For decades, Excel has been the backbone of financial controlling and planning. It offers flexibility, familiarity, and direct control. But at scale and complexity, those strengths become constraints: monthly cycles that consume hundreds of hours in data collection and reconciliation, formula errors buried in multi-layered driver models, and a structural inability to incorporate real-time market signals.

The result? Finance teams that are technically competent but strategically reactive.

Why Agentic Planning Changes the FP&A Equation

Agentic planning represents a genuine paradigm shift.

This is not about replacing finance professionals. It is about augmenting them — giving FP&A teams "digital colleagues" that handle the repetitive, computationally intensive work, freeing CFOs and planning leads for higher-value strategic advisory.

The key benefits I have observed in practice include:

• Planning velocity: Cycle compression from weeks to days, or hours for ad-hoc scenarios, enabling faster response to supply chain shocks, currency shifts, or demand changes.

• Forecast granularity: AI models that combine internal drivers (production throughput, material costs, labour efficiency) with external signals (commodity indices, lead time trends, customer sentiment) consistently outperform static statistical methods.

• Scenario resilience: Autonomous generation and ranking of hundreds of stress scenarios, including black swan events, to support capital allocation and risk decisions.

Let me make this concrete with an example drawn from my experience managing finance across multi-entity, multi-system environments.

In a traditional setup, producing a reliable 13-week cash flow forecast meant assembling inputs from AR ageing reports, AP payment runs, inventory valuation files, and bank statements — each sitting in a different system, updated at different frequencies, owned by different teams.

The deeper problem is not the effort, but the lag. Napoleon once wrote that “space we can recover, lost time never.” In finance, the same principle applies: effort can be redeployed, but lost decision time cannot always be recovered.

With an AI-augmented approach across the full cash conversion cycle, the architecture changes fundamentally. Agents connected to ERP, banking APIs, and procurement systems monitor DSO trends, DPO commitments, and inventory turnover in near real time.

This kind of early signal, roughly three weeks ahead of what a manual process would have surfaced, created meaningful optionality.

The 18-month rolling forecast layer adds a second dimension by connecting working capital drivers (seasonal inventory build, capex timing, supplier payment terms) to longer-horizon cash planning, so that treasury and FP&A are working from the same forward view rather than separate models reconciled quarterly.

The value of AI here is not eliminating the need for judgment. It is compressing the time between signal and decision and expanding the range of options still available when finance teams act.

Accountability: There Is No “The System Did It”

Early in my career, I was sent to Panama as part of a SWOT audit team reviewing significant gaps between local and group ledger reconciliations. The local team was welcoming and genuinely willing to support, but as I worked through the reconciliations, I came across a large non-balancing item that stopped me cold.

The local finance manager had signed the reconciliation. I approached them — a thoughtful, agreeable individual — and asked about the entry. The answer, repeated in different forms across every question I raised, was the same: 

"This part I don't know. I just ran the program and reported the difference. What the system does, I don't know. This large imbalance — the system did it."

The system did it.

I couldn't interview the system. Someone had a logic behind that program. Someone had made a decision about what those entries should represent. But that knowledge had been completely severed from the person responsible for signing the reconciliation.

That moment has stayed with me. Because twenty years later, I see the same risk re-emerging — this time dressed in the language of AI.

A critical point that often gets lost in the enthusiasm around agentic planning tools: the model does not own the plan. FP&A does.

AI can surface relationships, flag anomalies, and generate scenarios at scale. But the planning logic, the validation of assumptions, the challenge of outputs that are arithmetically correct but commercially implausible — that remains firmly human.

The finance manager in Panama was not negligent. The system had simply created a gap between process and understanding that no governance structure had closed.

We cannot afford to replicate that gap with AI.

Model outputs are a starting position for professional judgement — not a conclusion, and never a signature.

A Practical Implementation Approach

A question colleagues and finance leaders often ask me is how to implement AI in their finance functions.

Based on my experience, a pragmatic rollout follows three phases:

Phase 1 — Foundation: Establish a unified, clean data layer with real-time accessibility. Robust governance, auditability, and data lineage are non-negotiable prerequisites. Skip this, and the AI models will amplify noise rather than signal.

Phase 2 — Intelligent Drivers: Introduce machine learning models that enhance existing driver frameworks — predictive variance analysis, anomaly detection, and natural language querying of planning models.

Phase 3 — Full Agentic Planning: Deploy multi-agent systems — specialised forecasting, optimisation, risk, and reporting agents — operating under human orchestration, integrated with ERP, BI platforms, and external data feeds.

Three Lessons for Finance Leaders

Three observations I would offer to finance leaders beginning this journey:

1. Data readiness is the real gating factor. The technology is mature. What delays implementations is fragmented, inconsistent source data.

2. Speed and control must be balanced explicitly. The appeal of agentic planning is velocity. The risk is that compressed cycles reduce the space for human review. Governance frameworks need to clearly define where automated outputs can flow directly into the plan and where a human checkpoint is required.

3. Upskilling matters more than tooling. The CFOs who will get the most from these systems are those who invest in building FP&A teams capable of prompt engineering, model validation, and structured AI oversight.

The Strategic Imperative for the CFO Agenda

The CFO role is evolving from steward of numbers to architect of strategic velocity.

Finance organisations that make this transition successfully will move from being scorekeepers to genuine value drivers — providing real-time navigation for a business operating in an increasingly complex world.

Resources:

1. Napoleon Bonaparte, letter to Baron von Stein, Dammartin-le-St.-Père, 7 January 1814.