Is Your Finance Ready for SAP’s AI Future?

Finance has always been the last place organisations want to get wrong.

That makes what is happening inside SAP’s current AI roadmap worth paying attention to — carefully.

SAP’s autonomous enterprise direction points to a finance function where systems do far more than record and report. They will monitor financial activity, identify exceptions, perform close tasks, support reconciliations, assist with cash forecasting, and help finance teams respond more quickly to changing conditions. AI assistants are being positioned across financial close, billing, accounts receivable, tax and compliance, treasury, and planning.

For CFOs under pressure to close faster, forecast more accurately and make sense of complex operating environments, that is a significant opportunity.

But there is a practical catch — and it matters more in finance than almost anywhere else.

Autonomous finance will not work simply because AI is available. It will only work if the data, processes, and governance underneath it are ready. If company data is inconsistent, approval rules are undocumented, or finance processes rely on informal workarounds,the level of autonomous capability will be impaired.

The first step towards autonomous finance is not choosing a tool. It is preparing the environment so AI can operate safely, accurately and transparently.

Data is the foundation on which everything else rests on

Finance AI is only as good as the data underneath it.

Autonomous agents need reliable master data, clean transactional records, consistent business rules, and clear ownership. In a manual environment, experienced finance professionals fill gaps all the time. They know which figures need a second look, which records to check, and when an exception is normal. Without that business context, AI cannot move beyond generic outputs to the kind of context-aware intelligence finance teams need.

For autonomous finance to work, data needs to carry meaning — not just values. A transaction connects to a customer, supplier, contract, purchase order, approval path, business unit, tax treatment and reporting obligation. When that context is weak, fragmented or inconsistently maintained, AI inherits the problem.

In finance, a data issue can quickly become a control issue:

If the data issue is…	The finance risk could be…
Duplicate customer records	Incorrect AR ageing or poor collections decisions
Inconsistent vendor data	Payment errors or reconciliation failures
Poorly governed cost centres	Unreliable management reporting
Unclear approval rules	Weak control over automated decisions
Undocumented manual adjustments	AI cannot explain or repeat the correct treatment

The objective is not perfect data — very few organisations have that. It is fit-for-purpose data: accurate enough, governed enough, and connected enough for the level of automation being considered.

Before expanding AI’s role in finance, CFOs should be asking: which data do we trust today? Which reports still require manual checking? Where do people rely on workarounds outside the system? The answers will show where the real readiness gaps sit.

Design the process before you automate it

Data readiness is necessary. But it is not sufficient on its own.

Finance processes become complicated not because people lack skill, but because the business around them changes. New entities are added. Acquisitions introduce different systems. Local workarounds become permanent. Manual checks accumulate. Over time, the process still works — but only because experienced people know how to navigate it.

What is actively manageable in a human-led environment, inhibits the level of autonomy possible.

If an AI agent is asked to support financial close, it needs to understand dependencies, tolerances, approval paths, timing, materiality and control points — not just the task list. If an AR assistant is monitoring ledgers and recommending dunning actions, it needs to understand customer segments, dispute rules, trading history, and commercial exceptions.

A poorly designed process does not improve because AI is added to it. It simply moves faster.

Before automation is expanded, finance leaders should be able to clearly answer: where does the process start and finish? Which decisions are rule-based, and which require judgment? Who owns each approval and escalation point? Which tasks are repeated manually because the system does not reflect how the business actually works?

This is the practical starting point. AI should be layered on top of finance processes that are already understood, mapped, and governed—not on complexity that has never been examined.

Clean core is a finance priority, not just a technology conversation

Clean core is often described as a technical system principle. For CFOs, it needs to be understood as a business risk issue.

In simple terms, clean core means keeping the core ERP environment standard and upgrade-safe, while placing extensions and custom logic in the right architectural layer. That has always mattered for system stability. It matters even more when AI is introduced.

Autonomous finance agents need to operate against systems that are predictable and explainable. If the finance environment is heavily customised, full of undocumented logic, or dependent on legacy workarounds, autonomous workflows become harder to trust and harder to audit.

A finance process does not only need to run. It needs to be controlled, traceable, and aligned with current policy. If the underlying system is too complex or inconsistent, AI does not simplify that complexity — it inherits it.

For finance leaders, the clean core conversation should move beyond IT architecture. It should become part of finance transformation planning.

The useful question for CFOs is not: “Can this still run?” It is: “Can this be trusted, governed and explained when AI starts using it?”

Governance must be built in from the start

Autonomous finance introduces a different governance challenge.

Standard automation follows a fixed rule — if the condition is met, the action is triggered. AI-supported finance is different. An autonomous agent may interpret context, assess exceptions, recommend a course of action, prepare a journal entry, or support a compliance decision. The more responsibility the agent takes on, the more important governance becomes.

Finance cannot operate on black-box decision-making. If AI-supported processes affect reporting, tax, billing, collections or treasury, the organisation must be able to explain what happened, why it happened, which data was used, who approved the action, and whether the result complied with policy.

A practical governance model for autonomous finance should define:

What the agent is allowed to do — recommendation, preparation, or execution within limits
Which data sources it can use — reduces the risk of incomplete or unreliable outputs
What requires human approval — especially for material, compliance-sensitive or customer-facing decisions
How decisions are logged — supports audit, risk review and regulatory obligations
How exceptions are escalated — AI should not silently process unusual cases

Autonomy should increase in stages. AI supports analysis first. Then it recommends action. Then it prepares action for human review. Then it executes within clearly defined thresholds. At each stage, the controls should be in place before the autonomy expands — not added afterwards.

A readiness checklist for CFOs

Autonomous finance readiness does not require a large-scale programme to begin. It starts with an honest assessment of where the finance function stands.

For each priority finance process, work through these questions:

Data Provenance and Contextual clarity: Do we trust our customer, vendor, chart of accounts and cost centre data? Would we be confident letting AI make recommendations based on it today?
Process design: Is this process clearly mapped — including the workarounds, informal approvals and recurring exceptions? Does the actual process match the documented one?
Rules and ownership: Are approval thresholds, delegation of authority and control points documented and current? Do we know who owns the rules AI would follow?
Auditability: Could we explain an AI-supported decision to internal audit, the board or a regulator? Can we show which data were used, which recommendation was made, and who approved the outcome?
Operational connections: Does finance have timely, trusted access to data from HR, procurement, supply chain and operations? AI-supported forecasting and compliance depend on more than financial data alone.
People readiness: Are finance teams ready to supervise AI-supported workflows — reviewing exceptions, challenging outputs, and deciding where human judgement still belongs?

The goal of this exercise is to classify processes into three honest categories: ready to explore, needs preparation, or not yet ready. That gives finance leaders a practical way to move forward without overcommitting.

The foundations come first

Autonomous finance is a genuine shift in how finance operates — and a significant opportunity for organisations prepared to approach it seriously.

Discovery Consulting helps Australian finance and technology leaders assess data readiness, map processes, align SAP architecture, define governance, and build staged roadmaps for AI-enabled finance transformation. The work starts with the business problem, not the technology, and focuses on building the foundations that give AI the right conditions to operate safely.

The organisations that gain the most from autonomous finance will not necessarily be the ones that adopt AI earliest. They will be the ones who prepare their data, processes and controls well enough for AI to be trusted with material finance work.

That preparation is where the value is built.

Preparing your finance function for what comes next?

Talk to Discovery Consulting to assess your SAP finance readiness and build a practical roadmap for autonomous finance.