That's how Jamie Lyon, Chief Product and Strategy Officer at Lucid Software, opened last week's webinar, where we revealed the new airfocus: The product intelligence platform for teams who need direction, not just speed, in the AI era.
airfocus by Lucid’s Head of Product and co-founder, Malte Scholz, and Product Manager, Spencer Cowley, showed what this update looks like in practice.
Here are your top questions, answered.
can support your team.
Book a demo
Is the Insights agent just auto-tagging, or is it doing something more?
Auto-tagging is part of it, but that’s just the baseline.
Yes, the Insights agent automatically handles field-filling as feedback comes in, so your team doesn't have to manually tag before they can get to the actual signal. But there’s more to it. What it's really doing is connecting feedback to meaning.
As Spencer explains: "As feedback flows into our workspace, the Insights agent will automatically triage that and identify relevant pieces from that feedback that can be linked to different product opportunities that may already be in development or on our backlog, so that we can start to transform that qualitative feedback into real signal that we can prioritize with."
This distinction matters. Auto-tagging organizes data. The Insights agent builds a case. Each piece of feedback that comes in gets extracted, interpreted, and tied to existing opportunities, so when you're making a prioritization call, you're not looking at raw submissions, but an aggregated insights score sitting alongside other signals. The picture gets clearer the more feedback flows through.
The efficiency gain is real too, and Malte has seen it in his own team:
But the bigger shift isn't the time saved. It's that the output is actually useful for decisions, not just filed away.
What makes the airfocus agent different from just asking Claude or ChatGPT about my product?
This is probably the question we get most, and it's a fair one.
You can absolutely use Claude or ChatGPT for product work. But when you ask them something like "Why did we prioritize this feature?" or "What feedback is driving this initiative?", they don't know the answer. They're starting from zero every time.
As Malte puts it: "You have all these things that connect to each other and that are moving in real time. You have the opportunity that's connected to a bunch of feedback, the opportunity is a child of a larger strategic initiative, which is connected to key results and strategy documents. You have a lot of context, and relationships between all these objects. That context is incredibly important for the AI and the humans to constantly monitor and act upon."
That's what the airfocus agent works with: a data model where those relationships already exist. When you ask a question, it's navigating structure – your structure, your data – not reconstructing it from scratch.
Spencer adds the practical angle: "If you rely on other tools to reconstruct all these connections between the data and have it do it live, it fills up the context window with things that might not be related, and you also can't validate and point the AI and agents in the right direction." airfocus surfaces the signal that's actually relevant, not everything at once.
How far back can the airfocus agent read data?
There's no hard limit. The agent has access to the full item activity history in your airfocus instance. Spencer explains: “It starts with the most recent data and after that, it can pull more and more from the past as it needs.”
The older the data, the more valuable and extensive context you can provide to the agent: decisions made six months ago, feedback that shaped a roadmap, priorities that shifted with strategy.
How quickly does the intelligence layer start to feel valuable?
It depends on how you're using it, but value shows up earlier than most people expect.
If you're using airfocus for feedback management, the Insights agent gets to work immediately: automatically triaging incoming feedback, linking it to relevant opportunities, and reducing the manual effort of sorting and tagging.
Then, the compounding effect comes with breadth. The more of your product work lives in airfocus – OKRs, strategy documents, roadmaps, decision history – the richer the context, and the more useful the AI becomes.
What can external AI tools access through the MCP server, and how current is the data?
The airfocus MCP server gives external tools (Claude, Copilot, ChatGPT, or any MCP-compatible tool) access to your workspaces, items, comments, insights, item activity, and semantic search across your data.
But more than just reading existing items, it can also create new ones. Spencer demoed this live, using Claude via the MCP server to identify a product gap, create a new initiative, and tag Malte with the reasoning – all without leaving Claude. “You can continue this train of thought and work from whatever tool you might be in, all through the airfocus MCP server," he says.
Regarding for currency: The data is live. Every time the LLM calls a tool, it pulls the latest version from your airfocus instance. There's no sync delay.
Can I tell which insights were created by the AI versus a human?
Yes, and the team was deliberate about this. Spencer explains, "There’s a way that you can tell on each individual insight. There's a sparkle icon; you hover over it, and it says this was created by the Insights agent. We are also being very explicit in the work that we're doing to label where that source came from. This is part of us wanting to be transparent and helping teams to understand: This change was made on this item, and this is who did it."
The full activity log tracks every change – human or AI – including which tool triggered it. And if you want to revert something the AI created, you can.
Will users be able to create their own AI skills or standard operating procedures?
Not yet, but it's actively in development. Spencer described the vision: "You can not only have skills for you as a user, but also at a workspace or team level, because often in these companies, one or two product managers may find a certain skill really valuable and wish to scale it to the rest of the org."
He called it moving from individual AI to institutional AI, and it's something both airfocus and Lucid are building toward right now.
How long before product intelligence becomes mainstream, as DevOps did?
DevOps took roughly a decade. The team's take: This time it’s different.
Malte pointed to a telling signal: "Even in large, more traditional Fortune 500 companies – sometimes even the more restrictive product orgs in banking or insurance – these people very often use Claude Code and a lot of these tools. If you take that as a proxy for the speed of this movement, it's a very strong sign that things are moving much faster this time."
Adoption won't be uniform: Some teams are already operating this way, others are earlier in the curve. But as Jamie Lyon framed it in the webinar opening:
The Insights agent, airfocus agent, and MCP are now live. Explore them here.
Francisca Berger Cabral
Read also
Explore how airfocus can support your team
