Connect
Add Spark to Slack, connect Xero through OAuth. That is the setup.
Spark, by Practora
Spark prepares bills, journals and workpapers as Xero drafts, straight from Slack. It cannot post, pay or authorise—that stays with you.
Interface
Slack-native requests
System of record
Xero stays the ledger
Authority
Draft-only architecture
# month-end
June close - workpapers and approvals

Message #month-end
Drafts only. Nothing posts to Xero without your approval.
How it works
No new platform to learn, no migration, no logins for the team. Xero stays your source of truth - the work happens where finance already talks.
Add Spark to Slack, connect Xero through OAuth. That is the setup.
Spark reads your Xero history, coding patterns, and open items - so its first drafts already look like your team's work.
Bills, journals, and workpapers arrive ready for review. You approve in Slack before anything posts.
Not another chatbot
A chatbot gives you an answer. Spark gives you the deliverable - drafted, supported, waiting for your approval. The work is done, not described.
A generic AI tool
Explains what an accrual journal is.
Spark
Drafts the journal in Xero with the workpaper behind it, and waits for your sign-off.
A generic AI tool
Tells you how a bank reconciliation works.
Spark
Builds the reconciliation from your actual ledger and hands you the Excel.
A generic AI tool
Summarises what a P&L shows.
Spark
Pulls your P&L and posts the file in the thread. Ask a follow-up, get an answer.
Month-end
Ten minutes a bill. An afternoon a rec. Two days of workpapers. Every month. Tools that stop at invoice capture or approval routing leave the rest with you. Spark prepares it all - from the bill to the workpaper to the reporting question after - and hands the hours back.
built from your ledger, formulas included
schedule calculated, journals drafted
matched, documented, ready to review
checked against your AP ledger, exceptions flagged
the return support, prepared not pasted
calculated with the working shown
an editable spreadsheet, not a screenshot
P&L, balance sheet, aged payables - in the thread
one view of what is done and what is left
Glass box controls
Workpapers with live formulas, not magic numbers. Drafts only - Spark never posts, pays, or authorises on its own. Every action logged: who asked, who approved, what changed.
Connects via Xero OAuth. Your ledger stays in Xero. Revoke access anytime. SOC 2 in progress.
The review contract
How every Spark workflow runs today.
Named approver on every draft
Nothing reaches Xero on AI judgement alone.
Edit or reject before posting
The reviewer changes the coding, not the other way round.
Append-only audit log
Who asked, who approved, what changed - kept for good.
What Spark learns
Supplier coding rules
Approval thresholds
Recurring journals
Entity quirks
Preferred workpaper formats
Who signs off on what
It learns the approved version
Spark learns from the work you approve: supplier coding, thresholds, evidence formats and who signs off. The rules stay visible and editable.
FAQ
Straight answers on time saved, accuracy, and control.
A finance teammate in Slack for teams on Xero. It reads invoices, drafts the bills and journals, builds the workpapers behind them, and answers reporting questions in the thread. A named person approves everything.
The preparation hours - the biggest share of every close. Bills arrive coded, workpapers arrive built, reports arrive on demand. Your team's time moves from building the numbers to reviewing them - the work that actually needs a finance brain.
It removes the places errors hide: no re-keying, coding applied consistently from your own history, live formulas instead of pasted numbers, duplicate bills flagged before they land. And because everything is a draft, mistakes are caught in review - not found in the audit.
Capture tools stop once the invoice is in. Approval tools stop once the bill is routed. Spark does the intake and the approval, then keeps going - journals, workpapers, reconciliations, and reporting questions - in the same Slack thread, with every action logged.
No. Xero stays your ledger and your source of truth. Spark prepares work in it as drafts and never posts on its own - nothing to migrate, and no new system for the rest of the team to learn.
The close checklist shows what is done and what is outstanding, per entity, so nothing slips quietly. Deadline reminders and automated document chasing are in development with pilot users.
No. Everything arrives as a draft. A named person reviews, edits if needed, and approves before anything touches the ledger.
Two connections: add Spark to Slack and connect Xero through OAuth. Spark reads your history and coding patterns and starts preparing work for review.
Spark runs in Slack today. Microsoft Teams is planned, with the same approval-first model.
In progress. Today: OAuth-only access, encrypted transport, tenant isolation, audit logs, and human approval on every write.
Start with one close task
Connect Xero, get a free health check on your books, and decide where Spark starts. Approval-first from day one.