Why do traditional ROI models fail for AI projects on legacy ERP systems?

Traditional ROI models assume predictable, linear returns, but AI on legacy systems involves unpredictable discovery costs like data quality remediation, API integration, and iterative model tuning. They also miss non-linear value realisation where the first use case has modest returns but subsequent ones leverage shared infrastructure for disproportionately higher returns.

How should CIOs structure the financial case for AI on legacy ERP?

Use a three-tier model: Tier 1 covers hard savings like reduced manual data entry (CFO-ready), Tier 2 addresses productivity gains such as faster reporting cycles (for COO and business unit leaders), and Tier 3 frames strategic optionality showing AI buys time and reduces future migration costs (for CEO and board).

How long does a phased AI rollout on legacy ERP typically take?

A typical phased rollout spans 12-14 months across three phases. Phase 1 (months 1-3) focuses on a foundation and quick win using 20-25% of budget. Phase 2 (months 4-8) expands to 2-3 additional use cases. Phase 3 (months 9-14) scales across the organization and establishes a centre of excellence.

What are the biggest pitfalls when building a business case for AI on legacy ERP?

The five most common pitfalls are overpromising on timeline, ignoring data readiness, underinvesting in change management (allocate at least 15% of budget), failing to establish baseline metrics before starting, and treating the business case as a one-time document instead of updating it quarterly with actual results.

The CIO's Playbook: Making the Business Case for AI in Legacy ERP Systems

You know AI can deliver value on top of your legacy ERP. Your team has identified use cases, tested a proof of concept, and the results are promising. But now comes the hardest part — convincing the CFO, the board, and every business unit leader that this investment deserves priority over a dozen competing demands on a finite budget.

Building the business case for AI in a legacy ERP environment is fundamentally different from pitching a greenfield technology project. You are layering intelligence onto a system that has accumulated years of technical debt, organisational workarounds, and deeply embedded processes. That context shapes everything — from how you model ROI to how you sequence the rollout to how you manage the politics.

According to Gartner's 2025 CIO and Technology Executive Survey, 64% of CIOs ranked "building a compelling business case" as the single greatest barrier to scaling AI beyond pilot stage. The technology is rarely the bottleneck. The bottleneck is organisational: securing sustained funding, aligning stakeholders with conflicting incentives, and demonstrating value fast enough to maintain executive confidence.

Why Do Traditional ROI Models Fail for AI on Legacy ERP?

Traditional ROI models fail because they assume predictable, linear returns — and AI deployments on aging infrastructure rarely behave that way. A standard capital expenditure model expects a defined investment, timeline, and payback. AI projects layered onto legacy systems involve discovery costs that are difficult to estimate upfront: data quality remediation, API integration, change management, and iterative model tuning.

Three specific problems undermine conventional ROI calculations:

Hidden infrastructure costs. Legacy ERP environments — especially on-premise SAP ECC or Oracle E-Business Suite instances — often require middleware, data extraction layers, or cloud staging environments before AI can access data reliably. These costs are easy to underestimate by 30-50%.
Non-linear value realisation. The first AI use case on a legacy system delivers modest returns because much of the investment goes toward foundational work (data pipelines, governance frameworks, security configurations). The second and third use cases leverage that foundation and deliver disproportionately higher returns.
Soft benefits that resist quantification. Faster decision-making, improved employee satisfaction, reduced compliance risk — these outcomes are real but difficult to assign a dollar value that a sceptical CFO will accept.

The fix is not to abandon ROI modelling. It is to use a framework designed for the reality of AI investments.

How Should CIOs Structure the Financial Case?

CIOs should structure the financial case as a phased value model with three tiers: hard savings, productivity gains, and strategic optionality. Each tier speaks to a different audience in the approval process.

Tier 1: Hard Savings (CFO-ready)

Quantifiable cost reductions that survive scrutiny in a finance review:

Reduction in manual data entry and reconciliation hours. AI-assisted automation can recover 40-60% of analyst time in targeted workflows. Our guide on AI data entry automation for finance teams details the calculation methodology.
Vendor consolidation savings. AI often reveals redundancies across the tooling ecosystem built around the legacy ERP — duplicate reporting tools, overlapping middleware, underutilised add-ons.
Reduced error and rework costs. McKinsey's 2025 analysis of AI in enterprise back-office operations found that organisations deploying AI for invoice processing on legacy ERP systems reduced error rates by 35-50%, translating to measurable savings in rework, late-payment penalties, and customer credits.

When presenting Tier 1, use conservative estimates and show your assumptions. A CFO who trusts your numbers at the lower bound is far more useful than one who questions your optimistic projections.

Tier 2: Productivity Gains (COO and Business Unit Leaders)

Frame these as capacity unlocked rather than headcount reduced:

Faster reporting cycles — AI-augmented reporting on legacy ERP data can compress month-end close activities by 3-5 days
Accelerated procurement analysis — AI tools that parse legacy procurement data can surface vendor contract insights in hours rather than weeks
Improved demand forecasting accuracy — particularly valuable for manufacturers running legacy planning modules

Tier 3: Strategic Optionality (CEO and Board)

This tier reframes the AI investment not as a cost but as an option on the future. Legacy ERP systems are depreciating assets. Every year without modernisation increases the risk of a forced, unplanned migration — which Deloitte Canada estimates costs 2-4 times more than a planned, phased approach for mid-market enterprises in the $100M-$500M CAD revenue range.

