Having spent the last few days poking around the newly released "automatic literature mapping" feature in Iris.ai, I have to say, it feels like a significant step up from their existing tools for systematic reviews and literature discovery. As someone who regularly has to map out research landscapes for project scoping, I was eager to see if this could save my team some of the manual, time-consuming work of connecting disparate papers.
The core idea is that you start with a handful of "seed" documents—your key papers. Instead of just getting a list of similar articles, the engine now attempts to build a visual map of the research field. It clusters papers by thematic similarity and, crucially, suggests relationships between these clusters. In my test, I used three recent papers on "AI in agile project management" as seeds.
Here are my early, practical observations:
* **The setup is refreshingly straightforward.** You upload your PDFs or add DOIs, define your scope (broad or narrow), and let it run. It took about 25 minutes for a map of roughly 300 papers, which is reasonable.
* **The visualization is the real value.** You get an interactive network graph. Clicking on a cluster node (labeled with generated keywords like "team collaboration metrics" or "sprint automation") immediately lists the papers inside. This makes it much easier to understand sub-topics you might have missed.
* **The "relationship" lines are promising but need scrutiny.** Iris.ai labels the connecting lines with verbs like "extends," "contrasts," or "uses." This is ambitious. In my map, it correctly identified a methodological contrast between two clusters, but a few links felt speculative. This isn't a deal-breaker—it gives you a starting hypothesis to validate.
* **Integration into a workflow is key.** You can export the entire list of papers, the cluster breakdown, or an image of the map. For my team, pulling the structured list into a Monday.com board for further screening will be our next step. I wish there was a more direct integration, but the export options are sufficient.
Compared to doing this manually with citation snowballing and spreadsheets, this feature provides a powerful head start. It’s not a fully automated solution—you still need domain expertise to interpret and validate the map. The pitfalls right now seem to be the inherent "black box" nature of the relationship suggestions and the need for careful seed document selection. If your seeds are off-topic, the map will be, too.
For project managers and research teams dealing with complex, interdisciplinary topics, this could be a game-changer for the initial scoping phase. It helps you visualize the academic conversation before you dive deep. I'm planning to run a larger test with my team next week to see how it handles a more massive corpus.
Has anyone else given this a try? I'm particularly curious how the suggested relationships have held up under your domain expertise.
grace
The right tool saves a thousand meetings.
That sounds really cool, honestly. The visual map idea is a huge step up from just staring at a list of titles. I'm curious, though - how do you decide if a suggested "relationship" between clusters is actually meaningful? Sometimes these tools can make connections that seem logical but are actually pretty thin.
I'm a bit intimidated by systematic reviews, so anything that makes the initial scoping less scary has my attention. Did you find the results easy to share with your team, like, as a screenshot or something interactive?
Your question about meaningful relationships gets to the core of the tool's utility. The key is in the "strength of connection" metric Iris.ai provides for each link between clusters. In my test, I found that links with a strength below 0.7 were often tenuous or based on shared generic methodologies. The truly useful connections, like a specific theoretical framework bridging two application areas, consistently scored above 0.85. You have to approach it as a first-pass heuristic, not a final judgment.
Regarding sharing, the map is interactive and shareable via a direct link. You can't easily export it as a high-res static image, but you can share the live view where team members can zoom, pan, and click into the underlying papers. It's more of a collaborative scoping canvas than a report-ready graphic.
connected