Strategic Finance Leadership · Insight Article
Are we delivering the business we intended?
Can AI connect strategic ambition, commercial commitments and operational execution so leadership can allocate capital, track results and change course while outcomes remain influenceable?
Leadership · Capital allocation · Strategy execution · Financial intelligence · Subscription economics · Commercial context · AI-era CFO
A leadership team can approve an investment in an hour. Working out whether it delivered what was promised can take rather longer.
The investment case sits in a presentation. Spending appears in the accounts. Future commitments sit in contracts. Progress is discussed in project meetings. Customer response appears in usage, renewals and correspondence.
Each source describes part of the initiative. Leadership needs to understand whether the pieces still add up to a worthwhile investment.
That is the question behind this research:
Can AI help connect strategic ambition, commercial commitments and operational execution—so a business can allocate capital, track results and change course while outcomes are still influenceable?
Why I am asking
In my own work, understanding financial performance has repeatedly meant bringing together information that was never designed to tell a joined-up story.
Accounting exports provided one view. Customer subscriptions and billing schedules provided another. Vendor invoices explained costs, often using different descriptions and billing periods. Partner arrangements introduced another set of commercial rules. Depreciation analyses and other schedules sat outside the accounting system.
I also used financial and customer data to examine subscription economics: customer retention, revenue retention, concentration, revenue quality and profitability.
The difficult questions lived between the sources.
A vendor invoice could be accurate. A customer invoice could also be accurate. Comparing them could still reveal that the business was paying for something it was not recovering from the customer.
Revenue growth could look encouraging while customer losses, discounts or concentration told a less comfortable story. An operational team could report a completed project while finance was still waiting for the evidence needed to bill it.
Each record could pass its own checks while the commercial relationship between them went unexamined.
An experienced CFO already knows a great deal
This research starts with respect for commercial judgement.
A good CFO may already suspect that a customer is unprofitable, a forecast is optimistic or an initiative is drifting. The challenge is assembling the supporting evidence, keeping it current and translating the conclusion into coordinated action.
The CEO needs to choose between competing opportunities. The board needs to understand whether the economics behind an approved strategy still hold. The executive team needs to know where intervention could change the outcome.
Those decisions cannot always wait for another reporting cycle.
The question is how AI might make experienced judgement better informed, easier to test and more actionable across an organisation.
A reporting pack can be accurate and still leave you guessing
Much financial analysis begins with a snapshot: export the accounts, add supporting schedules, reconcile differences and prepare an explanation.
Before that explanation reaches the decision-maker, someone extracts the data. Someone else maps it. Another person supplies a spreadsheet. Operations explains an exception. Finance makes an adjustment.
Every person contributes knowledge. Every handover can also lose some.
Which version was used? Was the customer change reflected? Does the cost relate to this month or next month? Was the assumption confirmed, or did it simply survive enough meetings to become accepted fact?
By the time the answer arrives, the business may have moved on.
A margin issue can continue through another billing cycle. Completed work can sit unbilled. A renewal can pass before anyone connects the supplier cost with the customer’s current requirement.
The eventual explanation may be correct. The opportunity to act may have become smaller.
The missing connection is often the original promise
Actual versus budget is useful, but a budget rarely contains the whole investment argument.
An initiative might have been approved because it would reduce service costs, improve retention or release capacity. Spending within budget does not establish that any of those benefits occurred.
Leadership needs to follow the chain:
What did we approve? What assumptions supported it? What did we commit? What has been delivered? What benefit has appeared? What should we now expect?
A system implementation can be on budget while the intended operating savings remain absent. Expenditure can also exceed budget while delivering an attractive incremental return. The explanation matters more than the colour of the variance.
That requires preserving the original assumptions and subsequent decisions. If the original plan quietly disappears into the latest forecast, tracking performance becomes an impressive exercise in moving the goalposts.
What leadership should be able to ask
| Decision | Evidence to connect | What the answer should support |
|---|---|---|
| Fund, stage or stop an initiative | Investment case, assumptions, spending, commitments, delivery milestones and benefits | Whether to release more capital, revise scope, pause or stop |
| Protect customer profitability | Agreements, pricing, supplier costs, usage, delivery effort and invoices | Where to reprice, change scope or address delivery costs |
| Understand subscription economics | Customer cohorts, billing, cancellations, expansion, discounts, costs and renewals | How to respond to retention, concentration and revenue-quality risks |
| Turn delivery into cash | Completion evidence, acceptance correspondence, billing terms, invoices and receipts | Which blocker to resolve, who owns it and when cash is realistically expected |
| Track benefits against commitments | Original targets, operational results, financial actuals and forecasts | Whether a variance reflects timing, execution or a flawed assumption |
| Choose between competing investments | Pipeline, capacity, hiring plans, cash forecasts and existing commitments | Which opportunities to prioritise and which demands to defer |
These are capabilities to build and test. My current transaction experiments examine a small part of the foundation: whether connected evidence can help explain what a record represents and reveal where the first answer is incomplete.
