a
Insight · FP&A & Planning

When the Forecast Is Ready
but the Decision Is Not

Faster analysis creates limited value if the business still waits to agree the assumptions behind it.

FP&A & Planning Decision Inc. Australia 9 min read
Andrew Jerogin
Andrew Jerogin
Engagement Manager, Decision Inc. Australia
1MO
Budgeting cycle for a global mining technology client, down from three months, after implementing Workday Adaptive Planning
2WKS
Length of our Planning Diagnostic to map the planning estate and find the first intervention

Faster analysis creates limited value if the business still waits to agree the assumptions behind it. The starting point for planning AI should be the work between a business change and an accepted financial position, including what can proceed automatically and what requires judgement.

01

What the Forecast Doesn't Resolve

A major customer contract slips into the next quarter, just as the business is considering additional recruitment and delivery expenditure. Finance produces a revised revenue scenario, but leadership cannot yet decide which commitments to change. Sales believes the delay can be recovered, operations has allocated people, and the contract determines when invoicing can begin. The numbers are available before the business has agreed what they should represent.

Giving the analyst a faster way to produce those numbers addresses only part of the problem. The remaining work involves finding evidence, establishing which assumptions are acceptable and securing authority to act on them.

Our position is that this interval should be measured before the next planning AI investment is approved.

It reveals whether the organisation needs more analytical capacity or a different way of moving work towards a decision.
02

Measure the Delay the Model Can't Explain

A planning calendar records when a forecast is due, but it may reveal little about why reaching that date consumes so much effort. A submission can arrive on time and still be returned repeatedly because its assumptions are unclear. A completed analysis can wait for an approver who lacks the evidence needed to accept it. Counting the hours spent inside the planning application misses the work that determines when its output becomes usable.

The complete response, not the calendar

The useful unit of measurement is the complete response to a business change. It begins when the change becomes known and ends when the relevant owner accepts a financial position or authorises a decision. Within that interval, leadership should distinguish active analysis, waiting for information, reconciliation and review. These activities have different causes, so combining them into a general complaint about slow forecasting makes the investment case less precise.

Where delay is justified

Some delay is justified. A material customer commitment may require negotiation, and a significant change to expenditure may deserve careful challenge. The objective is to identify where judgement improves the decision and where people are compensating for missing information or unclear responsibility. Reducing the latter creates room for the former without treating every approval as an obstacle to remove.

03

Make Assumptions Usable Before You Delegate

Experienced finance teams can resolve inconsistent information because they understand what different records mean. A sales opportunity date may express commercial expectation, while a planning date records the assumption finance has accepted. An operations schedule may describe a resource commitment that cannot move with either. These records can coexist legitimately, but an assistant needs explicit context to use them appropriately.

Connecting the systems does not establish which date should govern the forecast. The business must define what confirms a change, which evidence is sufficient and who resolves a disagreement. Those decisions need to be represented in the workflow alongside the data. Otherwise, a more capable assistant may simply retrieve competing answers faster and leave finance to reconstruct the reasoning.

The opportunity for Autonomous Finance grows as organisations make these boundaries explicit enough to enforce, rather than relying on individual memory.

This is the operating change behind meaningful delegation. A defined rule can automate a repeatable step; an AI assistant can help a planner assess uncertainty; an agent can carry authorised work within agreed boundaries. For the delayed contract, gathering evidence and preparing scenarios may be suitable for delegation while the disputed assumption remains with its business owner.

04

What the Mining Client Case Shows

Our work with a mining technology client shows the value of addressing planning structures and participation together. The organisation operated across countries and currencies, with dispersed finance teams and inconsistent approaches. Manual budgeting and consolidation made the cycle demanding, while business units and central finance lacked a common way to collaborate. The challenge extended beyond the speed of a calculation.

We implemented Workday Adaptive Planning with consistent salary structures, cost drivers and planning assumptions across regions. Shared submission and review workflows allowed business units to contribute while central finance retained oversight. The client reported a reduction in its budgeting cycle from three months to one, alongside reduced manual consolidation and greater capacity for analysis. The engagement demonstrates a planning improvement; it does not establish an AI benefit or isolate assumption approval as the cause of the result.

That evidence supports examining the whole process before selecting the intervention. An organisation waiting for missing actuals needs a different response from one spending days reconciling competing assumptions. Where the model itself is slow, optimisation may be justified. Where responsibility is unclear, adding another analytical tool leaves the underlying decision unresolved, regardless of how persuasive its output appears.

