Last Updated: June 2026
XeroForce setup starts before you open the builder: XeroForce is Xero's no-code agent builder for financial workflows on your Xero data, currently in invite-only alpha with broader release later in 2026. To build agents you can trust, first get your data foundation right, with every receipt and invoice captured, coded to your chart of accounts, and free of duplicates.
What XeroForce Is, and What It Isn't
Xero introduced XeroForce in May 2026 as a natural-language custom agent builder. You describe a repeatable process in plain language, including review steps, client sign-off limits, and specific tax rules, and XeroForce turns it into a durable workflow. No code, no technical setup. An agent can be live within seconds.
The use cases Xero demonstrated at launch are the ones that eat your week: month-end close, ongoing reporting, organising tax documents, purchase-order validation, payrun approval, and reconciling uncoded statement lines. Agents are always-on. They watch for events like an email reply or a filing date and complete tasks in the background over days or weeks, rather than waiting for you to trigger each step. Crucially for practice work, every action is logged and traceable, and workflows can run across your whole client base at once rather than one conversation at a time.
It helps to place XeroForce in context. It is one of three AI products on Xero OS, alongside JAX (Xero's financial superagent) and the Xero plus Claude integration. Xero's framing is that it is moving from a system of record to a system of action, and XeroForce is the build-your-own layer of that shift.
What XeroForce is not is a document collector. It orchestrates work on the data already inside Xero. It is the conductor, not the source of the music. An agent told to "organise this client's tax documents each month" can only organise the documents that have actually made it into Xero.
One practical note before you invest time designing agents: XeroForce is still alpha and invite-only. If you are not in the program yet, the waitlist is open, and the most useful thing you can do while you wait is get your data house in order. That work pays off the day your invite arrives.
Why an Agent Is Only as Good as the Data It Reads
Here is the principle that should govern your XeroForce setup: an agent's output quality is capped by its input data quality. What makes an agent riskier than a person doing the same task is scale. A bookkeeper reconciling statement lines by hand pauses when something looks off; an agent running practice-wide across forty clients does not. It applies the same logic to every client, every month, on whatever data is present. If half a client's expenses never reached Xero because they were buried in a Gmail thread, the month-end summary is wrong, and it is wrong silently. Worse, the action is logged and traceable, so the audit trail documents the error rather than catching it.
This is why getting the data right comes first. Reporting, organising tax documents, and reconciliation, the three workflows Xero highlights most, all depend on completeness. A reporting agent that sees only 80 percent of spend produces a number that looks precise and is materially false.
The Data Foundation to Get Right First
Three things determine whether your Xero data is ready to support an agent. Work through them in order.
1. Complete document capture
Every deductible expense and every payable needs a corresponding document in Xero. The hard part is that financial documents do not arrive in one place. Subscription receipts land in email. A tradesperson photographs a fuel receipt on their phone. A supplier invoice comes as a PDF attachment. Xero's native capture handles manual upload, email-in forwarding, and the mobile app, but it does not connect directly to a Gmail, Outlook, or IMAP inbox and pull documents out automatically.
Receiptor collects financial documents from supported email sources, WhatsApp, iMessage and uploads. Define how each source reaches the document workflow, how exceptions are reviewed, and how the evidence reaches the appropriate Xero record.
2. Accurate coding against the chart of accounts
Capture gets the document into Xero. Coding makes it usable. An agent asked to report spend by category, or to flag unusual tax-document activity, relies on transactions being mapped to the right chart-of-accounts code. Miscoded transactions do not just produce a wrong line in one report. They train the agent's view of "normal" for that client, so anomalies get harder to spot over time.
The goal before building is a high rate of correct coding at the point of entry, not coding you fix later. Manual recoding does not scale to practice-wide agents, and it reintroduces exactly the human bottleneck XeroForce is meant to remove.
3. Clean, deduplicated records
Duplicates are the quiet killer of agent reliability. The same invoice forwarded twice, or captured once from email and once from a photo, becomes two transactions. A reconciliation agent now has a line it cannot match. A reporting agent overstates spend. A tax-document agent files the same receipt twice. None of these failures are dramatic, which is why they survive until an agent compounds them across every client.
Before building, your records should be deduplicated at the source, so the agent reasons over one clean set of transactions per client rather than a set you hope is clean.
Where the Documents Actually Come From
Closing the capture and coding gap is its own workflow, and it sits upstream of anything you build in XeroForce. The practical question is how you keep it complete without hand-keying everything.
Receiptor checks applicable line-item calculations, subtotals, tax, discounts, and totals. Validation problems can trigger AI review against the source and another check. Unresolved discrepancies remain flagged for review. Legal entity assignment helps identify the billed company, and each entity can use its corresponding accounting connection. Duplicate copies and related invoice/payment documents are handled as different evidence for the transaction. Receiptor supports Xero Bills, Spend Money transactions, and Bill payments, with checks for supported existing records and originals attached to matched or newly created records.
Receiptor prepares document evidence for supported Xero workflows. Review the captured documents, company context, and accounting records before using downstream automation that relies on them.
Your XeroForce Setup Readiness Checklist
This is the XeroForce setup work to finish before you design your first agent. Confirm:
- Capture is complete. Every channel that produces a financial document, email, messaging, paper, and direct upload, has a defined path into Xero. At any real volume that means an ingestion layer like Receiptor AI doing the capture, not manual forwarding that quietly misses things.
- Coding is accurate at entry. Transactions are mapped to the correct chart-of-accounts code when they post, not corrected weeks later.
- Records reflect the transaction. Distinguish duplicate copies from related invoice and payment evidence, and inspect the accounting result for the purchase.
- The chart of accounts is current. Agents reason against it, so a stale or inconsistent chart produces stale, inconsistent output.
- You have decided what stays human. XeroForce lets you build in review steps and sign-off limits. Define those before the agent goes live, especially for anything client-facing or tax-relevant.
Work through this list once per client before building, and revisit it whenever you onboard a new one.
Limitations: When This Guidance Doesn't Fully Apply
A few honest caveats. XeroForce is in invite-only alpha as of June 2026, so exact capabilities and availability may shift before general release. Treat Xero's published use cases as the current scope, not a guarantee.
This readiness approach also assumes your practice runs on Xero as the system of record. If a client keeps significant financial activity outside Xero, in spreadsheets or a separate system, an agent built on Xero data will not see it, and no amount of capture hygiene fixes that. And if your workflow is genuinely low-volume, a handful of documents a month that you already enter cleanly by hand, the data-foundation work matters far less, because there is little for an agent to get wrong. The completeness problem scales with volume and with the number of clients. The more of either you have, the more this guidance applies.
