AI promises outcomes. Scaling them is the hard part

Published in Strategy, AI

Eveliina Lakka

Business Strategist

Eveliina Lakka is a Business Strategist at Nitor and recently completed her doctorate on data-driven business models. In this article, she looks at what it takes to scale outcome-based business models in the AI era. The article builds on her research published in Industrial Marketing Management in 2026.

November 11, 2026 · 8 min read time

Everyone wants to sell outcomes, but only a few succeed. For business leaders and P&L owners, this is the ultimate dilemma: As AI increasingly carries out tasks independently on behalf of users, more companies can charge for outcomes instead of products, used hours, or work done. Yet, most struggle to turn this shift into a scalable, profitable business model.

Some of the best-known outcome-based business models come from industry. Rolls-Royce sells engine flying hours instead of engines, and Michelin charges per kilometre instead of per tyre. Today, AI is lowering the threshold for autonomously delivering, measuring, and acting upon results, bringing outcome-based models within reach for almost any business. MIT CISR expects business models to become steadily more outcome-oriented as AI takes on more of the work itself, with customers pushing for a tighter link between price and the value they actually receive.

In practice, however, the numbers look different. Bain's analysis of around 200 B2B software companies found that only about one in ten AI pricing models is tied to outcomes. The rest charge for the work the AI does, such as computing time or credits used, or for outputs it produces, such as drafts or updated records.

While there is no shortage of pilots and experiments with outcome-based models, scaling them into lasting business remains hard. Many industrial companies have spent years learning to deliver, measure and monetise outcomes. Their experience shows where scaling these models often gets stuck. This article draws on a study of nine industrial companies with outcome-based business models that I conducted with fellow researchers, recently published in Industrial Marketing Management.

Scaling is not the same as growth

Scaling has become a catch-all word, often used as a synonym for growth. But a business can grow without being scalable. At its core, scaling is about continuous growth: creating more value without costs rising at the same rate, so that outputs grow faster than inputs.

AI highlights the distinction between growth and scaling. In traditional software, one more user adds almost nothing to computing costs, but with AI, every task carries a running cost. These costs are falling fast, but AI is also taking on longer and more complex work, so the cost of serving one customer can be very different from the next.

Scaling should not be treated as a separate phase that begins only after initial pilots and experiments. Instead, companies need to build it directly into their business model: how value is created, delivered, and captured. And as their business matures, the best logic to scale also changes.

Four ways to scale

In our study, we examined the scaling of industrial outcome-based business models through four distinct scaling logics. Because these logics are drawn from general scaling research, they offer a useful lens for any business that commits to delivering outcomes:

Economies of scale

How can we deliver more value with fewer resources through standardisation and repetition?

AI-enabled customer support is a good example. It is one of the few use-cases where outcome-based pricing has gained traction, because a resolved conversation is often easy to measure and can be defined the same way for every customer. If the outcome has to be negotiated customer by customer, every deal becomes a customised project, and scaling gets much harder. And the same applies to delivery: if every new customer requires a customised set-up or data clean-up, volumes may grow, but margins will not.

Economies of scope

How can we broaden and deepen what we do for customers by using the same capabilities across more offerings?

Customers see the most value in AI that completes whole jobs. The vendor that owns the whole job can also own the outcome.

The study found the same pattern in industrial companies. The most mature companies took over entire processes from their customers and expanded their services within them. One chemicals company built services that predicted what was about to happen at a customer’s plant and adjusted the chemicals automatically, based on real-time data. When AI handles monitoring and adjustment automatically, the collaboration no longer requires an expert at every customer’s side, so the model can be scaled to more customers. At the same time, the risks grow: the more closely the provider is involved in the customer’s processes, the more its results depend on the customer’s systems, data and decisions, as well as on the other partners involved.

Economies of learning

How can we learn faster than our competitors and turn that learning into better processes and offerings?

