The million-token question: how do you maximise your return on AI investment?

Over the last few years, Interactive has invested significant time and effort into building AI capability across the business. With AI set to transform how we work, we wanted our people to understand the tools available to them. That meant encouraging experimentation, exploration and hands-on learning.

The initiative was successful. So successful, in fact, that when a frontier model moved from being included in our subscription to a pay-as-you-go offering, we temporarily paused access while we evaluated its costs and business value. The level of interest that followed was telling. It wasn’t limited to AI enthusiasts or early adopters. People across the organisation had incorporated the model into their day-to-day work and were identifying tangible benefits from its use. 

We’re currently running the tool through a pilot group so we can better understand where it delivers the most value, the associated usage costs and how it should be used. That insight will help us make more informed decisions about future investment and budgeting.

We’re now entering a new phase of the AI journey. As tools such as Copilot, ChatGPT and Claude introduce more consumption-based pricing, organisations are being asked a different question. Not simply how to adopt AI, but how to maximise the value it delivers. For many organisations, that means building a clearer understanding of where AI creates measurable outcomes, what it costs to deliver those outcomes, and how those investments should be prioritised. At Interactive, that’s the challenge we’re focused on solving. While the answers will look different for every organisation, there are several lessons we’ve learnt so far.

The AI cost problem

 Most AI tools entered the market with simple pricing models, typically a flat subscription fee or generous usage allowances. Like many SaaS offerings before them, the focus was on driving adoption rather than closely managing consumption. As organisations embraced AI, the predictability of those costs meant usage rarely needed to be scrutinised.

That dynamic is now changing. As AI providers introduce more consumption-based pricing, organisations are having to pay closer attention to both usage and value. What was once a straightforward monthly expense is increasingly becoming a variable operational cost.

Forecasting those costs is far from simple. Even as the price per token continues to fall, overall spend can rise as employees adopt AI more broadly and newer, more capable models consume greater resources to complete the same tasks. Adding to the challenge is AI’s inherent variability. The same model can generate different responses to the same prompt, with each response consuming a different number of tokens. As a result, AI costs can be difficult to predict and manage with the same level of certainty as traditional software investments.

Cost is only part of the equation. Organisations must also account for the governance, security and risk management measures that support responsible AI adoption. As AI expands an organisation’s attack surface, it introduces new risks that need to be understood and managed. Taken together, these considerations reinforce the need for a clear business case and a strong understanding of where AI delivers meaningful value across the organisation.

The business cases (and hard calls) to come

The emergence of AI pricing transparency is forcing organisations to ask an important question: where is AI creating measurable value? 

Answering it means lifting the curtain on AI use across your organisation and gaining clarity on two key questions: how is AI being used, and what value is that usage generating? There is no universal answer. Every organisation will have different use cases, priorities and measures of success. The principle, however, remains the same: high AI costs aren’t necessarily a problem if they’re delivering meaningful business outcomes. The inverse is also true. Low-cost use cases that generate little or no value can quickly become a drain on resources. Left unchecked, they risk undermining higher-value initiatives when AI spending inevitably comes under scrutiny.

This is where some tough decisions will need to be made. Until now, the return on investment was largely about experimentation and capability building. Organisations invested in helping their people understand and adopt AI, with AI literacy becoming the primary measure of success. The focus is now shifting from adoption to outcomes. Leadership teams need to ensure AI investments are aligned to meaningful business objectives and be prepared to reassess use cases that deliver only marginal value. Importantly, this responsibility sits with leaders across the organisation, not just technology teams.

Your organisation already makes decisions like this every day. You determine which markets to enter, which capabilities to invest in and how your operating model should evolve. You also decide where technology can create the greatest impact. Viewed through that lens, AI is not a special category of investment. It’s another business decision, and it should be evaluated with the same level of discipline and accountability.

From AI adoption to AI maturity

Avoiding AI adoption because of concerns about cost is counterproductive. The real challenge is ensuring adoption is intentional, governed and tied to measurable outcomes. The business functions where AI can add value are well documented. Take McKinsey’s State of AI 2025 report. It points to cost benefits in software engineering, manufacturing and IT, and revenue gains in marketing, sales, strategy, finance and product development.

But potential value alone isn’t a justification for continued investment. The same report found that only 39% of organisations see any enterprise-level EBIT impact from AI, with most of those reporting less than a 5% improvement. McKinsey attributes this to organisations layering AI onto existing ways of working rather than redesigning work around it. Measurement has lagged behind adoption, too. Those trade-offs may have made sense when the priority was experimentation and uptake. As the conversation shifts towards ROI, they’re becoming harder to justify.

The next step is to bring AI investment closer to the business outcomes it’s intended to support. If AI consumption happens where the work happens, shouldn’t the people responsible for that work also have a role in deciding how AI investment is allocated?

Accordingly, we’ve adopted a model where business heads approve access to PAYG AI tools. If you own a P&L, whether internal or commercial, you’re responsible for approving employee access to PAYG AI. Business leaders are best placed to assess the value of an AI use case because they’re accountable for the outcomes it delivers. This also reduces the burden on central teams to define a universal “value delivered” metric, which can be difficult without understanding how AI is being used within each function. As AI use cases mature, organisations may also find that alternative deployment models, including local AI, become increasingly viable for specific workloads. Read our guide to sovereign AI to learn more and keep an eye out for our next article on why tokens aren’t always the whole story.

Make AI work for your organisation 

Recent discussion around rising AI costs is a sign of the market’s growing maturity. As organisations move beyond experimentation, the focus is shifting from AI adoption to AI value. The question is no longer whether to use AI, but where it creates the greatest impact and which deployment model makes the most sense for each use case.

Importantly, not every AI workload needs to run through a large language model. Increasingly, organisations will have options ranging from frontier models and pay-as-you-go services through to smaller, purpose-built models running locally on devices, at the edge, or within sovereign environments. The right answer will depend on factors such as cost, performance, security, governance and the value being created.

That means taking a more deliberate approach to AI strategy. Organisations need to understand where AI delivers meaningful outcomes, how those outcomes should be measured, and which technologies are best suited to delivering them. At Interactive, that’s the challenge we’re focused on solving.

Wherever your organisation is on its AI journey, the next phase is about maturity: matching the right AI capabilities to the right business outcomes and making informed decisions about where to invest. If you need help bringing AI ambition, AI economics and business objectives together, we can help.

 Find out more about our Data and AI solutions.