ERP modernization: SAP, Oracle & the CFO's technology roadmap

What makes 2026 different from previous ERP cycles is the regulatory stack now sitting on top of the finance function:

• OECD Pillar Two requires entity-level effective tax rate calculations across every jurisdiction with a 15% global minimum tax. Legacy ERPs weren't designed for the GloBE data model. PwC estimates 71% of multinationals had to deploy bolt-on tax engines in 2024-25 because their core ERP couldn't generate the required data.
• CSRD demands audit-grade ESG disclosures for ~50,000 EU-operating companies, with double materiality assessments tied to financial statements. This requires a single source of truth for both financial and non-financial data, the original promise of ERP, finally enforceable.
• IFRS 16 / ASC 842 lease accounting continues to expose ERPs that handle leases in spreadsheets, particularly painful as interest rate volatility makes discount rate selection material.

The cloud question is already decided

Five years ago, "on-prem versus cloud" was a real debate. It isn't anymore. SAP's RISE and GROW offerings, Oracle Fusion, Microsoft Dynamics 365, and Workday have collectively captured 82% of net-new ERP deployments since 2023. The remaining on-prem holdouts are typically defense contractors, certain financial institutions with data residency mandates, and process manufacturers with deeply customized MES integrations.

But "cloud" is not one thing. The CFO needs to understand four deployment patterns:

1. Public cloud SaaS (Oracle Fusion, Workday, SAP Public Cloud Edition): Lowest TCO, fastest deployment, but minimal customization. Best for businesses willing to adopt vendor-defined processes.

2. Private cloud / RISE with SAP: Hyperscaler-hosted (AWS, Azure, GCP), customer-specific instance. Allows moderate customization. SAP's flagship migration path.

3. Hybrid: Core financials in cloud, specialized modules (manufacturing, complex consolidation) on-prem or in private cloud. Common for industrial conglomerates.

4. Multi-tenant industry cloud: Vertical-specific (Oracle NetSuite for mid-market, SAP IS-U for utilities, Infor for healthcare). Strong fit if the vertical matches.

🎬 [VIDEO: "SAP S/4HANA vs Oracle Fusion: The CFO's Decision Framework" - youtube.com/results?search_query=SAP+S4HANA+vs+Oracle+Fusion+CFO - Walkthrough of the architectural differences, licensing models, and total cost of ownership comparison between the two dominant enterprise platforms]

Case study: unilever's RISE migration, the counter-narrative to hershey

Unilever's "Horizon" program, announced in 2022 under then-CFO Graeme Pitkethly and continued by current CFO Fernando Fernandez, is the most-watched ERP transformation of the decade. The company is migrating 190 country operations off a heavily customized SAP ECC environment to SAP S/4HANA on RISE, consolidating from over 40 instances to a single global instance running on Microsoft Azure.

What Unilever did differently, and what every CFO should study:

1. They treated it as a business transformation, not an IT migration. Pitkethly publicly committed that the program would *reduce* customization by 60%, forcing business units to adopt standard SAP processes ("clean core" in SAP parlance). This is psychologically harder than it sounds: every country GMGMGross margin is the share of revenue left after subtracting the direct cost of producing goods or services, expressed as a percentage of revenue.View full definition → has reasons their version of order-to-cash is special. Unilever's executive committee pre-committed to overruling them.

2. They sequenced by complexity, not by revenue. The first wave (2023) targeted mid-complexity markets like Indonesia and the Philippines, large enough to matter, simple enough to fail safely. The US, UK, and India go last (2026-27). Hershey's, by contrast, attempted a Big Bang go-live across order management, transportation, and warehousing simultaneously.

3. They budgeted for failure. The reported program cost is €800 million, €1 billion over five years, with explicit contingency for parallel running and rollback. Unilever's investor communications have been deliberately under-promising: "modest efficiency gains in Phase 1, compounding through 2028."

The results so far: 14 markets live by Q3 2025, days-sales-outstanding improved 11% in migrated markets, monthly close reduced from 8 days to 4 days. The stock has not moved on ERP news, which, perversely, is the success metric. Investors haven't had to think about it.

