What Grassroots, Employee-Built AI Agents Reveal About Real Adoption

Workday's fiscal Q2 results showed more than 5,500 customers with an AI agent moving from pilot into actual production use — and inside its own walls, Workday's Sana rollout reached 80% of its roughly 21,000 employees with 12,200 weekly active users, in what the company calls a record 60-day internal rollout. Two different numbers, the same underlying pattern: once an AI tool clears a real usefulness bar, adoption spreads fast and mostly on its own — the harder, more useful question is what actually gets a tool to that bar in the first place.

Mandated Adoption Versus Adoption That Spreads Itself

There's a meaningful difference between an AI tool employees are required to use and one that spreads because people who tried it told colleagues it was actually worth using. A 60-day internal rollout reaching 80% of a 21,000-person company with genuine weekly active usage, not just provisioned accounts, looks far more like the second pattern than the first — mandated tools rarely hit that kind of active-usage ratio that fast, because compliance and genuine daily use are different things entirely.

What Makes an Internal AI Tool Actually Spread

• It solves a task people already do often enough that saving time on it is immediately, personally noticeable — not a hypothetical productivity gain described in a training deck
• The first attempt has to work well enough that someone tells a colleague unprompted — a tool that requires three tries before it's useful loses most of its potential word-of-mouth spread before it ever gets any
• Access has to be genuinely frictionless — an approval process, a separate login, or a request form between an employee and trying the tool kills a meaningful share of the curiosity-driven trials that grassroots spread depends on
• Leadership visibly using it matters less than a peer using it — grassroots adoption spreads through lateral trust between colleagues doing similar work, not primarily through top-down example-setting

Why This Matters More Than the Vendor's Own Adoption Numbers

When a vendor reports a customer adoption number like Workday's 5,500-plus pilot-to-production customers, that's a useful external signal — but the more actionable lesson for your own organization is the internal-rollout pattern, because that's the one you actually control. You can't replicate a vendor's customer base, but you can replicate the conditions that make an internal tool spread on its own: genuine task-level usefulness, low-friction access, and a first impression good enough to generate unprompted word-of-mouth recommendation.

Applying This to Your Own AI Rollout

Before mandating any AI tool company-wide, consider running a genuinely voluntary pilot with a small group first and measuring whether it spreads on its own beyond that initial group without additional prompting — that's a far more honest read on real usefulness than a survey asking people if they'd find a hypothetical tool valuable. A tool that has to be mandated to reach meaningful usage is telling you something important about whether it's actually solving the problem it was bought to solve, information worth having before a company-wide rollout, not after.