05

Give Each Adaptive Capability a Defined Job

Workday Adaptive Planning provides capabilities that can improve specific parts of this sequence. Their value should be assessed against the work identified in the cycle, with clear criteria for what improvement would look like. Predictive modelling, Adaptive Decision Intelligence and MCP connections are relevant in different ways. Treating them as one undifferentiated AI proposition obscures what each can actually change.

Predictive Modelling Predictive modelling helps finance challenge the forecast using patterns in historical data and other relevant inputs. Workday provides confidence measures that support assessment of the prediction. A useful evaluation compares its performance with the existing approach and examines where judgement changes the result. A prediction can inform the contract scenario, but it cannot establish whether a customer has made a new commitment.
Adaptive Decision Intelligence Adaptive Decision Intelligence supports what-if analysis and submission of changes through existing security and approval chains. It can help planners compare the consequences of the original and revised dates while their status remains clear. That analysis may allow leadership to make a conditional decision before the uncertainty is fully resolved. The business must still decide which assumption enters the accepted plan and who has authority to select it.
MCP Connections MCP makes planning information available to external AI environments such as Claude, ChatGPT and Gemini. Workday's listed Adaptive MCP tools retrieve planning data and model context, rather than write forecast changes. This can support evidence gathering, but an assistant's answer is not a saved or approved reforecast. With production delivery listed for 16 October 2026, organisations should confirm availability and configuration before committing to deployment.
06

Design the Handover From Evidence to Authority

The next design question concerns what happens after information has been assembled. For the delayed contract, a proposed workflow could present the competing dates, identify their sources and show the financial consequences of each. It should also make clear which assumption has been approved and which remains provisional. The reviewer can then assess the unresolved judgement without rebuilding the entire analysis.

Where the evidence conflicts, the workflow should send the issue to the person authorised to resolve it, together with the reason for escalation. A recorded decision would establish which assumption subsequent steps may use. Any execution beyond that point would require appropriate integration, permissions and approval controls, including a way to stop or correct an action. These are requirements for the wider workflow, not functions supplied independently by the listed MCP retrieval tools.

Access security is necessary, but not sufficient

Access security remains necessary throughout. Workday states that Adaptive MCP enforces API permissions and access rules and records calls in audit logs. Those controls govern retrieval; the organisation must additionally decide how external services retain information and how answers are reviewed. Legitimate access to two different assumptions does not determine which one is appropriate for the decision.

Keep the CIO's role in view

The CIO's role is to ensure that these requirements do not become a separate set of definitions and controls inside every new application. Shared customer identifiers, documented billing terms and controlled access can support both revenue forecasting and cash analysis, where their requirements overlap. Collections activity would introduce additional authority, including rules for customer contact. Reusing information reduces duplication, but each new action still needs its own justification and controls.

07

Fund the Change That Improves the Complete Cycle

The business case should compare the total effort required to reach an accepted position before and after the intervention. Preparation time, reviewer effort, exception handling and operating costs all belong in that comparison. A workflow that produces analysis quickly but requires extensive checking may relocate effort rather than reduce it. A simpler integration may offer greater value if missing information is the dominant constraint.

Model choice should follow the task and its consequences. Routine retrieval should not automatically use the same model and review process as a complex assessment of commercial alternatives. Finance should define the required quality and escalation conditions, while technology teams evaluate reliability, consumption costs and support requirements. Broader rollout should follow evidence that the first workflow performs usefully under real operating conditions, including incomplete information and exceptions.

The offer: a two-week Planning Diagnostic

We offer a two-week Planning Diagnostic to establish where that first intervention belongs. With the finance sponsor, FP&A, business contributors and the planning-system owner, we map the planning estate, trace one recent reforecast and quantify the time, people and reconciliation effort involved. The outputs include a costed cycle and a prioritised sequence of changes. This gives finance and technology leadership a common basis for deciding whether the next investment belongs in the model, the data, the process or greater delegation.

The Close

For the executive team considering recruitment after a contract delay, the test is whether the revised process provides a dependable basis for action while commitments can still be changed. Investment should follow the part of the work that prevents that outcome. Book a session with us to examine how much of your planning cycle is spent analysing, waiting and reconciling, and establish which change is worth making first.

Decision Inc. FP&A & Planning

Before You Invest

Talk to us first. A short conversation can tell you whether your planning cycle has a measurement problem, a delegation problem, or both, and what the first intervention should be.