Rebuilding Enterprise Operating Capability with AI
After the monthly close, the finance team explains why actual spending differs from budget. The operating lead uses that commentary to decide which variances to investigate and which warrant action from a business team. When introducing AI to this workflow, start by defining the information the lead needs to make those decisions.
This research note uses a fictional month-end close to apply the source framework’s ideas on stakeholder value, the four SOP scales and evidence of results.
Work back from the management decision
Agree on the deliverable with the operating lead who uses it. For each material variance, state the amount, the basis for comparison, the causes already verified, the questions still open and the person responsible for following up. The lead needs enough information to decide whether action is warranted and a way to trace the supporting data. Finance and the business team agree on what counts as material and apply the same threshold throughout the pilot.
A one-off purchase and a supplier price increase can produce similar variances. The first might be fully explained by this month’s transactions, while the second could affect next quarter. Good commentary makes those implications clear and identifies any explanation that still needs checking.
Connect the operating outcome to concrete tasks
At the operating level, the desired result is an action list backed by evidence. The business process starts once the close is confirmed and runs through variance identification, fact-checking, finance review and delivery to the operating lead. At the functional level, the analyst gathers evidence, the finance manager accepts the commentary, and the business owner confirms causes and agrees follow-up actions. Each responsibility translates into tasks that someone can carry out and check.
For example, the analyst reconciles actual amounts to the correct budget version before identifying the accounts or transactions that drove the variance. AI can use agreed rules to flag items, organize references and draft commentary. The analyst verifies each explanation and marks where evidence is incomplete. The finance manager reviews the amounts, comparison basis and support before accepting the package.
Check the deliverable and readiness separately
The finance manager first checks the package against the acceptance criteria. A variance calculated against the wrong budget version goes back to the analyst for correction. An accepted package may still contain an item that the operating lead cannot resolve until the business team confirms a supplier’s renewal terms. Record that item as pending confirmation and assign an owner and a due date.
Keeping these states separate lets the team tell where time is lost. A problem with the commentary calls for an improvement in analysis or review. A missing downstream input calls for coordination with the team that can provide it.
Use evidence of results to decide whether to continue
Choose one expense category for the pilot. Keep a baseline for effort, rework and on-time delivery under the current process. Record the steps actually performed, how many items passed the first review, which explanations the operating lead used and whether they informed a follow-up decision. Count time spent on human verification, waiting and handling exceptions.
On the next close, ask a second analyst to work from the same playbook and AI skill. Check whether that analyst can meet the same standard. Expand the pilot when the outputs are reliable and reviewers can handle the work required. If the evidence is inconclusive, investigate the specific sources of variation before deciding to adjust or stop.
Source
Read the original research · Andrew David W.Y.