10-lesson course · about 14 hours

Build proactive finance agents that prepare reporting updates you can trace

For CFOs, controllers, FP&A teams, accountants and finance operations professionals

Everything for Finance teams

Set up data, reconciliation, variance, forecast and Reporting Watch agents that work from accepted figures, route exceptions to a finance reviewer and can be paused or retired.

Try lesson 1 freeSee what you’ll learn

What you’ll be able to do

By the end, you can…

Turn arriving files into one traceable datasetDefine an intake agent that sorts each batch into new, restated, duplicate or rejected files. It builds a versioned dataset where reporting period, currency, units and source stay visible on every row, and rejected or duplicate files stay out of totals.
Reconcile every accepted version without silent fixesBuild a reconciliation agent that runs agreed matching rules and tolerances. It records each check and its evidence in a ledger matching workbook and tie-out log, and sends exceptions to a reviewer instead of making adjustments.
Keep variance commentary tied to evidenceCreate a variance review agent that refreshes the review when reconciled results change. It produces a driver bridge that adds up to the total, and it keeps a possible cause out of management commentary until evidence supports it.
Refresh forecasts without blurring the modelSet up a driver-update agent that recalculates an approved sample cash model only when an approved assumption changes. The model, the assumption register and the resulting scenarios stay separate, so you can see which assumption set each figure came from.
Run a Reporting Watch agent on a schedule you agreeConfigure a named Reporting Watch role on OpenAI Dots or Claude Duet. You set agreed intervals, change triggers, approved connections, agreed reports and a named reviewer, and choose between a single run and a continuing brief.
Test, then pause or retire with a clear recordTest your agents against missing, invalid, extreme, wrong-unit, stale and late inputs across several simulated periods. Then decide whether to continue, adjust, pause or retire the work, and record why.

How it works

A tutor that meets you where you are

  1. Lessons that adapt to your levelTen lessons across five modules follow the reporting cycle from intake to the management pack. Each lesson adjusts its explanations and examples to what you already know about reconciliations, variances and forecasting.
  2. An AI tutor you can ask anythingAsk questions by typing or out loud while you work: why a tolerance matters, how to word an evidence request, or whether a single run or a continuing brief suits your case. The tutor takes the place of a live trainer.
  3. Quizzes and flashcards for each lessonShort quizzes and flashcards tied to each lesson help you check key distinctions. Examples include observation versus supported explanation, and model versus assumptions versus scenarios.
  4. Progress tracking and a certificateTrack which lessons and outputs you have completed, and receive a certificate of completion when you finish the course.

Curriculum

The curriculum

The course follows one small reporting cycle, built from sample data, from start to finish. Each module adds one proactive agent and a practical output, and the final module connects them around a recurring management pack.

Module 1Keep the reporting inputs current2 lessons

  1. 1Define the data proactive intake agent
    • Explain how a proactive agent differs from a chat assistant that waits for a prompt
    • Write an intake agent brief covering role, inputs, acceptance rules, approved connections, outputs, reviewer and pause or retire steps
    • Design a target dataset layout that keeps reporting period, currency, units and source visible on every row
    • Choose between a single run and a continuing brief for the first version of the agent
  2. 2Process a second reporting batch
    • Classify each file in a successive batch as new, restated, duplicate or rejected
    • Build a versioned dataset that records what changed between batches instead of overwriting earlier data
    • Test that rejected and duplicate files stay out of totals
    • Explain which dataset versions may be passed on to downstream agents

Module 2Catch differences before reporting2 lessons

  1. 3Reconcile an accepted dataset
    • Define which comparisons the reconciliation agent performs and what triggers it
    • Build matching rules and a tolerance agreed by the finance owner
    • Produce a Ledger matching (XLSX) workbook and a Tie-out log (CSV) that record every check and its evidence
    • Explain why the agent identifies differences but does not adjust them
  2. 4Resolve or escalate exceptions
    • Build an exception record that captures amount, evidence, possible cause, owner, status and age
    • Choose escalation rules based on size, age and how certain the cause is
    • Produce an Open-item list (PDF) for a finance reviewer
    • Explain how proposed adjustments reach a person for approval instead of being posted by the agent

Module 3Keep commentary tied to the numbers2 lessons

  1. 5Generate the recurring variance review
    • Define the trigger that starts a fresh variance review
    • Build a variance table with stated comparisons, sign convention and materiality threshold
    • Produce a Driver bridge (CSV) that adds up to the total variance
    • Explain how the agent marks changed figures and changed commentary between reviews
  2. 6Request support for explanations
    • Explain the difference between an observation, a possible cause and a supported explanation
    • Build evidence requests that say which figure, which question and what evidence would support an answer
    • Apply a rule that keeps unsupported causes out of management commentary
    • Produce Variance commentary (DOCX) with each statement linked to its evidence

