BARC, Horváth and PwC paint a consistent picture: the question is not if, but when.
99%
of CFOs report material challenges in their finance function.
BARC CFO Agenda 2025 · n=194
71%
feel growing pressure for daily steering information.
Horváth CFO Study 2025
35%
still rely on manual data transfers between planning and consolidation.
BARC FPM Score 2025
34%
of companies use AI agents in accounting and finance today.
PwC AI Agent Survey
Agentic AI is fundamentally transforming forecasting, planning and closing processes, not in the future but right now. 34% of companies are already using AI agents in finance. Those who start today will secure faster decision-making and higher-quality forecasts. This white paper covers: what Agentic AI is (and isn’t), six specific use cases, a five-stage maturity model, a practical checklist and honest answers to the questions that board members really ask.
This whitepaper answers three questions that are currently on the minds of finance professionals: What exactly is agentic AI, and what isn’t it? Where is tangible value being generated in BI and planning today? And how can we achieve this without jeopardising the data foundation?
Chapter 02
DEFINITION & SCOPE
What sets Agentic AI apart from copilots, RPA, predictive analytics and generative AI in a clear 5×4 comparison matrix.
Chapter 04
VALUE POTENTIAL
From data provider to strategic business partner: what exactly is changing and why the data foundation is crucial.
Chapter 05
RISKS & GOVERNANCE
The real-world challenges of AI laid bare: data quality, explainability, control limits, IT security, data protection, and change management.
Chapter 07
MATURITY SCORECARD
Self-assessment across 5 dimensions: Where does your organisation really stand? Interactive scoring tool directly within the document.
Chapter 08
PRACTICAL CHECKLIST
15 yes/no questions across four areas: data infrastructure, governance, processes, and technology. Print it out, discuss it, and share it with your teams.
Chapter 09
Conclusion
Agentic AI is not hype. It is the next stage of development.
34% of companies are already using AI agents in finance and accounting. Those who start now will build a real competitive edge, and those who wait will fall behind with every passing quarter.
The whitepaper shows how the remaining 66% can keep pace, step by step, with governance built in and no big-bang migration.
Not in the future. Not just in theory. These are use cases that are already available in production EPM systems.
Use Case · 01
Automated variance analysis & root-cause investigation
An AI agent continuously monitors key performance indicators and automatically triggers a multi-stage analysis in the event of significant deviations, meaning results are delivered in minutes rather than days.
Use Case · 02
Dynamic forecasts & rolling planning
Instead of monthly forecasting cycles, an agent updates the rolling forecast based on current transaction data and market information, continuously and consistently.
Use Case · 03
Scenario planning & sensitivity analyses
Best-case, base-case and worst-case scenarios based on varying assumptions regarding exchange rates, commodity prices and demand trends. Days are compressed into hours.
Use Case · 04
Financial close & consolidation
Agents assist with intercompany reconciliations, general ledger mapping and the identification of genuine exceptions, providing explanations of the causes and recommendations for correction.
Use Case · 05
Regulatory reporting & compliance
Agents continuously monitor CSRD, transfer pricing and IFRS metrics and proactively report risks, resulting in a measurable reduction in manual effort.
Use Case · 06
Data quality assurance
Before feeding data into decision-making models, the agent checks for consistency, completeness and plausibility and escalates any anomalies along with the relevant context.
From the data foundation to autonomous control. Not as a sudden, radical change, but as a structured process with clear milestones.
01
Data Foundation
“Is our data reliable, consistent and accessible?”
Consolidation of the data foundation: unified data definitions, clean MDM structures, a clear semantic layer, documented data governance. Without this foundation, all downstream AI initiatives fail.
Integration
“Do financial, operational and market data flow seamlessly into a single steering architecture?”
EPM, BI, ERP and operational source systems interact in an integrated architecture. APIs, modern data pipelines and semantic layers form the technical bridge.
03
Augmented Intelligence
“Are our teams actively using AI support in planning and analysis processes?”
Copilots, predictive models and NLP-based analysis tools are introduced. Humans retain decision authority, and AI delivers recommendations, explanations and patterns.
04
Agentic Workflows
“Can defined planning and analysis steps be executed autonomously by agents?”
First real AI agents are integrated into processes: KPI variances, rolling forecast, scenario generation. Human-in-the-loop control points are defined and agents act only within set parameters.
05
Autonomous Steering
“Do humans and AI work together in a continuous, adaptive steering model?”
AI agents continuously plan, monitor, adjust and escalate, while human decision-makers set strategic guardrails and handle exception decisions. Autonomous steering does not mean human-free steering.
In 90 minutes we will jointly identify your real starting point, your biggest levers and the most sensible next steps on the path to Agentic AI in your corporate steering.
90 mins · no obligation · online or in person
Yes — in certain areas. Systems such as Board offer practical AI functions via FP&A Agent and Controller Agent: automated variance analysis with explanations of causes, support for intercompany reconciliations, and exception detection in financial reporting. Jedox enables conversational data queries and AI-powered forecasts using external market data. Fully autonomous, enterprise-wide control is the ultimate goal and the path to achieving this begins with what is available today.
Copilots and generative AI tools respond to human queries: they formulate, explain, and support. AI agents are given a goal, devise a solution, use tools, and carry out steps autonomously. The difference is the shift from reactive to proactive, from assisting to acting. GenAI comments on a deviation from plan; agentic AI detects the deviation, analyses the cause, simulates scenarios, and suggests countermeasures.
You don’t need perfect data, just sufficiently consistent data: clearly defined KPIs, clear ownership, integrated actuals and targets, and a documented data model. Inconsistent data leads to poor AI outputs, faster and at greater scale. Start pragmatically with the use case where the data is already cleanest.
Control comes through clearly defined action limits, escalation procedures and human-in-the-loop checkpoints. Critical processes, such as budget approvals, regulatory submissions and major planning adjustments, remain under human control. Human-in-the-loop is not a temporary safeguard, but a permanent design principle: agents act autonomously only within clearly defined parameters.
No — but roles are changing. Anyone who currently spends a large part of their time on data preparation will find that AI takes some of that burden off their shoulders. The time saved can then be spent on what machines cannot do: understanding context, exercising judgement, building trust, and providing advice. The real question is: “How do I use the time that AI frees up for what really matters?”
The most common pitfall is waiting for the right moment. A tried-and-tested approach is to start with a clearly defined use case where data and processes are already stable enough, then build a prototype within a few weeks. Not as a finished system, but as real-world proof of what is possible. According to the PwC AI Agent Survey, only 34% of companies currently use AI agents in finance, meaning those who start today can build a real competitive edge.
A pilot in a clearly defined area, such as automated forecasting for a business unit or AI-supported variance analysis for a sub-plan, can be delivered with manageable investment. ROI can be measured through reduced manual effort, improved forecast accuracy, shorter planning cycles and better-quality decision-making. The key point: ROI does not come from the technology alone, but from redirecting the capacity it frees up into more strategic work.
