“Sign the contract. Deploy the tool. Train the users.”
I’ve watched this playbook work for twenty years of enterprise software. Slack, Jira, Salesforce. Buy it, roll it out, mandate adoption, move on.
I’ve been watching enterprises try the same playbook with AI agents. It doesn’t work. And I think I know why.
You Can’t Mandate Trust
Here’s the thing: traditional tools ask people to share information. Drop a message in Slack. Log a ticket in Jira. Update a field in Salesforce. You’re giving the tool access to data.
Agents ask for something fundamentally different. They ask you to delegate authority. Draft my email. Make this decision. Talk to my systems on my behalf.
That’s not sharing information. That’s handing over the wheel.
And the stakes feel completely different. Worst case with Slack? An embarrassing message in the wrong channel. Worst case with an agent? Your reputation. Your relationships. A decision made with your name on it that you didn’t actually make.
You can mandate Slack. You can’t mandate trust.
Why This Is Harder Than It Looks
There’s another layer here that I don’t see people talking about enough.
When someone uses a calculator, nobody says “oh, so you can’t do math?” When someone uses spell-check, nobody questions their writing ability. But when someone uses an AI agent to draft a proposal? Suddenly it’s “did you actually write this?”
It feels like cheating. Not because the output is bad. But because the AI does the “hard part” and there’s no established norm yet for what’s okay. We’re in this awkward middle where using an agent feels like getting away with something. And people won’t spread a tool they’re embarrassed to admit they use.
And honestly, there’s an even harder issue underneath that. Some people don’t just feel awkward about agents. They feel threatened by them.
With Slack, the worst pushback was “I don’t want another chat app.” Annoying, but not existential.
With agents, the pushback is “this thing is coming for my job.”
I’ve seen what happens when that fear takes hold. People don’t just decline to adopt. They become antibodies. They find reasons it doesn’t work. They point out every mistake. They make sure the pilot fails. Not because the tool is bad, but because it threatens their identity.
You can’t argue with existential fear. You can only let results speak.
The Trust Ladder
So if top-down mandates don’t work, what does?
I’ve been thinking about this a lot. And I keep coming back to the same pattern: the playbook that worked for Slack, Notion, and Figma. Bottom-up adoption. But adapted for higher stakes.
I think about it as a trust ladder:
Level 4: Culture Shift ← "This is just how we work now"
↑
Level 3: Enterprise ← IT blesses it, leadership adopts
↑
Level 2: Teams ← "You gotta try this"
↑
Level 1: Individual ← Start hereEach level requires earning trust at the level below. You can’t skip rungs. And each rung has its own moment that pulls people up.
The Moments That Build Trust
Level 1: Skeptic to Individual
This is the foundation. The moment someone goes from “this is probably hype” to “okay, that actually worked.”
Picture this: you’re a backend engineer. Python is home. But a task lands on your desk. Fix a bug in the iOS app. Swift. You’ve never written a line of it.
Old world: spend two hours reading docs. Copy-paste from Stack Overflow. Break things. Ask the mobile dev to bail you out. Feel like an imposter the whole time.
New world: you describe the bug to the agent. It reads the codebase, finds the issue, explains why it’s happening in terms you understand, and drafts the fix. You review it, learn something, ship it.
That’s the moment. Not “AI wrote my code.” But: “I just did something I couldn’t do alone, and I understood it well enough to own it.”
That’s trust. Not blind faith. Earned confidence.
Level 2: Individual to Team
This is when trust becomes contagious. Someone had an experience impressive enough to grab a colleague and say “let me show you this.”
I’ve noticed the stories that spread aren’t about efficiency. They’re about surprise. “You know that monthly compliance report we all hate? I described what I needed, and the agent pulled data from three systems and had a draft ready in minutes.” That kind of thing travels fast at lunch.
The key? The pain has to be shared pain. If everyone on the team dreads the same task, and one person makes it disappear, the tool sells itself.
Level 3: Team to Enterprise
This is where it stops being “that cool tool” and becomes infrastructure. I’ve seen this tipping point in organizations. It usually sounds like a question in a leadership meeting: “wait, how many teams are using this already?”
Multiple teams depend on it. There’s real data showing impact. Someone puts together a slide showing hours saved or velocity gained. The question shifts from “should we adopt this?” to “we can’t NOT have this anymore.”
Level 4: Enterprise to Culture
Now it’s identity. The org sees itself as AI-augmented. It shows up in job postings. New hires expect it. Not having agents would be like not having email.
This is the hardest rung, and honestly, most companies aren’t here yet. But the ones that get here will have a compounding advantage that’s almost impossible to catch.
What This Means For You
Here’s where this gets practical.
If you’re building an agent product, stop selling to CIOs. Win individuals first. Design your entire onboarding around that Level 1 moment. Make the first five minutes magical. If someone can’t feel earned trust in their first session, your enterprise sales motion doesn’t matter. Your go-to-market is bottoms-up, even if your contract is eventually top-down.
If you’re buying for your organization, don’t mandate adoption. Look for organic traction. Ask your teams: “what are people already using? Why do they love it?” Follow the trust. It knows something that procurement doesn’t.
If you’re an individual deciding whether to try one, your personal experience IS the pilot program. You don’t need permission from IT to find out if something earns your trust. And when it does, your recommendation will carry more weight than any vendor deck ever could.
Where It Actually Begins
The enterprises that win with AI won’t be the ones with the biggest contracts or the most ambitious rollout plans.
They’ll be the ones with a thousand individual yeses.
You can’t buy trust at the enterprise level. You have to earn it one person at a time.
Start with one person. One real task. One moment of earned trust.
That’s where enterprise AI actually begins.