Q: Can we develop a robust staff AI usage policy that prioritizes security and safe usage without stifling people’s ability to be creative and explore new tools and use cases?
—Matt Kung, Marketing and Data Systems Manager, Mission Ridge, Wash.
A: It’s all about your data. Control the data, then let people play freely everywhere else. Don’t try to control every tool.
Start with a traffic light system. Sort your information into three buckets and build the policy around those.
Green is anything public: snow reports, event copy, social posts, trail descriptions, published pricing. Use it in any tool.
Yellow is internal but not sensitive, like meeting notes, draft budgets, marketing strategy, and vendor comparisons. That data only goes into company-approved tools on business plans, where your data isn’t used for training and you have admin controls.
Red is the sensitive stuff: guest and pass holder personal info, payment data, waivers, ski patrol incident reports, and anything in HR. Red never goes into an AI tool unless IT has specifically signed off on that exact use.
Once people know which bucket they’re in, you’ve removed 90 percent of the risk, and staff know where they can experiment. To stop people from using sketchy tools, pay for a business-tier account on one or two main platforms (Claude, ChatGPT, Copilot or Gemini), making it the easy default. Then, create a fast lane for the approval of new tools: i.e., a short form that asks what it is, what data would touch it, what it’s for, and promise a “yes” or “no” within a few business days. If your approval process takes six weeks, people will route around it.
Newer AI tools connect to your email, drive, calendar, and CRM, and can take actions on their own. Require approval to connect an AI tool to a company system, and scope access to only what the task needs.
A human should review any AI-generated guest-facing or published content before it goes out. And be careful with AI-generated imagery. Real photos are better.
Three other things: First, share your policy with your seasonal staff; they’re often the most likely to use free consumer tools. Build a segment about AI use into seasonal onboarding. Second, create a no-blame “oops” channel. If someone pastes something red into the wrong tool, you want them reporting it in five minutes, not hiding it for five months, so make reporting safe and fast. Third, assign the policy an owner who reviews it quarterly. AI tools change every few months, so a policy written in 2025 is already stale.
Finally, spend more energy on sharing wins than on policing. A monthly 15-minute “here’s what I built with AI” show-and-tell promotes safe, creative adoption.
AI usage policy template. Feel free to steal Eternity’s template. Edit it to fit your resort. Download: https://bit.ly/ai-usage-policy-template.
Q: How is the merger of Cowork and Chat going to change how to use Claude and, specifically, the relevance of Projects?
—Stephen Kircher, President and CEO, Boyne Resorts
A: The short answer: users can stop picking a tool and start picking a workspace. Formerly, you picked a place for each task: Chat for a quick question, Cowork for work that needed files, connectors, and a longer run. Now, Claude will choose between quick answers and deeper work based on each prompt, while users can stop, redirect, or adjust its approach. So, a conversation can stay short, or turn into research, a report, a spreadsheet, or a presentation that Claude works on after you step away.
A few practical shifts come with this. Claude can use local files and apps while your computer is open, then keep working on its own cloud computer when you close it. The new Docs and Slides live right in the conversation, and users can edit the outputs and export presentations as PowerPoints or PDFs.
Projects gets more important after the merge. When every chat can turn into hours of autonomous work, a Project helps clarify the scope.
In the Cowork version, a project collects everything Claude needs for a recurring area of work: the local folders to read and write, standing instructions, useful links, and a dedicated memory store. Background work plugs in too, since long-running work picks up the same folders, instructions, and memory. Think of Projects as a smart folder that can store documents, chats, and memory. Cowork then has a clear direction of what to do with your request, instead of taking a best guess.
You might have one Project per recurring job: season pass pricing, capital projects, guest experience feedback. Each gets its own instructions, files, and memory, and the memory stays put.
For anyone not using Claude, Projects also exist in ChatGPT, and functions in exactly the same way, with what ChatGPT calls “Work” instead of Cowork.
Got a question for the AI guy?
