From Access to Enablement: Why AI Investment Alone Will Never Deliver ROI

Organizations have moved quickly to put AI into the hands of employees, driven by a fear of being left behind and pressure from executives to deliver on AI’s productivity promises. Consequently, we saw enterprise AI spending jump from $11.5 billion in 2024 to $37 billion in 2025, a 3.2 x increase within a single year. Despite this momentum, CFOs are now pushing back with many continuing to struggle to answer a fundamental question: what measurable business impact is actually being created from AI adoption? 

In the haste to introduce AI, the terms access and enablement have blurred into one, and organizations are left reeling from huge bills. It is time for a more considered approach to AI that takes into account the nuances of different roles and skill levels in driving scalable AI efficiency. 

Cost versus Value 

When an organization rolls out a new tool, it normally follows a familiar pattern: a problem is identified and measured; the tool is introduced into the relevant workflow, and its impact is measured against a pre-defined baseline. When it came to AI, that rule book went out the window, and overnight everyone was under pressure to be using it everywhere and for everything. This has resulted in organizations measuring cost and calling it value, with the rollout of thousands of licenses presented as evidence of being an AI-first company.  

Global organisations such as Uber and ServiceNow have been publicly burned by this, consuming their annual AI budgets within just months of 2026. Internal memos from Meta tell a similar story, with employees having racked up 73.7 trillion tokens in just over 30 days after AI usage was announced as a “core expectation” in performance reviews. However as the CTO Andrew Bosworth put it, “token usage alone is not a measure of impact of any kind.”  

Throwaway Work 

Fundamentally, employees are being equipped with tools they have never been taught to use and two errors follow. Contamination, which happens when someone runs their entire working month through a single chat, mixing customers and tasks with personal questions, without understanding that the model pulls the whole conversation back in as context. Information about customer A begins shaping answers about customer B and the output degrades in ways the employee cannot explain. Alternatively, the implication is erosion, which happens over long conversations as the context window fills, the tool compresses and summarizes, and the careful instructions given at the beginning quietly fall away. 

To combat this, employees need the skills to understand how these tools actually work. When equipped with foundational AI knowledge employees are empowered to pick-up on nuances in responses, for instance an employee who notices AI losing details will know to instruct it to write a handoff packet, review it for the rules it lost, and carry the work into a fresh context window with nothing missing. Without this fundamental understanding people conclude the chat is broken, throw away the work, tokens and time and start again. 

The Key to Transformation  

Instead, enablement is when real transformation happens, driving AI value through workflow understanding and the enforcement of reusable skills. The challenge in this instance is understanding that the opportunities for AI vary greatly across every business function.  

The way sales generates value differs from marketing, which differs from finance, so a single generic training program will never deliver consistent results across an organization. Instead, the answer is role-specific guidance and giving every role practical AI training tied to the work they actually do, so employees move beyond basic tasks into driving valuable outcomes.  

AI proficiency varies greatly across a workforce, so it is also crucial that training meets people where they are, equipping employees with the right tool for the right task. When tools bear the blame for missing the mark, the real failure is usually that nobody taught an employee which tool fit that specific moment in their workflow. 

The Importance of Proof 

The fundamental change required to prove the benefits of AI is to stop measuring usage, which matters only insofar as it costs money. Instead, catalog who has access, what they are actually doing with it, and frame that against what an optimized version of their role would look like. By applying this framework, value can then be measured in a way that connects directly to business objectives. 

One of the smartest approaches I have heard involved giving developers a fixed monthly token allotment while holding delivery expectations constant. Suddenly the work stopped being throwaway and engineers asked which genuinely painful problems were worth their tokens, underscoring the importance of guardrails in a successful AI roll-out.  

This field changes so quickly that a conversation from six months ago already feels dated, therefore enablement must be continuous rather than a one-time workshop. 

Organizations that stall at access will conclude that AI is a waste of money and abandon it, but if they had measured the right things and given employees the right training, they could have seen real business transformation. 

Those who achieve the greatest returns will have invested time and effort into helping people apply AI within meaningful workflows, measuring outcomes rigorously, and turning access into capability.