What hershey got wrong (and most cfos still get wrong)

Hershey's failure modes are textbook. The board can quote the case study but still authorize the same mistakes:

• Compressed timeline driven by an external deadline (Y2K for Hershey, the 2027 SAP cutoff today). A regulatory or vendor deadline is the *worst* reason to go-live. Push the deadline or push the scope, never the testing.
• Big Bang go-live instead of phased rollout. The math of integration testing is unforgiving: combinatorial complexity scales with the square of modules deployed simultaneously.
• Treating the SI (systems integrator) as the accountable party. Deloitte, Accenture, IBM, and Capgemini will all happily take the fee. They will not lose their jobs if the implementation fails. The CFO and CIO will.
• Underestimating data migration. Lift-and-shift of dirty master data produces dirty cloud data. Unilever spent 14 months on data cleansing *before* the first technical migration.

The CFO's ERP decision framework

The temptation is to delegate ERP selection to the CIO. Resist it. The CFO owns the business case, the change management mandate, and ultimately the financial reporting risk. Here is the framework I recommend running with your steering committee.

Step 1: define the "why" before the "what"

Before evaluating SAP versus Oracle versus Workday, document the *business outcomes* the ERP must enable in 2030. Examples:

• Close the books in 3 days, not 10
• Generate Pillar Two GloBE returns in-system, not via Excel reconciliation
• Embed CSRD ESG metrics in every transaction, not bolted on at year-end
• Support 5x M&A volume without proportional finance headcount growth

If you cannot articulate 5-7 specific outcomes, you are not ready to select a vendor. You are ready to do strategy work.

Step 2: run a true total cost of ownership model

Vendor list pricing is a starting point, not an answer. A realistic 7-year TCO model includes:

| Cost Category | Typical % of Total |

|---|---|

| Software subscriptions / licenses | 22-28% |

| Systems integrator fees | 35-45% |

| Internal labor (often hidden) | 12-18% |

| Data migration & cleansing | 6-10% |

| Change management & training | 5-8% |

| Contingency (real, not theoretical) | 10-15% |

Most CFO presentations under-count internal labor by 50-70%. Your best FP&A and controllership talent will be pulled onto the program for 18-36 months. That has a cost, both in cash and in the deferred work they aren't doing.

Step 3: pre-commit to the operating model

ERP forces process standardization. The single most predictive variable for success is whether the executive committee has pre-agreed to adopt the vendor's process versus customize the software. Procter & Gamble's Lawrence Kuhn famously called this "fitting yourself to the suit, not retailoring the suit." Customizations are the silent killer: they balloon implementation costs, complicate upgrades, and accumulate technical debt that returns in 7 years as the next migration crisis.

A useful test: for every customization request, ask the business owner to quantify the dollar value of the deviation from standard. If they can't, deny the customization.

Step 4: govern like a public company audit

The most successful programs run like an SEC reporting cycle:

• Monthly steering committee chaired by the CEO (not the CIO)
• Quarterly external assurance from a Big Four firm separate from the SI
• A named "Executive Sponsor" with veto authority, typically the CFO or COO
• Pre-defined go/no-go criteria, in writing, signed before each phase

When Lidl abandoned its €500 million SAP project in 2018 after seven years, the post-mortem found that bad-news escalation had been culturally suppressed for years. Build a structure where the project manager can tell the CEO "we are not ready" without it being career-ending.

The AI layer: what's actually new in 2026

Both SAP and Oracle have aggressively embedded generative AI and AI agentsAI agentsAgentic AI refers to AI systems that pursue goals autonomously by planning, taking actions through tools, and adapting based on results, with minimal step-by-step human direction.View full definition → into their platforms over the past 18 months. SAP's "Joule" copilot and Oracle's Fusion AI AgentsAI AgentsAgentic AI refers to AI systems that pursue goals autonomously by planning, taking actions through tools, and adapting based on results, with minimal step-by-step human direction.View full definition → are no longer demoware, they are GA features with real adoption.

What CFOs should care about, beyond the hype:

• Automated journal entry classification and anomaly detection, Salesforce reported a 30% reduction in close-cycle exceptions after deploying Oracle's AI close agents in 2025
• Natural-language financial query, analysts asking "show me variance to forecast for EMEA Q3 by product line" and getting an answer in seconds, not after a JIRA ticket to FP&A
• Predictive cash forecasting, moving from monthly direct-method forecasts to rolling 13-week forecasts