From AI-Licensed to AI-Driven Success: How to Generate Measurable Value in the Finance Sector

High Adoption, Low Return

The market figures paint a clear picture. Gartner expects that by 2026, around 90% of finance departments will have at least one AI solution in use – a figure that has nearly doubled within two years. According to Deloitte’s CFO Signals Survey, 54% of CFOs have made the integration of AI agents a top priority for 2026. This is offset by a sobering statistic: In a survey of 200 U.S. CFOs conducted by RGP, only 14% reported a clearly measurable impact from their AI investments.

The report State of AI in Finance 2026 by CFO Connect highlights where the problem lies. Although 56% of finance executives now use AI, the finance department ranks last among all corporate functions in terms of adoption. Forty-five percent of teams are stuck in a limited pilot phase, while only 17% use AI in their core processes. The report cites several hurdles: busy closing cycles that leave no time for experimentation, security and confidentiality concerns, a lack of training – and a striking sense of uncertainty: 68% of CFOs say they don’t know where to start.

The technology has long been available.

It’s not a matter of the tools. These capabilities are currently making the transition from prototype to production: According to BCG, agent-based AI functions are already in productive use in high-volume processes such as invoice processing, payment allocation, and account reconciliation – specialized providers report contactless automation rates of over 90% for payment allocation. In planning and performance management, AI-supported model generation, dialog-based planning, and continuous reforecasting are nearing maturity.

Why is so little of this translating into results? The Boston Consulting Group provides the clearest answer in its analysis “The CFO’s AI Agenda,” produced in collaboration with SAP: According to BCG research, only about 10% of AI success is attributable to the models themselves, and another 20% to the technology platform. The remaining 70% is determined by organization, processes, and competencies – precisely the foundations that are still being established in most finance organizations. AI creates value only where it operates on a trustworthy data foundation, closely aligned with the processes it is intended to support.

THE 70/20/10 RULE

Only 10% of AI success is attributable to the models, and 20% to the platform. 70% is determined by the organization, processes, and people – and that’s where consulting comes in.

Five Consulting Approaches for Achieving Measurable Benefits

Based on these findings, we at BDF EXPERTS have developed five concrete consulting approaches. They build on one another but can be implemented individually – starting from where your organization stands today.

01 · DATA OPTIMIZATION FOR AI IMPLEMENTATION

When AI falls short in the finance sector, the model is usually called into question. Almost always, the problem lies with the data. A general ledger consisting solely of raw account numbers and posting codes provides an AI agent with nothing from which to draw conclusions. Furthermore, most finance organizations work with a patchwork of ERP systems, planning tools, and Excel – each system offers only a partial view of the business. People bridge this gap with their experiential knowledge. AI cannot do that.

Our approach therefore starts with the data foundation in the SAP core: harmonized master data, a uniform chart of accounts, consistent cost center, profit center, and product hierarchies—and a live connection to transaction data rather than static extracts. The semantic layer builds on this: driver relationships are explicitly modeled (revenue in relation to price, quantity, and channel mix; costs to production and personnel), and transactions are enriched with context such as region, product, and customer segment. Only when “margin” means the same thing in Europe as it does in Asia does an agent have a solid foundation for analysis – a human analyst tacitly reconciles differing definitions, while AI adopts them uncritically.

It’s important to note: This is not a multi-year data migration. We work incrementally and in a use-case-driven manner – one planning domain at a time – with each stage measurably improving both AI maturity and financial performance. Momentum beats perfection.

02 · PROCESS STANDARDIZATION AND USE CASE PRIORITIZATION

Most AI pilot projects fail not because of the technology, but because intelligence is applied to processes that were never designed for consistency. Automating a fragmented process simply scales that fragmentation. We reverse the usual order: The starting point is not the technology, but the value – faster completion, more accurate forecasting, less coordination effort. Data, process, and technology requirements are derived from this. Standardization occurs in parallel with AI implementation, not before it; early efficiency gains finance the next wave.

03 · AI GOVERNANCE AS ARCHITECTURE

The finance department operates under conditions that other functions do not face: regulatory precision, documentation requirements, and separation of duties. “The AI said so” is not a valid argument before any supervisory board or auditor. Governance must therefore be built into the architecture from the start, not added later as a compliance exercise: data provenance that makes every result traceable back to the source entry; escalation rules specifying when an agent acts and when it hands off; approval processes that maintain the separation of duties. The guiding principle is: AI suggests, humans decide – with increasing autonomy only once reliability has been demonstrated. We are factoring in the requirements of the EU AI Act from the very beginning.

04 · AI AGENTS IN THE SAP LANDSCAPE

Agent-based AI doesn’t wait for input: It continuously monitors data, detects anomalies, analyzes causes, and suggests responses – this is already in production for order reconciliation and receivables management, and is about to be implemented in planning and performance management. We bring these capabilities to your SAP landscape: from activating AI functions in standard SAP to custom-developed agents that access your planned and actual data in a controlled manner via open interfaces such as the Model Context Protocol (MCP) – with clearly defined write permissions and regulated verification. This creates what BCG describes as a “digital team”: agents whose work remains under the ownership and responsibility of business owners.

05 · EMPOWERING THE FINANCE TEAM

A Gartner survey from early 2026 identifies building AI expertise in finance as the most pressing task for CFOs – ahead of technology and budgeting. This is because roles are shifting: Those who prepared variance reports yesterday will be reviewing AI-generated analyses tomorrow and deciding which findings require action. We support this transition with role-specific skill development rather than one-time training: AI fundamentals and limitations, effective prompting, and recognizing plausible-sounding errors – practiced using real-world workflows, starting with low-risk applications. A 30/90/365-day roadmap serves as a guide: first, a friction-intensive process and an assessment of existing tools; then, establishing structure with AI champions and governance rules; and finally, scaling up based on a robust data foundation.

From AI-Licensed to AI-Driven Success

For BDF EXPERTS, this is the decisive step: turning licensed tools into robust finance processes. As a leading SAP partner, we combine deep process and data expertise in the SAP core with SAP’s Business AI capabilities – and with a governance framework that also supports financial reporting and regulatory compliance.

In SAP-integrated financial and treasury planning – for example, with the Liquidity Planning Cockpit (LPC), the Cash Position Cockpit (CPC), or PPG MIKA (real-time calculation) – it becomes clear what makes the difference: AI is effective when it is based on reliable data and embedded in transparent processes. This transforms pilot projects into results that stand the test of time in reporting.

Sources: Gartner (2026 Adoption Forecast, Q1/2026 CFO Survey); Deloitte, 4Q25 CFO Signals Survey (54%); RGP survey of 200 U.S. CFOs, 2025 (14%); CFO Connect, “State of AI in Finance 2026” (56% / 45% / 17% / 68%, barriers, 30/90/365 roadmap); Boston Consulting Group & SAP, “The CFO’s AI Agenda: From Automation to Advantage,” May 2026 (70/20/10, automation rates, semantic layer, governance principles).