The 80/20 of AI: who's winning, who's just spending

16%.

That's the percentage of companies getting measurable business value from artificial intelligence — according to a new Harvard Business Review study published in April 2026.

Despite billions of dollars in enterprise AI spending, despite 59% of organisations now running AI in production, and despite virtually every Fortune 500 announcing an "AI strategy" over the past 24 months — only one in six companies can point to genuine business impact.

The remaining 84%? They are spending. Deploying. Adopting. And wondering why the returns haven't arrived.

Meanwhile, according to a 2026 PwC study: 20% of companies are now capturing 74% of AI's total economic value.

This article is about why that gap exists, what the winners are doing differently, and how any organisation can begin to close it.

The two camps: Automators and Transformers

Deloitte's research on AI maturity divides organisations into two camps: Automators and Transformers.

• Automators add AI onto the business they already have. They deploy chatbots, copilots, and predictive tools alongside their existing workflows. The team structure, the data architecture, the decision rights — none of it changes. AI is a productivity layer.
• Transformers rebuild the operating model first — workflows, decision rights, accountability — and let AI integrate inside it. AI becomes a structural component, not a tool.

Same technology. Different architecture. Different outcomes.

The HBR data confirms it: companies that embed AI directly into their core processes report substantial or moderate value 71% of the time. Companies that run AI alongside their processes report a much weaker return profile.

The category your company belongs to is determined by three operational decisions.

Decision 1 — Where AI lives

The single most predictive question for AI ROI is also the simplest: does AI sit close to the work, or inside it?

Only 18% of organisations have rebuilt their workflows so that AI runs inside the process — automatically, every time. 34% run AI as a standalone tool people use when they need it. Another 34% are in transition.

The data on what happens at scale is clear: 71% of companies that embed AI into their processes report real business value. The rest are largely running pilots that never reach production.

When AI lives outside the flow of work, it lacks the context and connectivity needed to drive meaningful business outcomes. It becomes optional, episodic, and easy to ignore. When it lives inside workflows, it becomes structural.

Decision 2 — What gets fixed first

The instinct of most companies is to deploy more AI tools. Add another copilot. Pilot another use case. Sign another vendor contract. The data says the opposite is what works.

• Modernising legacy infrastructure delivers the highest ROI of any AI investment in the HBR study: 76% of organisations doing it report substantial business value.
• Yet only 45% of companies are actually doing it. Integrating data sources comes in second at 75%.
• Orchestrating processes across systems comes in third at 73%.

In other words: the foundational work is the differentiator, not the model.

This is also where 69% of organisations report being held back. Legacy systems are limiting their ability to scale AI across the enterprise.

Decision 3 — How AI is governed

Governance is the area where executive sentiment and executive action are furthest apart.

• 92% of executives in the HBR study agree that AI agents need rules-based guardrails to operate safely and effectively.
• Only 48% say their company has defined those rules.

That 44-point gap is the distance between "we take AI safety seriously" in a public statement and the day a regulator, a customer, or a journalist proves the company doesn't.

The risk profile escalates sharply with agentic AI. The HBR data already shows organisations encountering meaningful problems:

• 36% report difficulty integrating agents into systems,
• 34% report factual errors and hallucinations,
• 34% say agents lack access to the right data,
• 31% cite a lack of governance over agents specifically.

Transformers write the guardrails, audit trails, and escalation paths before AI scales. Automators figure it out after something breaks.

Why productivity gains are not a competitive advantage

Most executives, when asked what AI is delivering, point to productivity. The HBR data confirms it:

• 64% of organisations report productivity improvements as their top AI outcome,
• followed by operational efficiency at 58%.

This is the trap.

Productivity gains are valuable, but they are not durable. Every competitor will achieve similar gains within 12 months as the same tools become commoditised and the same playbooks become public. Spending tens of millions to arrive at parity is not a strategy.

Transformers are pursuing something fundamentally different: growth.

• New revenue streams.
• New business models.
• New ways to compete that didn't exist 24 months ago.

The HBR data shows just 30% of companies are seeing new revenue from AI, and only 35% are seeing improved ROI — the highest-leverage outcomes are concentrated in the smallest group of companies.

The leadership signal hidden in the data

What separates Transformers from Automators is a clear-eyed view of what AI is, and what it requires.

Transformer leaders have stopped chasing pilots and tools. They have stopped reorganising around every new model launch, every benchmark, every product release from OpenAI or Anthropic or Google. They have focused, instead, on the only question that matters at the C-suite level: where does AI change outcomes — revenue, cost, customer experience — in our business?

That clarity is the real edge. Not budget. Not talent. Not roadmaps. They stopped chasing the technology. They started using it.

Every real transformation eventually comes back to the same truth: if it doesn't change the operating model, it's just a budget line. AI is no different. If your processes, your data, and your teams look the same 18 months in — you didn't transform. You repainted the surface.

The path forward for organisations still in the 80%

If you're reading this and thinking "we'll wait and see" — that is the Automator choice. And every quarter you wait, the gap to the Transformers increases. You're not standing still. You're falling behind.

The encouraging news is that becoming a Transformer doesn't start with a $50M programme, a new Chief AI Officer, or a board-approved roadmap. It starts smaller — with three actionable moves we recommend leadership take in the next 90 days:

1. Pick one core process and embed AI inside it. Train the people running that process to work alongside AI as part of their job. Measure the before-and-after delta on a single metric.
2. Identify one piece of legacy infrastructure that is blocking AI value and fix it. The one that, if integrated and modernised, would unlock the most downstream AI capability. Start there.
3. Write one page that defines what AI is allowed to do in your company and what stays human. One page is enough to start. It will evolve, but the act of writing it forces conversations among leaders.

The bottom line

The Transformers aren't smarter. They aren't better funded. They didn't have a clearer roadmap. Their leaders understood that AI is not an upgraded tool for the business — it's an invitation to redesign it.

The 20% of companies capturing 74% of AI's value have already made that decision. The 80% still have time.

Sources: HBR Analytic Services, "What Drives AI Value: Why Modernisation and Workflow Integration Matter," April 2026 (sponsored by Appian); PwC, 2026 AI Performance Study; Deloitte, AI Maturity and Digital Value research.