Beyond tokenmaxxing: is AI governance the ultimate value driver?

AI success can’t be measured in prompts and tokens alone. Increasingly, governance is how organisations turn AI investment into value.

18 November 2026

Publication

Loading...

Listen to our publication

0:00 / 0:00

For the past couple of years, the success of enterprise AI has mostly been judged on activity (more users, more prompts, more tokens).

That made sense in the first wave of adoption, when the focus was on experimentation and scale. But AI is no longer a novelty; organisations are realising that not every employee needs the most advanced tools, and the economics of AI use are changing.

As AI is adopted at scale and fixed licensing models give way to consumption-based pricing, the cost of adoption rises. Advanced models are more expensive to run. Every prompt, workflow and automated process has a cost, and the rise of agentic AI could push consumption higher still.

So boards and investors have started asking a different question: what value is all this activity actually creating?

Some call this a move from "tokenmaxxing" to "tokenminning". "Valuemaxxing" may be a better label: the transition from maximum AI activity to maximum business impact.

AI governance is still essential for control, risk management and compliance, but leading organisations are also using it to create value, by deciding where AI belongs, what success looks like and whether an initiative should be scaled, changed or stopped. In this article we explore three areas where governance is central to maximising real value from AI, and offer key considerations for leaders to bear in mind.

1. Surfacing and prioritising high-impact use cases 

Most organisations can identify hundreds of possible AI use cases. Yet not all deliver the same value, and not every business problem requires an AI solution. With implementation time and costs rising, the priority should be the use cases where AI makes the biggest difference.

Effective governance frameworks give organisations a clear view of how AI is used, and a way to weigh new opportunities against strategic fit, expected business impact, cost, risk and scalability.  This creates a structured basis for investment decisions, implementation of new tools, and ensures resources go to initiatives that can deliver real value, instead of being dispersed across disconnected experiments and low-priority tasks.

This assessment should be informed by a clearly articulated AI risk appetite. Sorting use cases into "permitted" and "prohibited" categories is only part of it;  organisations also need to decide where they are willing to accept uncertainty, where additional safeguards or senior approval are needed, and where the risk outweighs the likely benefit.

In practice, that means a defined process for identifying and evaluating AI use cases, keeping them aligned with business objectives, maintaining visibility across the organisation and tracking return on investment, with regular reviews to capture lessons and check whether the expected value is being delivered. A defined process will typically include:

  • a single route for submitting or identifying proposed use cases;

  • clear business ownership;

  • an initial value and risk assessment;

  • proportionate approval pathways;

  • agreed success measures and review dates; and

  • criteria for scaling, modifying, suspending or retiring the use case.

2. Measuring what actually matters

Governance also shapes how organisations measure success.

Early AI adoption was usually measured by prompt and token volumes, active users and adoption rates. Those figures say something about engagement, but reveal very little about business impact.

Regular use of AI systems doesn't necessarily mean productive or valuable use. A high prompt count could point to weak prompting skills, low-priority business activities, or AI being used for jobs that conventional tools could do just as well, such as searching databases or attaching documents. Equally, some of the most valuable AI deployments may see fairly modest use whilst significantly improving customer outcomes, accelerating decision-making, reducing risk or enhancing quality.

Better visibility of how AI is adopted and used can help organisations to understand their return on investment, but monitoring needs careful design. Detailed employee monitoring raises privacy, data protection and employment law issues, particularly for multinational organisations operating across different regulatory regimes.  Poorly designed monitoring can create incentives to generate activity for its own sake rather than value. However, approaches that are proportionate, transparent and clearly explained to the workforce are more likely to support adoption and long-term value creation.

Time saved isn't the same as value created, either. Saving an hour only creates value if that hour is redirected to a more meaningful activity.

So when looking at the value being delivered by AI, the measures need to go further than adoption figures like prompt volumes and active user numbers. Those show how widely the technology is used, but the more important question is what it enables people to do, and whether that use makes a difference. Useful measures might include cases processed, but also turnaround times, better-quality decisions or fewer errors. Over time, businesses will want to know whether those improvements are translating into something more tangible, whether that is increased revenue, lower costs, reduced risk or better customer experience.

