Priya was told to "figure out AI" for her 200-person company on a Monday.
By Friday she'd shipped a rollout plan: a Claude Team license for every employee, a lunch-and-learn, a Slack channel for tips.
Three months later, adoption had flatlined at people asking the tool to summarize meeting notes. Priya herself had used it maybe a dozen times, mostly to punch up an email.
When her CEO asked why "the numbers" hadn't moved, she didn't have an answer. In part, because she didn't know enough about AI herself to even know what "moved" should even look like.
It's hard to lead adoption of technology you don't use.
That's what this playbook is about, in three parts:
- Part I: Single Player: You start to play the game and become an AI power user with your own workflows and routines.
- Part II: Multiplayer. You help others to connect and benefit from their previously individual, isolated AI experimentation.
- Part III: Transformation. You change how the business works — this is the "AI native" everyone is talking about — in a durable way.
When the leadership team at the SaaS company where I run People & Operations asked me to steward internal AI adoption, I started with governance alongside L&D initiatives.
The challenge was: I wasn't fully bought into AI yet.
I hadn't had my own "aha" moments, so I couldn't talk about it enthusiastically, or predict the challenges our team were going to run into as they were ramping up.
So, I took a step back and challenged myself to spend 2-weeks where every single task started from within Claude (our preferred AI tool). That was the unlock.
It slowed me down for 2-weeks and has sped me up ever since.
Think About AI Adoption Like Becoming Familiar With A New City
1. Tourist. You’ve tried it. You ask it fun questions. It’s a novelty.
What it feels like: Confusion.
You can’t think of what you’d use AI for other than making funny cat videos (and the cast have 5 paws anyways). To level-up, don't wait for someone to hand you a use-case list. Get curious. Pick one thing you're passionate about or get frustrated by and try to see how AI can enhance or improve it.
2. Resident. You draft emails, docs, and slides with AI. Sometimes you get a faster output (other times you're slowed down), but you're mostly working the same way you were before with some new tools.
What it feels like: Anxiety.
This is where you start to suspect you’re behind, because you can tell other people are getting more out of AI than you are. That instinct might be right, but it's also dirty fuel. Nobody learns well while panicking. To catch-up, focus on finding something to get excited about with AI. Then, dive in.
3. Citizen. You’ve handed AI entire workflows. It now does things you used to do by hand, start to finish.
What it feels like: Amazement
The first Monday a routine drafts your manager update before you’ve finished your coffee, you start hunting for the next thing to hand over. This is the fun part of the curve. Ride it while it lasts.
4. City Planner. You’ve got multiple workflows and agents running simultaneously, and are building in a collaborative way with others.
What it feels like: Cost-conscious.
You start wondering what all this costs and whether you’re being wasteful. That worry is a milestone, not a problem (for now). It means you’re using it enough to matter. Step 9 will help you get a handle on it.
Every company measuring “percent of employees who logged in” is counting tourists and residents and calling it a win.
​
Priya’s Slack channel was full of resident-level stories.
How to Become an AI Citizen
All of these examples are based on Claude as the AI tool of choice:
- Create a Project.
​A Project is a workspace with its own memory and knowledge base, so context compounds instead of resetting every session.
​
- Write your context file.
​This is the most valuable hour you will spend all month.
​
In Claude.ai it’s the Project instructions panel; in Cowork or Claude Code it’s a literal CLAUDE.md file that gets read at the start of every session. Put in your role, your team’s names and what they own, your recurring meetings, your writing rules, and three real samples of your own writing.
​
There are a lot of online prompt templates to help with this. Here's one example from Ruben Hassid.
​
- Start everything from Claude for 2 weeks.
​This will slow you down. At first. But, it will help you learn the limitations, and true capabilities of Claude.
​
Did you know you can send Slack messages from Claude? You can. Emails? Yep. Draft a Notion doc and have it make those annoying databases? Oh yeah.
​
- Distinguish between Chat, Cowork, and Code, plus models.
​If you're using Claude and getting inaccurate results, you're probably just using the wrong tool or too basic of a model.
​
Experiment with starting different tasks in Chat, Cowork, and Code, using Haiku, Sonnet, and Opus models. See how the output varies.
​
No, you don't need Fable. Unless you're doing multi-day coding sessions, save the money.
​
- Talk instead of typing. ​
Install a dictation tool like Wisprflow and have a full conversation with AI. (Claude also has a native talk to text feature.)
​
Ruben Hassid clocks it at roughly 4x faster than typing, but speed isn’t the main gain.
​
When you talk, you dump 90 seconds of messy, useful context.
When you type, you write three tidy sentences and strip out context that would have been helpful.
​
- Connect one tool that you spend 1+ hours a day in.​
Connectors let AI read (and in some cases write or push information back to) your calendar, your notes, and your task list.
​
Check your company’s acceptable-use and data-privacy policy before you connect anything holding employee data or PII. We’re in HR. We don’t get to skip that step.
​
- Turn the prompt you’ve retyped five times into a Skill, then into a Routine.
​When you notice you’re pasting the same instructions weekly, make it a Skill you invoke by name. Once it’s reliable, put it on a schedule as a Routine so it runs whether you remember it or not.
​
Remember: Skill first, Routine (or Scheduled Skill) second. Scheduling something that isn’t reliable yet just means you get bad outputs automatically.
​
- Practice refining AI work. Don't ship AI slop.
​Once you start using AI more, it's tempting to ship it's first draft as a completed product. Don't. You'll get a reputation for someone who is using AI recklessly to replace original thought and taste.
​
Take time to make sure that the initial AI draft sounds like you with a Claude.MD setup, but always edit and refine.
​
- Learn what you’re spending, and why.
​A few things to know: First, models are tiered: use the fast, cheap one for mechanical work like reformatting or tagging, and save the deep-reasoning model for genuinely hard thinking. Most people do everything on Opus (High) and spend too much.
​
Second, your context file and your Project knowledge get read every session, so keep them tight. A bloated context file is a recurring cost, not a one-time one.
​
You can also use tips to reset token usage windows to ideal times (e.g., have a Routine run that pings "Hi" at 7 AM so your 5-hour window ends at noon and resets instead of at 2 PM).
​
If all of these felt basic to you, congratulations, you're doing well on your AI adoption journey.
​
Check out these more advanced AI use cases for HR from PromptMates, including one from one of my direct reports (Emily) on scaling recruiting ops as 1-person team.
The Four-T Playbook
Every edition, I share a proven tip, trick, tactic, or template. This time, it's a:
👉 Template: If you want a more guided step-by-step AI tutorial, I recommend Hilary Gridley's Couch to 5k for AI.
It has 30 days of content that take 30-minutes per day to complete. It's free and full of rich prompts and specific connector suggestions.
Final thoughts
This is Part I of the Multiplayer AI Playbook: single-player mode. If you follow steps 1 to 9, you'll advance significantly in your AI journey.
If this is useful, forward it to whoever at your company just got handed "figure out AI." They'll want all three parts.
Until next time,
Melissa
P.S. This newsletter represents opinions and experiences from various HR professionals, it is not advice. Please always consult with appropriate legal counsel on AI use at your company.