They do not yet demonstrate the full decision capability described here.
“Rubbish in, rubbish out” is only part of the problem
Often, the information is reasonable within its own boundaries.
Sales knows what the customer agreed. Operations knows what was delivered. Procurement knows what was ordered. Finance knows what was posted.
Nobody necessarily holds the whole picture.
Commercial context also changes. A customer’s requirements may change after signing. A service cancellation may be agreed in email while the supplier charge continues. A new commitment may affect cash before an expense appears in the accounts.
An assistant receiving only an accounting export inherits its boundaries. It may produce an excellent explanation of an incomplete picture.
Fluency does not fill the missing evidence.
A useful answer must distinguish recorded actuals, contractual commitments, operational signals and forecasts. A customer expressing concern is not yet a cancellation. A purchase commitment is not an expense already posted. Both may matter to the next decision.
Connected systems should help. Why does change still feel so difficult?
My experience working around Business Central brought another issue into focus: a capable system can feel like a black box to people unfamiliar with its configuration.
Changing a process may require specialist knowledge of mappings, permissions and existing dependencies. People become cautious because they cannot easily see what a change will affect.
That caution is understandable. Teams have workarounds they know how to operate. A new process may promise better information while introducing another specialist, another set of rules and another dependency to maintain.
People are being asked to change how they work while continuing to deliver the work.
An AI layer could make that easier—or add a more articulate black box.
If nobody can explain where an answer came from, how its assumptions were formed or who maintains the connections, the organisation has acquired another thing that needs holding together.
Faster answers need a route to action
A useful connected assistant should help leadership ask:
- What changed since we approved this?
- Which assumption is now under pressure?
- What costs are already committed but have not reached the accounts?
- Which benefits are evidenced?
- What happens to cash, capacity and returns if we continue, change scope or stop?
- Who needs to do what next—and did the action produce the expected result?
The response might support a revised forecast, a staged funding decision, a pricing review or an assigned investigation. Execution still needs appropriate responsibility and authority.
“Real time” also needs a definition. An answer assembled instantly from last month’s records is still last month’s answer, delivered with excellent service.
Leadership needs to know how current the evidence is, what is missing and which conclusions depend on assumptions.
Why the experiments start with small gaps
The broader ambition depends on foundations we can inspect.
Can an assistant find supporting correspondence? Can it challenge an accounting description? Can it explain a net receipt? Can it recognise when evidence conflicts or is insufficient?
The experiments examine different application workflows and connected assistants against those questions. Each records the components used, what was observed and what remains untested.
Assessing whether records reflect commercial reality is one necessary step towards evaluating whether the business is delivering its intentions.
The standard I want to work towards is practical:
Can leadership understand an important issue sooner, trace the explanation to evidence and act while it can still influence the outcome?
That is worth pursuing.
Preferably before the next reporting pack explains why we should have acted last month.
Supporting evidence
First-person reflection on working with disconnected accounting exports, customer and subscription data, vendor invoices, partner arrangements and external analytical schedules. Published transaction experiments examine selected foundations of cross-system investigation; they do not demonstrate the complete leadership capability proposed here.
Sources
- Author’s experience and ongoing experiments
Anonymised first-person reflections and the separately published Xero experiments. No company identities, internal project names, private correspondence links or private financial amounts are included.
Counterarguments
Connected sources can remain stale, incomplete or contradictory. Specialist dependencies, data ownership, access controls and maintenance remain material. Better explanations do not automatically produce better decisions or execution. No quantified reduction from weeks to real-time leadership insight has been established.
Current conclusion
Reliable evidence and financial completeness are foundations for a larger leadership capability: testing investment assumptions, tracking economic outcomes and intervening earlier. This is a research proposition rather than a claim of a completed system.
How my thinking has changed
The research extends from explaining recorded transactions to supporting investment choices, execution and accountability by connecting intentions, commitments, operational evidence and outcomes.