Strong governance helps organisations to move beyond activity-based metrics to measures such as output quality, efficiency, product development and testing timelines, customer experience, revenue, risk reduction and strategic impact. It also enables them to compare results across use cases and business functions, identify what works and move or reallocate investment accordingly.

The resources AI consumes should also be factored in when measuring value. Every prompt and automated workflow has an energy cost as well as a financial one, and for computationally intensive models or agentic systems running at scale, that cost can be significant. It's worth asking whether the model and level of use suit the task; a smaller model, a better-designed workflow or a conventional search may get the same result with far fewer resources. The discipline that cuts wasted spend can also cut avoidable environmental impact.In this sense, measurement is a mechanism for continuous improvement as much as for accountability.

Practical steps organisations can take to measure value include:

  • identifying a baseline for the activity before AI is deployed, then tracking the same measures after deployment and over time;

  • deciding how value should be measured for each specific use case;

  • tracking human intervention (correction and escalation volumes, for example) as an indicator of the value the AI tool is adding;

  • maintaining scorecards for each use case covering adoption, efficiency, quality, business value and risk;

  • appointing a named benefits owner for each use case, not just a technical or product owner;

  • reassessing each use case after a set period against its expected benefits and total cost, including integration, assurance, change management, human review and remediation, not just licence or token consumption; and

  • requiring business units to provide value metrics when submitting a use case for review.

3. Building trust 

With growing concerns about AI's impact on society, the environment and the workforce, trust is at a premium.

People trust AI when they have good reason to be confident in how it's being used. Employees, leaders and customers need to understand why a system is suitable for its intended purpose, where its limitations lie, and what happens when something goes wrong or an output needs to be challenged.

Each group needs something slightly different. Employees need confidence in the systems they use, leaders need assurance that risks are being properly managed, and investors want evidence that substantial spending on AI is paying off in order to have confidence in the decisions that leaders are making. Customers, users and other stakeholders increasingly expect AI to be used safely, securely, transparently and responsibly, with clear accountability.

Investors and other stakeholders are also looking more closely at AI's environmental footprint, including the energy and water used by the data centres behind it. Much of that impact sits in the supply chain rather than in an organisation's own operations, so governance that shows AI is being used intentionally , and not indiscriminately, will become part of how organisations account for their sustainability commitments.

In all of these ways, trust built through strong governance can help value creation. 

Strong governance builds trust through clear accountability, effective oversight and proportionate controls. It shows that AI systems are evaluated carefully and deployed responsibly, and that decisions can be understood, challenged and improved when needed.

Trust can also help organisations move faster. Employees are more likely to adopt tools they trust, leaders are more willing to back investment where governance is mature, and customers and users are more likely to engage with organisations that can show they use AI responsibly.  Governance shouldn't slow AI adoption down. Handled well, it gives organisations the confidence to scale quickly.

Conclusion 

Organisations using AI across different markets have a range of legal requirements to navigate, many of them not new. In the UK and EU, existing rules on data protection, cybersecurity, consumer protection, employment and equality still apply to how AI is developed and used. On top of these, the EU AI Act is coming into effect in stages, and what it requires depends on the organisation's role, the type of AI involved and how it's being used.

It's tempting to see governance primarily as a response to all this regulatory complexity, but the value that governance can create is much broader. As well as helping organisations keep pace with changing regulation, it points AI investment towards the right opportunities, supports proper measurement, brings more discipline to how resources are allocated and gives stakeholders confidence in how AI is used.

These objectives are becoming increasingly linked, and the same disciplines that support responsible AI deployment tend to help organisations achieve better ongoing business outcomes. Governance is becoming the connective tissue between innovation, value creation, trust and compliance.

This document (and any information accessed through links in this document) is provided for information purposes only and does not constitute legal advice. Professional legal advice should be obtained before taking or refraining from any action as a result of the contents of this document.