Enterprise AI Implementation: Lessons from the First 90 Days

There's a version of enterprise AI implementation that looks great in a vendor deck: plug it in, connect a few systems, watch productivity soar.

The reality is more nuanced. Not harder, necessarily — but different from what most teams expect. After implementing Glean and running AI readiness programs across Australian and New Zealand organisations, we've developed a clear picture of what the first 90 days actually look like.

Days 1-30: The quick wins matter more than you think

The temptation in the first month is to go big. Connect every system. Build the complete knowledge graph. Train the entire organisation.

Resist that temptation.

The organisations that get the most from AI in the long run are the ones that start narrow and prove value fast. We typically recommend connecting 3-5 core systems in the first week — the ones your teams use most. For many organisations, that's Slack or Teams, Google Workspace or SharePoint, and one domain-specific tool (Salesforce, Jira, Confluence).

Within days, people can search across those systems from one place. That alone — just unified search — generates the first wave of positive feedback. People start saying "I found that in 10 seconds instead of 10 minutes."

Those stories matter. They create internal advocates. And internal advocates are the single most important factor in long-term adoption.

Days 30-60: The adoption curve

Around week 4-5, something predictable happens. The initial excitement settles. The early adopters are using the platform regularly, but a significant portion of the organisation hasn't changed their habits.

This is normal. It's also the phase where many implementations stall — not because the technology isn't working, but because the change management isn't.

What works during this phase:

• Team-level champions: Identify one person in each team who's already using AI effectively and empower them to help their colleagues.
• Workflow-specific training: Don't train people on "the platform." Train them on their specific workflow. Show the sales team how to find competitive intel. Show HR how to surface policy answers. Make it relevant.
• Visible leadership use: When senior leaders use the platform in meetings — "Let me search that in Glean" instead of "Can someone find that and send it to me later" — it signals that this is how work gets done now.
• Remove the old path: If people can still do things the old way with no friction, many will. Look for opportunities to make the AI-assisted path the easier path.

Days 60-90: Agents and automation

By month three, your team has context. They understand what the AI knows and how it works. This is when agents become powerful.

We encourage teams to identify their first agent candidates during the adoption phase (days 30-60), but build them in the 60-90 window when they have enough experience with the platform to define the agent's behaviour effectively.

The best first agents are the ones that automate the most repetitive knowledge work:

• Answering the same internal questions repeatedly
• Compiling information from multiple sources into a summary
• Routing requests to the right team based on content
• Checking documents against standards or policies

What most teams don't expect

The permissions discovery

When you connect an AI search platform to all your systems, you sometimes discover that your permissions aren't as tidy as you thought. Documents that should be restricted are accessible to everyone. Team spaces that should be open are locked down. This isn't a problem the AI creates — it's a problem it makes visible. Most teams appreciate the transparency, even if it means some cleanup.

The knowledge gap revelation

AI search is very good at finding what exists. It's also very good at revealing what doesn't exist. When employees search for process documentation and find nothing, or ask about policies that have never been written down, it highlights gaps that were previously invisible.

This is actually one of the most valuable outcomes of the first 90 days. The gaps become a content roadmap.

The culture shift

The most underestimated change is cultural. When information becomes findable, the people who previously held power through knowledge — the "ask Sarah, she knows" dynamic — experience a shift. Some welcome it. Some resist it. Acknowledging this dynamic openly makes the transition smoother.

Measuring success

By the end of 90 days, you should be able to measure:

• Search adoption: What percentage of the organisation is using the platform weekly?
• Time saved: Survey data or anecdotal evidence of reduced search time.
• Agent usage: How many agents have been built? How often are they used?
• Satisfaction: Do people find value in it? Would they go back to the old way?

If those numbers are trending in the right direction, you have a foundation to build on. If not, the 90-day mark is a good time to diagnose and adjust.

Planning an AI implementation? JOURN3Y manages the full process — from readiness assessment through implementation and adoption. Talk to our team.