Module 4Refresh scenarios within clear limits2 lessons

  1. 7Build a driver-update proactive agent
    • Explain the difference between the model, its assumptions and the resulting scenarios
    • Build an Assumption register (CSV) that records owner, value, source, date, status and version for each driver
    • Configure the agent to recalculate the Cash model (XLSX) only when an approved assumption changes
    • Produce a Scenario grid (XLSX) and a Forecast summary (PDF) that show which assumption set each figure came from
  2. 8Test missing and extreme inputs
    • Build a set of test cases covering missing, invalid, extreme, wrong-unit and stale inputs
    • Choose validation ranges with the assumption owner and write them into the agent brief
    • Test whether the agent stops, flags or proceeds as expected for each case
    • Record test results and retest after changing the brief

Module 5Run the reporting cycle end to end2 lessons

  1. 9Assemble the recurring management pack
    • Write a Reporting Watch brief with agreed intervals, change triggers, agreed reports, approved connections and a named reviewer
    • Choose OpenAI Dots or Claude Duet and a single run or continuing brief, and explain the choice
    • Build a pack that shows the data version, date and status of every item and flags stale or provisional inputs
    • Explain how a person pauses or retires the Reporting Watch agent and what record is left behind
  2. 10Test late data and recovery
    • Test the full cycle across several simulated periods
    • Trace a late correction through intake, reconciliation, variance, forecast and the management pack
    • Test that the agents hold back, flag and recover correctly after missing data or an interrupted run
    • Choose whether to continue, adjust, pause or retire the Reporting Watch work, and record why

Who it’s for

Made for you if…

Made for: CFOs, controllers, FP&A teams, accountants and finance operations professionals

Controllers and accountants

You want successive data batches and reconciliations handled consistently, with every difference logged and evidenced and no adjustment made without a person's approval.

FP&A analysts

You want variance reviews and driver-based forecasts refreshed when accepted figures change, while commentary stays tied to evidence and assumptions stay visible.

CFOs and finance leaders

You want to understand how recurring preparation work can be given to AI agents within agreed limits, with review gates and a clear way to pause or retire them.

Finance operations staff

You run parts of the monthly cycle and want to stop arriving files from piling up as an unexplained collection of spreadsheets.

Questions

Good to know

Do I need programming experience?

No. You need no software development background. The course covers only as much prompt writing as the exercises need.

What should I know before starting?

You should be comfortable with formulas, lookups and pivot tables in a spreadsheet. You should know how a monthly cycle of actuals, budget, variance and a management pack works, and be able to explain a variance and its likely drivers in plain words. You should also have some hands-on experience with an AI chat tool.

Will the agents connect to my ledger, ERP or bank feeds?

No. The course uses sample data files, an approved sample model and simulated periods. Live connections to production systems are not part of it.

Can the agents post journal entries or change source records?

No. Agents in this course identify differences, preserve evidence and propose adjustments for a person to approve. They do not post entries or change source records.

Which AI tools does the course use?

For the Reporting Watch agent, you choose either OpenAI Dots or Claude Duet and explain your choice.

Does it cover month-end close, statutory reporting, tax or audit?

No. It does not cover real month-end close, statutory or regulatory reporting, tax positions, audit opinions, internal control certification, or valuation beyond the sample model.

How long does it take, and is there a live trainer?

The course takes about 14 hours and is self-paced. There are no live sessions. The AI tutor teaches you one-on-one.

Should I use my company's real data?

The exercises use sample data. The course also teaches you to keep customer identifiers, account credentials and nonpublic reporting information out of unapproved cloud tools.

Build and supervise your own Reporting Watch agent

$397

One payment · lifetime access

  • 10 self-paced lessons in 5 modules, about 14 hours
  • An AI tutor you can ask anything, typed or out loud
  • Quizzes and flashcards tied to each lesson
  • Practical outputs: ledger matching workbook, tie-out log, open-item list, driver bridge, variance commentary, cash model, scenario grid, assumption register and forecast summary
  • Simulated periods to test late data, recovery and ongoing runs
  • Progress tracking and a certificate of completion

Secure checkout by Stripe. Have a promotion code? Enter it at checkout.

Refunds: Full refund if you ask before you open your first lesson. Once you’ve started, purchases aren’t refundable, except where the law gives you that right.

Already bought it? Sign in