Fire Automations is my weekly newsletter where I cut through the hype and teach you when automation is worth it, what to automate, and how to implement it in real businesses
Thanks for reading - and let me know your thoughts down below.
You know what makes people hate “AI at work”? It’s not the tech. It’s the rollout.
One day you’re in a normal meeting. The next day someone says “we’re using AI now” and half the team hears “your job is being automated,” while the other half hears “cool, another tool I have to learn.” I’ve watched smart teams freeze up because the introduction was vague, mandatory, and way too big.
So here’s a calmer approach that actually works: start small, pick low risk wins, be transparent about what AI will and will not do, and let people opt in before you make anything a standard.
These are nine ways to introduce AI at work without spooking your team or lighting your Slack on fire.
1. Start with “AI for drafts,” not “AI for decisions”
The fastest way to freak people out is to let AI “decide” anything important before your team trusts it. The fastest way to build trust is to position it as a first draft machine. I’ve had the most success saying: “Use it to get to 60 percent, then you drive.” That framing keeps ownership with the human and makes quality control obvious. Start with things like rewriting an email, summarizing notes, or generating a rough outline for a doc. If the output is wrong, nobody dies. If it’s helpful, people feel it immediately.
2. Make it opt-in for 2 weeks (and say that out loud)
Mandatory AI adoption on day one is how you create quiet resentment. When I roll something out, I do a short opt-in pilot and I say the quiet part out loud: “Try it for two weeks. If it’s annoying, drop it. We’re learning.” That alone lowers the temperature. People who are curious will test it, and people who are anxious will watch from a safe distance. The funny part is the “watchers” usually become users once they hear a peer say, “Okay, this actually saved me time.”
3. Ban the phrase “everyone should use AI”
Nothing triggers skepticism faster than blanket statements. Instead, tie AI to specific roles and tasks. If you lead a team, pick one workflow per function: sales uses it for call recap follow-ups, marketing uses it for content repurposing drafts, ops uses it to clean up messy SOPs. You’re not preaching AI. You’re solving one pain point. And you’re giving people permission to ignore use cases that don’t match their job.
4. Use a “no secrets” policy about what’s happening to their work
A lot of fear comes from ambiguity. Tell your team what AI is doing with their content, what you store, and what you do not paste into tools. Even if you’re not going deep on IT policy, a simple set of guardrails reduces anxiety and prevents the classic mistake where someone pastes sensitive info into the wrong place. Keep it simple and human:
Don’t paste customer PII.
Don’t paste confidential financials.
Assume anything you paste could be reviewed.
Use approved accounts, not personal logins.
You own final edits and approvals.
That’s enough to make people feel safe without turning you into compliance theater.
5. Teach one “prompt pattern” and stop there
Most training fails because it tries to teach everything. Your team does not need prompt engineering. They need one repeatable template that makes the tool useful on day one.
The simplest pattern I’ve seen work is: “Context + goal + constraints + example.” For example: “Here’s a messy email thread. Write a reply that confirms next steps. Keep it under 120 words. Friendly but firm. Use this sentence as the opener.” When you hand people a format, they get better results and they stop blaming themselves when the first output is mediocre.
6. Introduce AI in a meeting by showing your own imperfect example
This is my favorite move because it kills the hype and the fear in one shot. In a team meeting, I’ll share a real before-and-after from my own work: a messy draft, the AI output, and what I changed. I point out the good parts and the garbage parts. That communicates three things immediately: it’s not magic, it’s not replacing you, and you’re still the editor.
If you want a script: “I used AI to get a draft faster. It hallucinated two details. Here’s how I corrected it.
Net result: I saved about 20 minutes and the final call was still mine.”
7. Pick one boring workflow and automate it end-to-end
AI feels scary when it’s abstract. It feels helpful when it removes a chore. The best early win is something boring that happens every week: meeting notes, weekly status updates, customer call summaries, internal FAQs.
A simple version that works without heavy engineering is: record the meeting, transcribe it, summarize it, then post the summary to your shared space. I’ve set this up in under an hour using common tools, and once it’s running, it saves a surprising amount of “Wait, what did we decide?” time. The key is to frame it as documentation hygiene, not surveillance.
8. Create an “AI wins” channel, but only allow real wins
If you want adoption without pressure, make it social and practical. Start a Slack or Teams channel where people share one screenshot or one sentence about what they used AI for and what it saved them.
The rule is simple: no vague posts like “AI is amazing.” Only tangible wins, ideally tied to time saved or a task completed.
When people see peers using it for real work, it normalizes the behavior. And it helps you spot patterns about where AI is actually delivering value versus where it’s just adding noise.
9. Put “human review required” in writing, then actually mean it
This is the trust anchor. If people believe AI output will be shipped without review, they will resist it. If people believe they will be judged against AI output, they will resent it. So write a simple standard: AI can assist, but a person is accountable for final quality.
I like to say: “AI can write it. You sign it.” It signals ownership, protects quality, and removes the weird pressure people feel to pretend the AI output is perfect. If you later decide to standardize certain workflows, you do it from a foundation of trust instead of forced compliance.
Final thoughts
The calmest way to introduce AI is to treat it like a helpful intern: great at drafts and busywork, not trusted with final calls without supervision. Start with one low risk workflow, run a short opt-in pilot, and make the rules clear enough that nobody has to guess. If you want a simple starting point this week, pick one recurring task that annoys your team and use AI to cut it in half, then share the before-and-after. Momentum beats mandates.