From Cost Reduction to Growth Engine: How AI Creates New Business

McKinsey Global Institute estimates that AI represents approximately $4.4 trillion in annual value creation potential. Most enterprise AI programmes are designed to access a fraction of it.

The majority of AI initiatives are scoped around efficiency: reducing the cost of existing operations, eliminating manual steps in existing workflows, improving the speed of existing processes. These are real and measurable returns. They are also the smallest category of available AI value.

The organisations that access transformative AI value — the kind that shows up in revenue, in new customer segments, in market position — are the ones that move through efficiency and keep going. Not because efficiency is unimportant, but because it is a foundation, not a ceiling.

This framework maps four phases of AI value creation, what distinguishes organisations that advance through them from those that plateau, and what the growth opportunities look like at each stage.

The Efficiency Plateau Is Real — and It Is a Trap

There is a dynamic that appears consistently across enterprise AI programmes: efficiency gains plateau earlier than expected.

The first wave of cost reduction from AI tends to be rapid. Routine tasks are automated, processing times improve, headcount requirements for structured work decline. The business case is confirmed. The programme is declared a success.

Then growth slows. The remaining efficiency opportunities require deeper integration, more complex workflows, or more significant operational change. The AI programme, now measured against its initial cost-reduction mandate, does not expand its mandate. It optimises within the scope it has been given.

The result is an AI programme that is doing what it was asked to do — and leaving most of the value it could create on the table.

The four phases above are interactive — select any phase to explore the central question it answers, the outcomes it produces, and the metrics that indicate you are operating there.

Phase 1: Efficiency — Necessary, Not Sufficient

Efficiency phase AI is characterised by a specific question: how do we spend less to produce the same output?

The answer is structured automation: AI handles data entry, categorisation, extraction, routing, compliance checking, and the structured parts of workflows that previously required human attention. Cost per transaction falls. Processing speed improves.

This is worth doing. It is not sufficient.

The organisations that treat efficiency as their AI horizon are building a cost reduction capability when they could be building a growth capability. The difference is not the technology — it is the question being asked.

Phase 2: Capacity — The Overlooked Growth Multiplier

When efficiency improvements free operational capacity, organisations face a choice: reduce costs by reducing headcount proportionally, or redirect the freed capacity toward volume growth.

The organisations that choose volume growth — handling two or three times their previous client volume at the same operational cost — are accessing a category of value that never appeared in the initial efficiency business case.

This is capacity-phase AI. The cost structure does not change dramatically. The revenue opportunity does.

The capacity phase is systematically underestimated in AI business cases because it does not fit neatly into cost-reduction logic. It requires a different question: what new volume can we take on with the capacity we have freed? That question is a commercial strategy question as much as an operational one — and it requires commercial and operational leadership to be aligned on the answer.

Phase 3: Capability — Competing on What Others Cannot Offer

Capability-phase AI is qualitatively different from Phases 1 and 2. It is not optimising existing services. It is making new services possible.

Real-time analysis that previously required three days of analyst time. Monitoring that operates continuously without fatigue. Personalised delivery at a scale that manual operations cannot achieve economically. These are not faster versions of existing services — they are services that did not previously exist in the organisation's portfolio.

MIT Sloan research on AI leadership found that organisations with deliberate AI strategies reach capability-phase operations approximately 14 months faster than those with reactive AI adoption. The difference is not model quality — it is whether the organisation has a defined ambition for what AI will enable, not just what it will save.

Phase 4: Market — When the AI Becomes the Product

Market-phase AI represents the category of value that most enterprise AI programmes never reach in their planning horizon.

In this phase, the operational AI capability is no longer internal infrastructure. It becomes the basis for a new product, a new service, or access to a new customer segment that was previously uneconomical to serve. Accenture's 2024 research found that companies which move AI beyond efficiency into new service delivery grow revenue 2.5× faster than those that remain in efficiency mode.

The transition to market phase tends to be recognisable in retrospect, and opaque in advance. Organisations that reach it usually do so because they built operational AI infrastructure with deliberate depth — not because they made a single strategic decision to commercialise it. The infrastructure creates the option; the strategic decision exercises it.

The Six Opportunity Categories

Every phase of the growth trajectory contains specific opportunity types. The taxonomy below maps what those opportunities look like in practice.

The six opportunity categories above are not exhaustive, but they cover the pattern of AI-driven growth most consistently observed across industries. The phase column is significant: each opportunity type becomes accessible at a specific level of operational AI maturity. Attempting to reach Market-phase opportunities without Efficiency-phase infrastructure is a reliable path to underdelivery.

What Determines Which Phase Organisations Reach

The research on this question is consistent across sources: what determines how far organisations advance through the phases is not technology quality or AI investment level. It is the quality of the question they are asking at each stage.

Efficiency-phase organisations ask: what can AI reduce?

Capacity-phase organisations ask: what can AI enable us to do more of?

Capability-phase organisations ask: what can AI enable us to do that we cannot currently do at all?

Market-phase organisations ask: what can we create for customers from the operational capability we have built?

The organisations that advance quickly are the ones that do not wait to exhaust one phase's opportunities before asking the next question. They run the phases in parallel — using efficiency as funding for capability investment, using capability investment as infrastructure for market opportunity exploration.

Designing for Growth, Not Just Efficiency

The practical implication for AI programme design is this: the scope of the programme determines the scope of the value it can create.

An AI programme scoped to efficiency will find efficiency opportunities. An AI programme scoped to growth will find growth opportunities — and it will find them within the same technology base, the same team, and a similar operational investment.

The difference is in the mandate, the questions being asked, and the metrics being tracked. Operational AI leaders who expand those three things — mandate, questions, metrics — consistently find that the value available to them is substantially larger than the efficiency case they started with.

Sources

• McKinsey Global Institute (2023): The Economic Potential of Generative AI
• Accenture (2024): AI: Built to Scale
• MIT Sloan Management Review (2024): AI Strategy and Leadership
• BCG (2024): AI's Moment of Truth