AI Adoption Is a Cultural Problem, Not a Technology One

More than a decade after Paul Boag published Digital Adaptation for Smashing Magazine, the consultant sees the patterns he documented in 2014 replaying across enterprise AI rollouts.

Boag argues that senior leaders and IT teams repeat the central error of the previous era: treating a people problem as a technology rollout. "Organizations need to recognize that AI is a cultural shift as much as a technology," he writes, "and that shift has enormous implications for their staff."

What does the rollout look like today?

The pattern is consistent. IT departments procure company-wide Copilot Studio licences, ship a single round of generic training, and consider the work finished. Meanwhile, frontline staff receive tools they do not know how to apply to their job environment and feel pressured to demonstrate productivity gains from systems they never asked for. This time, the pressure carries a sharper edge. Staff who have read industry headlines about AI-driven layoffs do not trust reassurances from management, and that anxiety blocks adoption at the moment organizations need it most.

Who benefits from AI productivity?

Boag points out that the productivity case for AI benefits the employer, not the employee. A worker who finishes tasks faster still works a standard schedule and receives no reclaimed time or pay adjustment. The case for adoption, he argues, must be built around what improves the working life of the individual rather than the executive spreadsheet. Practical entry points include helping staff identify the tasks they least enjoy and showing how AI can shorten those specific workflows. Over time, organizations may need to attach incentives such as career progression, compensation adjustments, or shorter working hours.

How much time does adoption actually require?

Adoption costs more than a training slot. Staff must rebuild decades of procedural knowledge around new tool behaviour, and most workers enter that learning process already overworked. Boag recommends role-specific training delivered in the language of the job. He draws a sharper line at peer-led adoption. Every organization contains a small group of unofficial enthusiasts experimenting on their own. Bring those people together, give them structured time to share findings, and let social proof carry the message faster than any memo from leadership can.

What happens when failure is not safe?

Organizations that demand AI adoption while continuing to grade staff against unchanged KPIs send a contradictory signal. Boag reports hearing about employees reprimanded for excessive token usage weeks after being told to embrace the tools, and others lectured into paralysis by governance documents. A culture of fear around mistakes halts experimentation, and halted experimentation halts learning. Management must adjust review criteria, relax governance during the learning phase, and accept a temporary drop in measurable output as the price of capability building.

What does shadow AI reveal about the toolset?

Shadow IT emerged during the digital transformation era when staff signed up for Dropbox and Google Docs because sanctioned tools could not do the job. The AI equivalent is already visible. Boag describes a UX team whose leadership supports experimentation and whose members proposed an AI research repository and AI personas. Months of steering-group bureaucracy followed because Copilot could not deliver the required capability, and the team's patience thinned. Other staff skip the steering group entirely and open personal ChatGPT or Claude accounts on personal devices, feeding company data into tools the organization cannot monitor. Restrictive policy pushes the risk outside the perimeter rather than removing it.

Boag treats shadow AI as a diagnostic signal rather than a disciplinary problem. "Instead of trying to stamp out shadow AI, organizations should learn from it," he writes. The next step is to ask staff what they use personal accounts for and what the official toolset prevents them from doing. Governance, in his model, fits on a handful of posters in the manner of gov.uk's design principles, and the rest relies on staff trust.

What is the cost of moving too slowly?

Boag concedes that adaptive rollouts feel dangerous to leadership teams trained to minimise variance. He argues the greater risk is delay. Organizations that centralise every AI decision through a steering group and lock down the toolset will lose the experimental energy required to identify real use cases. The companies that learn fastest will be those that treat AI adoption as change management rather than procurement, and that begin by listening to the enthusiasts already at work inside the organization.