ROAI 4-Category Framework

ROAI 4-Category Framework (v2026.1)

The ROAI 4-Category Framework is Advanta’s canonical way to partition Return on AI into four distinct value categories so that AI investments are not collapsed into a single, fragile “productivity” line. It is stamped as v2026.1 and applied consistently across diagnostics, AI Council decisions, and ROAI reports.

The Four Categories

Q1. Productivity

Direct, measurable operational gains:

• Labour-time savings (hours removed or repurposed)
• Throughput increases (more work per unit time)
• Unit-cost reductions (lower cost per transaction, case, or unit)

This is the most visible category and often the focus of early business cases, but it is typically the smallest contributor to value over a 3–5 year horizon.

Q2. Defensibility

Value from being able to evidence that AI is governed under a named methodology with audit-recognisable artefacts. It is realised when a regulator, customer, auditor, or internal supervisor can clearly distinguish your function from peers because:

• Decisions and models are traceable and documented
• Governance artefacts (policies, logs, approvals, risk assessments) are consistently produced
• The function can show it followed a recognised, named method

Q2 value often appears as:

• Faster or smoother regulatory reviews
• Preferential treatment in audits or vendor assessments
• Reduced risk of forced shutdowns, fines, or remediation programmes

Q3. Institutional

Enduring capability built into the operating model that survives any single tool or vendor. Examples:

• Trained staff with defined AI responsibilities and skills
• Named product and model owners with clear accountabilities
• Recurring cadences (councils, reviews, post-incident learning)
• Vendor and partner relationships that can be reconfigured as tools change
• Telemetry, data, and monitoring infrastructure that can support multiple use cases

Q3 is the category most likely to persist when underlying tools change, and the one most often undercounted if everything is forced into a single productivity metric.

Q4. Category Positioning

Value from being recognised as a category-shaper rather than a category-participant. This is realised through:

• Stronger hiring pipelines and better candidate quality
• Higher referral and word-of-mouth momentum
• Premium pricing or improved win-rates in competitive deals
• Inbound partnership and co-creation demand

Q4 is about how your AI posture changes your position in the market, ecosystem, or internal landscape, not just how it changes your cost base.

Why Four Categories, Not One

A single-line “ROI from AI” number is rarely defensible because it forces every benefit into the Q1 (Productivity) frame, where value is least durable and easiest to challenge. The four-category split:

• Surfaces durable value
• Makes trade-offs visible
• Enables comparability

How the Framework Is Operated

The framework is embedded into the operating rhythm via named artefacts and cadences:

• Use Case Intake (USE-01)
• Quarterly Cadence Retrospective (GOV-15)
  • Over- or under-performance in each category
  • Systematic bias (e.g., always overestimating Q1, underestimating Q3)
  • Whether governance and operating model changes are landing as expected
• Annual Charter Refresh (STR-07)
  • New regulatory expectations
  • New data sources and telemetry
  • Evolving business strategy
• AI Council Decisions
  • Approve a use case with modest Q1 but strong Q2/Q3 to strengthen defensibility and institutional capability
  • Sunset a use case that delivers Q1 but damages Q4 (market perception or partner appetite)

Distinction from Adjacent Constructs

• Risk Taxonomy 2026 vs. ROAI 4-Category Framework
  • ROAI is the value lens (what value is realised, and in which category).
  • Risk Taxonomy 2026 is the risk lens (what can go wrong, and how it is controlled).
• Maturity Bands vs. ROAI Categories
  • Maturity Bands describe the operating posture (how advanced and robust your AI operating model is).
  • ROAI categories describe value realised (what you are actually getting from AI in practice).
• Forecasting vs. Realised Value

Governance and Evidence

A function that runs the ROAI 4-Category Framework as a recurring artefact (with documented inputs, decisions, and retrospectives) simultaneously evidences:

• DE-2 (Methodology transparency) – by using a named, repeatable value lens with clear categories and measurement methods.
• DE-5 (Continuous learning) – by revisiting realised vs. intended value, updating methods annually, and feeding lessons into future AI Council decisions.

Over a 3–5 year horizon, the combination of Q2, Q3, and Q4 value typically dominates Q1, which is why the framework is designed to keep all four categories visible and comparable rather than allowing productivity to crowd out the rest.