Build a budget that reforecasts itself weekly.

What the AI will do

• Connects the budget to evidence of new expenses and income in messages and operating systems.
• Builds and tests automatic reconciliation and refreshed forecasts on the cadence you choose: weekly, daily, hourly or when relevant data changes.

What you can provide

• The budget or where it lives, plus the business area and period it covers.
• Access to relevant finance records, expense and income systems, email and messages; existing exports can help begin.

Where you'll take part

• Confirm what the forecast measures, how uncertain items should be treated and what changes you want called out.
• Review a worked reconciliation, choose the refresh cadence and authorize the tested automation.

What you'll get

An inspectable budget-and-forecast system that explains changes and preserves the original budget. If a required connection is blocked, you receive the working reconciliation and forecast with its prepared integration, clearly marked as not yet continuously updating.

Activity context

A budget records what you expected. Then a purchase happens, a new opportunity appears, or an expense lands earlier than planned. Keep those changes connected to the budget as they happen, with refreshed forecasts as often as they are useful—weekly is one possibility, not the limit.

For the user

Start with the budget, its location or a rough description of what it covers. The AI will find the relevant records, work through a reconciliation and build the refresh loop. You will settle the few financial definitions it cannot recover, review a worked result and choose how often it should run.

Guidance for the AI

These are general instructions, not a description of this user's particular AI system, tools, access, or circumstances. Adapt the method to the user's goal and the capabilities you can actually use, while preserving the activity's essential sequence, review points, and intended result.

Start with the activity and the user's context

Read the complete activity and guide the user through it. Use the activity prompt yourself; the user does not need to paste it again. Keep the responsibilities of the user, the AI, and any colleague clear.

Begin with what the user has already supplied. Establish the intended result and resolve missing context through focused questions when needed. Work within the activity's stated scope, permissions, required confirmations, and stopping points, including any deliberate order for introducing information. Make progress with the available information and reasonable, clearly identified assumptions where the activity allows them.

Check the capabilities needed for this work

Inspect relevant available tools, connectors, and data. Use small, relevant, non-destructive checks to establish whether necessary access or a capability actually works. Naming a product does not establish access to it. Let this activity determine what needs checking; do not impose a universal preliminary questionnaire or a full system audit.

When useful access is missing:

• Explain what you cannot reach and how that material would help.
• Investigate practical setup, connection, export, or other access alternatives before asking the user to gather material or troubleshoot.
• If progress depends on access, request only the specific user action or permission needed to enable it or provide the material another way. Continue useful independent work while awaiting the response.
• Offer to walk the user through an export or upload. Welcome files, links, copied text, rough notes, or an interview as supported by your system and the activity, and use the best available route. If the missing access is optional, continue with what is available.

Ask focused questions and respect the answers

Ask questions only when necessary to complete the work well or satisfy the activity's required review points. Build on answers already supplied. Ask one question at a time, or a small group of closely related questions. Wait on work that depends on the answer, and continue useful work that does not. Explain unfamiliar language with examples where helpful.

Confirm answers or decisions you supply on the user's behalf before treating them as the user's final answers or decisions. This does not require separate approval of every sourced finding; follow the activity's specific review requirements.

Keep evidence and independent contributions distinct

Distinguish evidence, inference, and missing information. Check your work against the source material.

Use independent agents when your system actually supports delegation and independent investigation or review would improve the work:

• Give each agent a bounded brief and the context appropriate to its role.
• Keep discovery independent until synthesis where the activity requires it. Preserve any requirement for the user or colleagues to contribute before seeing other explanations.
• Follow the activity's stated fallback and any required confirmation when independent agents or fresh contexts are unavailable. Preserve reviewer isolation where required; describe shared-context or sequential passes accurately.

Produce, check, and revise the result

Do the substantive investigation and production. Carry the activity through to the most complete, usable result within its scope, rather than ending with a plan or handing the main work back to the user. When missing information does not prevent useful work, proceed with clearly identified assumptions where the activity permits them.

Check the result against the activity's completion criteria. At each substantive delivery, state what is complete and any material unfinished work, assumptions, helpful missing information, and remaining uncertainty. Explain how these affect what the user can use, and help the user revise the result.

Preserve the activity's stopping point. Do not turn a rehearsal, investigation, or test-design activity into execution that the user has not authorized.

Your AI prompt

Build a system that detects new expenses and sources of income, reconciles them with my budget and refreshes the forecast at a useful cadence. It should connect to the relevant records and keep running after this conversation, with weekly, daily, hourly or event-triggered updates as the actual sources support.

Understand the budget and its uses

Start with the budget or description I supply. Inspect the workbook or planning system, including formulas, period definitions, categories, assumptions and version history. Identify the business area, forecast horizon, currency, and whether the intended view concerns cash, revenue and expense, or several separately defined measures. Find who uses it and what decisions it informs in finance review notes, calendar invitations, leadership correspondence and prior forecast discussions.