AI layered onto legacy ERP buys strategic time. It extracts additional value from existing systems while building the data infrastructure and organisational capabilities that make a future migration cheaper and less risky.

What Does Stakeholder Alignment Actually Require?

Stakeholder alignment requires understanding that every executive in the approval chain has a different definition of success, a different risk tolerance, and concerns they will never raise in the formal meeting.

Map each stakeholder to their core concern:

Stakeholder	Primary Concern	How AI Business Case Addresses It
CFO	Capital efficiency, payback period	Phased investment model, Tier 1 hard savings
COO	Operational disruption during rollout	Parallel-run approach, targeted use case sequencing
CISO/CTO	Data security, integration complexity	Architecture review, enterprise data security framework
CHRO	Workforce impact, skills gap	Training plan, role evolution narrative
Business Unit Heads	Team productivity impact	Use case relevance, quick wins in their domain
General Counsel	Regulatory compliance, liability	PIPEDA compliance plan, AI governance framework

The mistake many CIOs make is presenting the same business case to all stakeholders. Tailor the narrative to each audience, and conduct pre-meetings to surface objections before the formal approval session.

One effective approach: identify one senior leader outside IT who will co-sponsor the initiative. When the CIO and the VP of Supply Chain present jointly, it reads as a business initiative that happens to involve technology — not a technology project. That framing difference matters enormously in budget discussions.

How Should You Phase the Rollout to Protect the Business Case?

Phase the rollout to deliver visible value within 90 days while building toward the larger strategic vision. A phased approach limits financial risk and generates the early evidence needed to secure continued funding.

Phase 1 (Months 1-3): Foundation and Quick Win

Select one high-visibility, low-complexity use case — typically reporting automation or anomaly detection on financial data
Establish the data extraction and integration pipeline from the legacy system
Deliver a working prototype that demonstrates tangible value to at least one business unit
Budget: 20-25% of total programme investment

Phase 2 (Months 4-8): Expansion and Validation

Extend AI capabilities to 2-3 additional use cases leveraging the Phase 1 data infrastructure
Implement governance and monitoring frameworks
Begin change management programme for broader user populations
Budget: 40-45% of total programme investment

Phase 3 (Months 9-14): Scale and Optimise

Roll out across remaining prioritised use cases
Integrate AI outputs into executive decision workflows
Establish centre of excellence for ongoing AI development
Budget: 30-35% of total programme investment

The critical discipline is killing use cases that are not working. The business case is strengthened — not weakened — when you demonstrate the rigour to reallocate resources from underperforming initiatives.

Mini-Case Study: A Calgary Energy Company's Phased Approach

A Calgary-based energy services company with $240M CAD in annual revenue was running Oracle E-Business Suite 12.2 — deployed in 2011 with extensive customisations. The board had rejected a full Oracle Cloud migration proposal ($4.2M CAD over three years) but acknowledged the legacy system was constraining operations.

The CIO reframed the proposal: deploy AI on top of the existing Oracle instance to extract immediate value while building migration readiness.

Phase 1: AI-powered accounts payable automation. Delivered 42% reduction in invoice processing time and $185K CAD in annualised savings within 90 days.
Phase 2: Predictive maintenance integration with the Oracle asset management module. Reduced unplanned equipment downtime by 28%.
Phase 3: Natural language reporting layer for executive dashboards, eliminating 25 hours of manual report preparation per month.

Total programme investment was $1.1M CAD over 14 months, with a measured payback period of 11 months. The data infrastructure built during these phases reduced the estimated cost of the eventual Oracle Cloud migration by approximately 30%.

The CIO's retrospective observation: "The business case that got rejected was a technology project. The one that got approved was a business improvement programme that happened to use AI."

What Are the Common Pitfalls That Undermine the Business Case?

Even well-constructed business cases can fail. The most common pitfalls:

Overpromising on timeline. AI on legacy systems takes longer than on modern cloud platforms. Factor in data quality remediation, legacy API limitations, and integration testing. Add 30% contingency.
Ignoring data readiness. If your legacy ERP data has not been cleansed, your first three months will be spent on data preparation, not value delivery.
Underinvesting in change management. Gartner estimates that organisations allocating less than 15% of their AI budget to change management are 2.4 times more likely to see adoption stall.
Failing to establish a baseline. You cannot demonstrate ROI if you never measured the current state. Document process times, error rates, and costs before Phase 1 begins.
Treating the business case as a one-time document. Update it quarterly with actual results versus projections to build trust with finance and create the evidence base for future AI investments.

Key Takeaways

Structure the financial case in three tiers — hard savings, productivity gains, and strategic optionality. A single ROI number rarely survives scrutiny; a layered model that addresses each decision-maker's concerns is far more durable.
Phase the rollout to deliver visible value within 90 days. Early wins are not optional luxuries — they sustain executive confidence and continued funding through the harder phases.
Treat the business case as a living document. Update it quarterly with actual results, kill use cases that underperform, and use the evidence base to secure funding for the next wave of AI initiatives.