AI isn’t going to fix your intranet, it’s going to expose it
AI doesn’t improve messy information, it reveals it. Tools like Copilot don’t “understand” content the way people do because they retrieve, summarise, and recombine what already exists across your environment. Remember, AI will always give an answer, even when the underlying information is unclear.
Which means:
- Good content becomes powerful
- Confusing content becomes inconsistent answers
- Outdated content becomes confidently wrong answers
Takeaway: AI performance is not a tool problem. It’s a content quality problem.
Your intranet was built for humans, not machines
Most intranets have grown organically over time:
- Multiple content owners
- Inconsistent naming conventions
- Duplicated or overlapping pages
- Pages that try to do too many things
Humans cope with this because we interpret context instinctively. We fill in the gaps, but AI doesn’t and it relies on signals. When those signals are inconsistent, the outputs are too so make sure you consider:
- Titles
- Headings
- Metadata
- Structure
- Freshness
- Ownership
Takeaway: AI doesn’t read between the lines. It reads what you explicitly structure.
The shift: your intranet is becoming a knowledge base
This is the mindset shift that matters most. Your intranet is no longer just a communications channel or a document library because it is becoming a knowledge base for AI and that changes how content should be written. When those signals are inconsistent, the outputs are too. Instead of broad pages trying to cover everything, each piece of content should have:
- One clear purpose
- One primary audience
- One job to do
For example:
- ❌ “General Information”
- ✅ “Eligibility Criteria for Professional Staff Leave”
Or:
- ❌ “Requests should be timely”
- ✅ “Staff must submit leave requests at least 10 business days in advance”
Takeaway: If a human has to interpret it, AI will likely misinterpret it.
Three small structural changes that matter most
You do not need a rebuild. You need consistency in a few key areas.
1. Metadata that actually means something. Keep it simple and consistent as this helps both people and AI understand relevance and trustworthiness.
- Clear title
- Short description
- Content owner
- Last reviewed date
2. Content lifecycle discipline: AI doesn’t know what’s “out of date”. It only knows what it can access so if old content is still live, it will still surface it. Start to:
- Archive redundant content
- Remove duplicates
- Introduce review cycles for critical pages
3. Clear structure over volume:
More content does not equal better outcomes, better structured content does. So think:
- One page per topic
- One answer per question
- No “everything pages”
Takeaway: Structure is what makes content usable by AI at scale.
Where to start (without overwhelming your team)
You don’t need to fix everything; start where it matters most:
- High-traffic pages
- Frequently referenced policies
- Content that causes confusion or repeated questions
Then introduce a simple standard:
- How a page is structured
- What metadata is required
- When content must be reviewed
Support content owners rather than centralising everything. This is iterative improvement, not a rebuild project.
Takeaway: Focus effort where AI (and people) already depend on your content.
What “good” looks like in an AI-ready intranet
When your intranet is in good shape, you’ll notice:
- Copilot answers feel more accurate and consistent
- Fewer “which version is correct?” conversations
- Less duplicated content creation
- More trust in surfaced information
- Faster decision-making because answers are clearer
AI doesn’t create these outcomes, it amplifies them. AI won’t make your intranet better. It will make its quality unavoidable. The organisations that benefit most won’t be the ones with the biggest transformation budgets. They’ll be the ones who focused on fundamentals:
- Clear content
- Simple structure
- Reliable governance
- Content people can trust
And the good news is none of that requires starting from scratch it just requires starting.
Final thoughts
If you’re thinking about, or in the process of implementing Copilot or agents in your intranet, it’s worth understanding what it will surface in your current environment. We’re working with organisations to assess intranet and digital workplace readiness for AI, identifying where structure, governance and content need to evolve to support accurate and trustworthy outcomes. If that’s something you’re exploring, I’d be happy to have a conversation on LinkedIn or via email.
About the author
Linda Brunetti is a Senior Consultant within Engage Squared’s Digital Workplace practice. She believes people are at the heart of every smart digital workplace and loves bringing a company’s culture to life via intelligent platforms that are designed to make peoples lives better. Linda is a Mum of 2 kids and a lover of the beach, so often spends her summer weekends packing the boot and vacuuming sand out of her car.