In early-stage companies, the study found that learning stays with a few experts. As one manager put it: “if it requires an experienced expert to analyse the data, how can we scale it?” AI can help address this challenge by turning expert know-how into models that work without the human expert in the room.

Automation has a downside, though. The more expertise is built into software, the more repeatable the work becomes, but the less room there is for judgement in situations that depart from the norm. For example, Zendesk, a customer service software company, spent months testing unusual cases and defining what counts as resolved before it could charge customers based on outcomes. Defining the outcome, in other words, is often the first thing one must learn.

Economies of structure

How can we build our organisation, technology and partnerships so that complexity does not grow faster than the business?

McKinsey estimates that every dollar spent on developing an AI model typically requires three dollars spent on change management, such as training and monitoring results. With outcome-based models, selling changes too. When revenue follows usage, the work does not end with the contract. What matters is helping the customer adopt the solution and get it working, so that the intended outcomes are actually achieved.

In the study, the most mature companies built separate teams around outcome-based offerings with dedicated roles, from solution sales to data analysts. The same is happening with AI: as AI takes on more of the execution, people move from doing the work to setting goals and limits, following results and answering for the outcome.

Maturity decides which scaling logic matters

Our study identified three maturity stages based on how far a company has progressed from one-off pilots towards a more established way of providing outcomes to its customers. The four scaling logics carry different weight at each of these stages:

Early stage: scaling through learning

At this stage, scaling is built above all on learning: each pilot shows what works and what can later be standardised. Outcomes are hard to measure, however, and delivering them depends on a few experts, so lessons do not carry over from one project to the next. The biggest barrier to scaling is often the lack of standardised processes and reliable data.

Intermediate stage: scaling through standardisation

At this stage, scaling is built above all on standardisation: the same model can easily be replicated across more customers. The promised outcome is therefore usually one the company can guarantee itself, such as equipment availability. The customer’s broader benefit, such as more efficient production, only materialises when the customer acts on the company’s analyses and recommendations. Yet the company often has only limited influence over this, which makes scaling further difficult.

Advanced stage: scaling through scope and structure

At this stage, the company often enters into performance-based contracts in which it promises the customer direct business results, such as a more efficient process. This is possible because the customer hands over responsibility for entire processes. Scaling is built above all on scope and structure: the company deepens and broadens its services within its customers’ processes. This often requires coordinating teams, partners and systems. The challenge is that the more parties are involved, the harder the whole becomes to manage. The biggest barrier to scaling, then, is aligning teams, partners and systems.

These stages are not steps that every company should climb in order. Which stage a company aims for is a strategic choice that depends on its capabilities, risk tolerance and scaling opportunities.

Three things follow in the AI era:

Outcome definition cannot be outsourced to AI. AI can analyse data, suggest metrics, spot exceptions and measure results automatically. But the company and the customer still have to agree on what is promised, how the result is verified, who carries the risk and how disputes are resolved.
The intermediate stage can also be a destination. According to our study, some companies may find a strong position at exactly this stage: standardised service agreements bring predictable revenue without the provider having to take deeper responsibility for the customer’s processes. Bain’s market analysis shows why predictability matters: about four in five vendors introducing AI pricing sell prepaid credit or usage packages, because they give customers predictable budgets and vendors stable revenue.
Trust determines how far you can go. In a buyer survey reported by BCG, about a quarter of buyers said results often depend on factors the vendor cannot control. AI can improve this: instead of handing recommendations to the customer and depending on them being acted on, it can act on them itself. But that only works if the customer hands over control of its processes, and customers only do that when they trust the provider. In our study, trust built up stage by stage: in the early stage, customers were cautious about sharing data; in the intermediate stage, results were monitored together; and in the advanced stage, customers handed entire processes over to the provider.

Whether outcome-based models scale depends less on the technology than on a company’s ability to standardise, learn, organise and earn the customer’s trust, so that it can gradually take on more of the customer’s process. The place to start is asking which stage the company is at today and which scaling logic fits it.