A technician performs maintenance and repairs on data-capture equipment at the Verlet Robotics office in Kigali, Rwanda. AFP
A technician performs maintenance and repairs on data-capture equipment at the Verlet Robotics office in Kigali, Rwanda. AFP

How AI tools can cause problems in the workplace

Alejandro Sposato
Alejandro Sposato

November 17, 2026


A professional told me recently that his employer gives him access to an approved AI platform, but the usage allowance is so small that he cannot rely on it for much of his work. His employer can legitimately say that it provides AI, yet employees cannot necessarily use it to the extent their responsibilities require.

This raises a practical question: are organisations paying enough attention to whether employees can actually use the AI tools they provide, rather than simply making those tools available?

Providing access to technology is not the same as making it useful.

There are good reasons to restrict AI usage. Advanced systems can be costly when deployed across large workforces. Employers must also manage cyber security, data confidentiality, regulatory obligations and inappropriate use. Unlimited access is neither financially realistic nor necessarily desirable.

Trouble arises when the resources provided fall short of what employees need to do their jobs.

Consider someone who regularly uses AI to analyse documents, prepare reports or evaluate information. Their organisation supplies an approved platform, but the employee quickly exhausts the allowance. The work still needs to be completed.

The employee can stop using AI, delay the task, request additional access or turn to another service through a personal account. For someone already paying for a personal subscription, that last option may seem like the easiest way to keep working.

Yet it is precisely the behaviour many organisational AI policies are designed to discourage.

The use of unauthorised AI applications, often called shadow AI, presents genuine concerns. Staff may expose confidential documents, customer information or internal communications when using platforms that have not been approved for work.

Providing a secure, authorised alternative is therefore sensible. But if that alternative cannot meet everyday needs, an organisation may have formally implemented AI governance while unintentionally creating incentives for employees to seek alternatives.

This does not have to reflect a deliberate decision. Budgets, cautious purchasing and uncertainty about demand can all explain why access is initially limited. Still, the effects deserve attention.

Suppose an employee uses a personal AI account and an information-security incident follows. The employer may reasonably point to the policy requiring staff to use the approved platform. The employee may have breached that policy. But the situation also raises a question about how workable the approved option was in practice.

Employees remain responsible for protecting confidential information. Running out of AI capacity does not justify sharing sensitive data with an unauthorised service. At the same time, employers should ask whether their rules give people a realistic way to do their work while following those rules.

Quote
A uniform allowance might be straightforward to administer, but it will not suit every role

Buying AI licences is relatively easy. Making them work across an organisation takes more thought. Employers need to know which tasks benefit from AI, how frequently people use it and where current limits become obstacles.

A uniform allowance might be straightforward to administer, but it will not suit every role. Someone who regularly analyses substantial documents may need more capacity than a colleague who uses AI only occasionally. Limits should reflect how the work is done, rather than an arbitrary figure.

Employees should also have a clear and reasonable way to request additional capacity. When someone says the approved tool is inadequate, that feedback may reveal a problem with implementation rather than an appetite for unnecessary spending.

If employees repeatedly turn to personal AI accounts, this is not only a compliance concern. It may also suggest that the approved tools are not adequately meeting their needs.

None of this requires unlimited AI access. It calls for restrictions that are proportionate, understandable and responsive to actual demands.

Organisations are right to establish approved platforms and clear safeguards. But those safeguards are more likely to work when employees have the resources to follow them. Otherwise, a gap can develop between what policies require and what day-to-day work permits, potentially increasing the very confidentiality and cybersecurity risks those policies seek to reduce.

Providing AI may satisfy a policy requirement. Making it usable is what makes the policy effective.

Updated: November 17, 2026, 4:00 AM