in Hubdoc. I made the case to build capture into Xero instead.

Hubdoc still had loyal customers, so retiring it right away made no sense. In 2024 I turned our research into a prioritization paper and took a milestone plan to the engineering leads. In early 2025 I pitched an end-to-end vision to product leadership, and they agreed. We also got a neighboring team to build capture into their workflow, so bookkeepers could match documents where they already work.

Looking back, building capture into Xero seems obvious. Keeping Hubdoc, or even improving it, still left many documents outside Xero. That meant lost files, duplicates and a clunky two-app workflow. Every major Hubdoc problem came from it being a separate product.

But the trade-off was running both products in parallel for years, and many accountants saw little reason to switch.

I set the direction and handed over in March 2025. The team launched it a year later, and Hubdoc still runs alongside it.

Uploading straight into Xero, where each document arrives with its match already found.

Decision twoWhich workflow should we improve first?

  • Not chosen

    In the bills workflow

    What the team assumed. But bills without a receipt rarely got past draft.

  • Chosen

    In bank reconciliation

    Where bookkeepers start their day. Shipped first, in Hubdoc.

  • Not chosen

    A standalone inbox first

    One place for everything. But we didn’t have the evidence for it yet.

Bank rec first, because that’s where bookkeepers start. The inbox came after uploads showed people wanted it.

We learned bookkeepers start with bank rec because it’s easy but tedious, making it the biggest opportunity for Smart Document Capture. So we built document upload, data extraction, and matching to the bank feed, which noticeably increased uploads. When users needed to manage documents outside bank rec, I designed an inbox to manage the rest, since testing showed little need for storage and raised permission concerns.

The trade-off: we depended heavily on other teams and their roadmaps; figuring out how we could contribute and pushing to build in their domain was a worthwhile challenge in the end.

Matching a document in bank rec.

Decision threeHow do we earn users’ trust in AI results?

  • Not chosen

    Create transactions automatically

    Fastest. But a wrong value lands in the books without anyone seeing it.

  • Not chosen

    Check a form of fields

    Familiar. But there’s no way to see where a value came from.

  • Chosen

    Review beside the source

    Document and fields side by side. Click the document to fix a value.

We staged the trust: read-only first, then editable, then automation.

I went with review beside the source: documents displayed next to extracted fields, making it easy to see and fix values. I rolled this out in stages: first read-only, then editable, with automation added once users built trust.

Data supported this: surveys and tests showed users valued seeing document sources as much as a central inbox.

The trade-off: review slows down high-volume workflows that prefer full automation. The review step was built so more automation could follow once people trusted it.

The Review state as shipped, the document beside its extracted fields.

Before

Hubdoc reviewing a store receipt, with a Publish All button sending it to QuickBooks

Hubdoc, a separate app that published documents out to the books.

After

Xero Files reviewing a receipt beside its extracted fields, with a spend money match found

Capture inside Xero, with the match already found.

Outcomes and how it shipped

  1. March 2026

    UK, 100% live

  2. 4 June

    North America and rest of world

  3. 6 July

    Australia and New Zealand open beta, then everywhere

  4. September

    Renamed Documents, moved into the Accounting menu

  5. October

    Homepage widget

Millions

Documents uploaded a month

More than 2×

Uploads turned into transactions

Global

Live in every region in 2026

CUSTOMER QUOTES

“

Data extraction looks really good though, picking up payment method, descriptions etc.

“

The new files experience is great to see. We will be recommending this to our clients going forward.

“

As much as I love some 3P tools, having it within Xero will streamline doc management for us.

Looking back

After launch, people acted on far more of the documents they uploaded.

I designed capture inside Files, which sat in the organization menu. Bookkeepers work in Accounting, so after launch they couldn’t find it. The team has since moved it there.