Why AI Efficiency Gains Aren't Enough: The Revenue Gap Most Organisations Miss

The Number That Should Bother Boards

Seventy-four percent of organisations globally want AI to drive revenue growth. Twenty percent can actually point to it happening. That is a 54-percentage-point gap between stated ambition and demonstrated evidence, from Deloitte's State of AI in the Enterprise report, which surveyed 3,000 senior leaders in 2025.

That gap matters because most organisations are not failing at AI. They are succeeding at the wrong version of it. As discussed on RNZ's Morning Report, AI is set to reshape thousands of roles within a year. The businesses positioned to benefit are not necessarily the ones with the most AI tools. They are the ones who have moved from efficiency to redesign.

If you are seeing time savings but not business outcomes, this article is for you.

Two Plays, Not One

Every AI investment sits somewhere on a spectrum between two fundamentally different plays.

The first play is "same output with less input." Do what you already do, but faster and with fewer people. Summarise reports. Generate first drafts. Automate data entry. These are real gains. Your team saves time. Operating costs fall. Productivity metrics improve. This is the efficiency play, and it is genuinely valuable.

The second play is "more output with the same input." Take that recovered capacity and redirect it toward new revenue, new clients, new products, or new markets. Serve more clients without hiring. Launch campaigns at scale. Move from reactive to proactive service models. This is the revenue play, and almost nobody is running it yet.

The Deloitte data makes this concrete: organisations are broadly achieving the efficiency play, while the revenue play remains elusive. The gap is not a technology problem. It is a design problem.

Why Efficiency Arrives First

AI has what researchers call a jagged capability profile. It is excellent at pattern-based, repeatable work: summarising, drafting, classifying, retrieving. It is uneven on tasks requiring novel combinations of judgment, relationship context, and real-time adaptation. Revenue plays tend to require the second type.

This means efficiency gains are structurally easier to achieve. You give an AI tool a repeatable task with clear inputs and outputs. It performs well. You measure time saved. The win is visible.

Revenue plays require something different: redesigning how work flows, not just how fast it happens. That requires understanding which processes, when changed, unlock capacity that can be deployed toward growth. Most organisations skip this step because it requires deeper analysis and leadership commitment, not just tool adoption.

Why Organisations Stall

In workshops across New Zealand, I hear a consistent pattern from leaders who feel their AI programme has plateaued. They report strong adoption metrics. Eighty percent of staff have accounts. Usage is up. But outcomes have not shifted.

The problem is what they are measuring. Login rates and activation figures are proxy metrics for tool adoption, not business change. I once spoke with a director who proudly cited "80% adoption across the organisation." When I asked which processes had changed as a result, there was a pause. They could not name one.

That is not an adoption success story. That is 80% of staff using AI to do the same things slightly faster, with no redesign of how value is created or delivered. The efficiency is real. The business impact is not.

The stall has a structural cause. Most AI programmes are designed around tool rollout, not workflow transformation. Training focuses on "how to use ChatGPT" rather than "which processes, when redesigned, unlock growth." Without that second question, organisations get efficiency by default and revenue by accident, if at all.

Where Most NZ Organisations Sit

Working with 400+ organisations since 2023, the pattern is consistent. Most sit between Rung 2 (Augment) and Rung 3 (Automate) on the AI Adoption Ladder. Fluency is building. Teams are using AI tools daily. Individual productivity is improving. But workflows have not been redesigned and processes have not been connected to AI in ways that change business outcomes.

That is the efficiency zone. It is a legitimate and important stage. It is not the destination.

The organisations capturing revenue upside are those that have moved past Rung 2 fluency into Rung 3 process connection and, critically, begun the Rung 4 redesign work where whole functions rethink how they deliver value with AI embedded.

What the Revenue Play Actually Requires

The path from efficiency to revenue runs through two specific transitions. Both require deliberate design, not just more tools.

Transition 1: Rung 2 to Rung 3 (Efficiency to Workflow Automation)

At Rung 3, you stop using AI for one-off tasks and start connecting it to actual business processes. Document processing pipelines. Automated reporting workflows. Customer correspondence at scale. The shift is from "individual uses AI" to "process runs with AI embedded."

This is where capacity recovery becomes substantial enough to matter. A team that spent 40% of its time on manual data extraction and reporting now has that time back. The question is: what do they do with it?

Sector examples where this transition unlocks capacity:

• Professional services: Automated document review frees senior advisers for higher-value client work
• Banking and finance: Automated compliance reporting frees analysts for relationship development
• Healthcare administration: Automated patient communication frees clinical coordinators for complex case management
• Government: Automated correspondence handling frees policy staff for stakeholder engagement

Transition 2: Rung 3 to Rung 4 (Automation to Redesign)

At Rung 4, whole functions rethink how they operate with AI capability available. This is where revenue plays become possible because you are not just doing existing work faster, you are doing different work.

A sales function at Rung 3 uses AI to generate proposals faster. A sales function at Rung 4 uses AI to serve three times as many prospects with the same team, qualifying and nurturing at scale before human relationship time is invested. The output is not faster proposals; it is a fundamentally different revenue model.

A marketing team at Rung 3 uses AI to produce content more efficiently. A marketing team at Rung 4 uses AI to run personalised campaigns at a scale previously requiring ten times the headcount. The output is not cheaper content; it is market reach that was not previously economically viable.

The distinction matters: Rung 3 recovers cost. Rung 4 creates revenue. Both require AI, but the design questions are completely different.

What to Do This Month

If you are in the 74% who want revenue outcomes but are not yet seeing them, the starting point is not more tools. It is diagnosis.

Three questions to answer before your next AI investment:

1. Which processes, when automated, recover the most capacity? Not which processes are easiest to automate; which automation, if successful, frees the most valuable time.
2. What would your team do with that capacity if they had it? If the honest answer is "the same things, but faster," you are at Rung 3 thinking. If the answer is "things we currently cannot afford to do," you are at Rung 4 thinking.
3. Where does AI capability match a growth bottleneck? Look for places where your business cannot scale not because of demand, but because of cost or capacity. Those are the revenue plays.

Take the AI Maturity Benchmark to identify where your organisation sits on the ladder and which transition is most relevant. The benchmark takes two minutes and gives you a starting point for the conversation your leadership team needs to have.

Then: pick one process. Map it. Identify the automation opportunity. Pilot it. Measure not just time saved, but what you did with the time you recovered. That second measurement is the one that tells you whether you are on track for the revenue play.