Shadow AI: Nope, Not the Latest Model Release and Why Attraction Operators Should Already Know the Term

Disney Imagineering made an interesting announcement this summer.

The team responsible for designing Disney parks and experiences is bringing generative AI directly into its design process through a partnership with Adobe. Imagineers will use customized AI models built around Disney’s own design catalog to help turn sketches into concepts, explore creative directions and produce franchise-accurate visual assets faster. The important part is not simply that Disney is using AI. It is that Disney is being deliberate about where the technology belongs, what information it is built on and how it fits into an existing workflow.

Now compare that with what is happening across the rest of the attractions industry.

A new UNESCO and International Council of Museums study surveyed more than 400 museums in 90 countries and found that 57% are already using AI. At the same time, 55% have no internal AI policy, strategy or guidelines, and only 17% provide AI training to their staff. Much of the adoption is still exploratory and driven by individual employees rather than an organization-wide plan.

That gap is where Shadow AI lives.

Shadow AI is simply the use of AI tools at work outside of an organization's approved systems, policies or formal processes. In an attraction, that might mean a marketing coordinator using ChatGPT to clean up an email, a manager asking AI to help with an Excel formula, a group sales employee summarizing meeting notes or a department head uploading a report because they need help understanding what the numbers are telling them.

Most of these employees are not trying to work around leadership or create a security problem. They are trying to get through the work in front of them.

That distinction matters because many attraction leaders are still discussing AI as though adoption is a future decision. Their employees may already have made that decision for them.

The policy conversation is already behind the behavior

The UNESCO/ICOM findings are particularly relevant because they look a lot like the attractions industry itself. AI is already being used for administration, communication, research, documentation, exhibition development and visitor engagement. In interviews and Attraxion.ai led group workshops, we uncover many use cases where our first ask back to the participant feels like it needs to be 'Is your boss aware you are using AI for that'.

That is also why a blanket instruction telling employees not to use AI will only address part of the problem. The workload that led them to use it in the first place is still there.

Someone still has to prepare the weekly report. Someone still has to respond to the guest complaint. Someone still has to summarize the meeting, organize the notes, write the follow-up or figure out why last Saturday's labor percentage moved in the wrong direction.

When employees discover a tool that can help them complete that work faster, some of them are going to use it.

The more useful leadership question is whether they understand the boundaries.

• Can a manager put guest information into a public AI tool?

• What about employee schedules, contracts, attendance figures, financial reports or membership data?

• Which tools has the company approved?

• Which information should never leave company systems?

• When does somebody need to check an AI-generated answer before relying on it?

• Without clear guidance, employees are answering those questions for themselves.
  

Disney's approach offers a useful contrast

Disney Imagineering is not simply giving employees access to a public chatbot and asking them to experiment.

Its work with Adobe involves customized Firefly models built around Imagineering's own licensed and proprietary assets. Adobe says those models are being incorporated into the tools and workflows Imagineers already use, with specific initial applications such as converting sketches into rendered concepts and generating creative assets that remain consistent with Disney franchises.

There is an important operating lesson in that, even for an attraction without Disney's resources.

The lesson is not that every operator needs a proprietary AI model.

It is that AI becomes much more useful when leadership decides where it belongs.

A regional waterpark does not need Disney's technology budget to decide that managers can use an approved AI tool to summarize non-sensitive meeting notes, but cannot upload employee records. A museum does not need an enterprise AI team to identify three repetitive administrative processes that employees are already trying to solve with ChatGPT. A family entertainment center can establish approved tools and basic rules without producing a fifty-page governance document.

The sophistication of the solution should match the organization. The need for clarity does not disappear because the organization is smaller.

Shadow AI can also tell you where the operation is struggling

There is another reason leaders should understand how employees are already using these tools.

AI use can expose inefficient work.

When a manager repeatedly turns to AI to summarize information, perhaps that information is too difficult to find or consume. When someone uses AI every week to rebuild a report, perhaps the reporting process itself needs attention. When employees are creating their own tools to answer common questions, perhaps important knowledge is living in too many systems or with too few people.

Those behaviors are worth examining before anyone decides which software to buy next.

One of the best questions an attraction leader can ask right now is simple:

What have you used AI to help you get done at work?

Ask it without making the employee feel as though they are admitting to something.

The answers may uncover a policy issue, but they may also reveal a three-hour process that should take twenty minutes, information that has become unnecessarily difficult to access or an employee who has already discovered a genuinely better way to complete part of the job.

The current AI conversation in attractions is not really about whether employees are going to use these tools. The evidence suggests that many already are.

The real work now is understanding where AI has entered the operation, setting reasonable boundaries around its use and deciding which of those experiments should become part of the way the organization actually works. That's part of the work we help operators uncover here.