PROMPT YOUR FINANCEYour close isn't slow. The chasing is.
Ask what is blocking the books. Payhawk's Agent works the list with you until every expense is coded, documented, reviewed, export-ready.

A prompt does a task. A playbook does the job
Run the month-end close without the chase.
Ask once. It works these steps the same way every month.
Finds and fixes every expense missing a category, tax rate or custom field.
Nudges the owner directly in Payhawk, so you write no reminder emails.
Clears the approvals waiting on you and releases expenses to export.
Names the template and the period, asks for confirmation and then runs the export.
Flags anything the ERP rejected, explains why, fixes what it can, re-runs the export.
Reads each corrected expense back from the ERP, and names what is still in flight.
Gives a shareable close summary: what was cleared, what is open, and who has it.

"I love the MCP connector. We have it in Business Central and Payhawk. It does checks on vendors, how we previously categorised expenses and provides a list of all expenses that may need to be allocated to a different GL and can then overrule it in BC."
A playbook runs as you, inside Payhawk's own controls. It cannot reach data your role cannot see, nor take an action your role cannot take.
- Runs with your own Payhawk permissions, never wider
- Previews every change, writing nothing until you confirm
- Names the actions it can take before it starts
- Records every action in your normal audit trail
- Never starts a payment or a transfer
Get your playbook walkthrough
Book a demo to see how you can close your books with one conversation inside Claude, ChatGPT and in Payhawk.
FAQ
A playbook is a finance job written down as a set of steps. Ask for one of Payhawk's official playbooks by name, and your AI Assistant works through those steps in order, the same way every time.
A plain question gets you data back. A playbook carries Payhawk's procedure for the job: which checks to run, in what order, what finished looks like. That procedural knowledge is what separates a playbook from a general connection to your finance data.
Yes, in the parts that need judgement. The sequence of steps is declared up front rather than improvised, and the steps that analyse, such as budget burn or benchmarking, call Payhawk Insights for the reasoning. Every step that would change your data still waits for your confirmation.
No change happens without your confirmation. Each playbook names the actions its steps can take before it runs, works only within your own Payhawk permissions, and records everything it does in your normal audit trail.
The playbook is not the automation, it is the follow-through. Payhawk's rules already code and route most spend; this works the remainder automation cannot decide: the missing receipt only one employee can produce, the approval sitting with a manager, the export the ERP bounced. That tail is where close time actually goes.
You are offered the playbooks your own permissions allow, resolved role by role and account by account. A colleague with different permissions, or working in a different entity, sees a different list.
Yes. Describe the job in your own words and Payhawk drafts the complete playbook for you, built only from actions you can already take. Nothing is saved until you review it and confirm.
No. An ordinary question, such as how many expenses are pending your review, gets an ordinary answer. A playbook runs only when you ask for the job itself, on request, never on a schedule.
In the Payhawk Web Portal, the Payhawk Mobile App, or the AI Assistant your company already uses. There is nothing for IT to configure first.
No. You ask in plain language, for example 'get us ready to close January', inside the AI assistant connected to Payhawk. Any technical steps you may have seen are the instructions the assistant follows, not something you run or maintain.