Search relevant email, chats and other messages, including WhatsApp where accessible, for purchases, spending commitments, new opportunities, delayed projects, cancellations and changed expectations. Search expense tracking, purchasing, invoices, accounting records, sales or other income systems and contracts for the underlying events. Follow financial discussion threads and policy documents to recover the organization's definitions, treatment of uncertain income, timing assumptions and existing reconciliation method. Use records to understand the needs of the actual budget owner and report readers rather than inventing a generic finance dashboard.

Test access to each relevant source. Record the available date range, latest update, event identifiers and fields it provides. Investigate supported connections or exports before asking me to act. If automated access is unavailable, use supplied records for a worked reconciliation while preparing the missing connection; distinguish a manually refreshed result from continuous operation.

For a large scope, split bounded units or source families among agents with common definitions, source references and required outputs. Reconcile overlaps centrally. Use successive passes and explicit handoffs when capacity is limited. Ask focused questions one at a time about consequential definitions or gaps, then confirm the intended financial view before calculating it.

Make a traceable reconciliation

Preserve the approved budget as a dated baseline. Reproduce its relevant totals before extending it. Build a working event record linking each detected item to its sources, amount, currency, timing, category, status and budget line. Distinguish a discussion or opportunity from a commitment and from a posted transaction. A message saying “we may buy this” must not become a completed expense.

Match multiple records of the same event: an email, purchase order, invoice and payment may describe different stages of one purchase. Use source identifiers and reviewed matching rules so they do not become four expenses. Handle revisions, cancellations and refunds with explicit links to the earlier item. Send ambiguous matches or classifications to an exception view instead of guessing silently.

Reconcile the latest available actuals with their authoritative source totals, showing unmatched items, source gaps and timing differences. Apply the organization's confirmed rules for moving commitments and opportunities into the forecast. Keep amounts and periods inspectable. If accounting treatment is unresolved, present the precise item and question for the appropriate finance owner; do not invent a policy or edit source accounting records to force agreement.

Calculate the refreshed view

Implement the calculation in the actual workbook, planning environment or a connected tool that fits the existing workflow. Carry forward observed actuals and recalculate the remaining period from current commitments, income expectations and relevant drivers. Use the confirmed recognition and timing rules for each view, and avoid counting the same item in both actuals and remaining forecast.

Show the original budget, latest reconciled actuals, current forecast and changes since the previous forecast. Explain material changes through the underlying events: a new purchase, revised delivery date, new income or removed opportunity. Retain assumptions and prior forecast versions. Make uncertain items visible through their status and a useful sensitivity or range where appropriate; a more frequent update alone does not establish greater accuracy.

Where dated history permits, replay earlier cutoffs using only information available then, and compare the resulting forecasts with later actuals at the relevant horizon. Forecasting: Principles and Practice explains this use of rolling historical cutoffs. Report the usable periods, forecast errors and limits of that check. If historical source versions are missing, demonstrate calculation and reconciliation correctness without claiming established forecast accuracy.

Connect the automatic refresh

Inspect the existing automation environment and current official documentation for the chosen source and planning integrations. Build the source readers, update detection, reconciliation, recalculation and change report. Use supported event notifications where suitable or scheduled checks where necessary. Agree with me on the cadence and destinations: weekly, daily, hourly or after relevant source changes. Explain the actual source-refresh limits; hourly polling cannot make a monthly source current.

Define what runs automatically, what requires finance review and which changes merit an alert. Keep the live forecast separate from changes to the approved budget and source accounting records. This activity updates the agreed planning view; any authority to modify other records must be separately explicit. Configure credentials through the supported secure setup and ask only for actions I must personally complete.

Retain run history, source timestamps, matched events, exceptions, calculation versions and report delivery results. Make repeated runs safe from duplicate transactions and duplicate alerts. If a source is unavailable, show which figures are stale, preserve the last successful result and flag the failed refresh rather than presenting old data as new.

Test, activate and deliver

Test a new expense, new income, an uncertain opportunity, a duplicate source record, a revised amount or date, a cancellation and a source outage. Check arithmetic, category totals, timing and original-budget preservation. Verify that the calculated changes follow the confirmed rules and that reports reach their agreed destination. Where available, give a fresh reviewer the actual calculation and source records to find errors without your explanation; repair findings and repeat affected checks.

Review one worked reconciliation and forecast with me, including material exceptions, proposed cadence and automatic actions. Once authorized, activate the schedule or event connection and verify a complete run from retrieved source data through the updated forecast and report. Show the next scheduled run or active event subscription, last successful refresh and how to pause or change it.

Deliver the usable forecast, underlying formulas or code, source mapping, reconciliation and exception view, first run evidence and brief operating instructions. If an essential connection remains unavailable, deliver the working calculation and prepared integration with its exact remaining action; describe that limit rather than calling it a continuously updating system.

Done when

The budget stays available as the original baseline, new source events flow through a tested reconciliation into the current forecast, and an activated refresh produces an inspectable change report at the agreed cadence. Any disconnected source or untested forecast accuracy is stated plainly.

These instructions were written in September 2026 using the frontier models available at the time. You may now have better tools or models, or restrictions that make some tasks harder. Your AI should adapt to the tools and access you actually have and find the best available way to help. Everything here is a starting suggestion for you and your AI, not a proven best practice for every setup. Feel free to rewrite any of it to better meet your needs.