I kept hearing the same embarrassed confession in user interviews, and it had nothing to do with Liminary.

Researchers would open up their screens and apologise, showing me a workflow that looked completely out of control. Notes were scattered across half a dozen docs, Chrome windows were packed with tabs, and files were stashed randomly in the Downloads folder. They often needed five or six different tools just to walk me through a single project.

A consultant described feeling guilty about her “Frankenstein setup”. She was simultaneously using Notion and Google Docs and Evernote because none of them quite did what she needed. An analyst joked about being a tool-hoarder while rattling off the dozen apps he’d tried and mostly abandoned.

These were smart people who genuinely hated having their information scattered everywhere, yet they kept adopting new tools. And they would say, “There doesn’t seem to be a better way.”

Information was coming at them from so many different sources and in so many different formats that no single tool could handle it all. They would start with one tool for a specific type of note or task, but when it couldn’t meet the needs of another part of the project, they would add a second tool to fill that gap. Before they knew it, their information was spread across multiple places, without a way to see everything together or make sense of it all.

What made this interesting from a product strategy perspective was that most tools are built to do one thing really well, but they all assume they can be the centre of your work if only they just added enough features. But knowledge work, especially the deep kind that involves research and putting ideas together, doesn’t follow neat steps. It doesn’t all come in the same format either. It’s flexible, changing with the project and with how each person likes to work.

So, the tool proliferation problem really becomes a workflow fragmentation problem. Forcing people to split their work across multiple tools breaks the cognitive flow that lets them see connections and generate insights. A consultant told me her best ideas came from noticing patterns across different client projects, but those projects lived in separate folders in separate tools, so surfacing those patterns required her to actively remember and seek out the connections.

That fundamentally changed what we’re building at Liminary. We needed to build something that could work across the contexts where knowledge workers operated, that could pull together information from wherever it lived and surface it when it became relevant. We think of it as an ambient AI, quietly supporting work in the background.

The promise of AI in knowledge work isn’t that it can automate individual tasks. It’s that it can finally bridge the gaps between all the disconnected places where knowledge lives. The companies that figure out how to do that well will have teams that can actually think faster and deeper, not just teams with more tools.

Why knowledge workers spend so much time searching

Across studies, one consistent finding emerges: knowledge workers spend a large chunk of their week just locating the content they need. For example:

• According to a survey conducted by Gartner, 47% of digital workers reported struggling to find the information or data needed to perform their jobs effectively. [1]
  
  
  

• Another survey found that knowledge workers may spend roughly 30% of their workday searching for information, or about 2.5 hours per day. [2]
  
  
  

• One study reported that UK knowledge workers wasted an average of nine hours a week locating information. [3]
  
  
  

Why does this happen? Because knowledge is not neatly organised. It sits in half-a-dozen tools, multiple tab sets, overlapping doc versions, and wandering downloads folders. Every time a worker has to switch context, e.g. bounce from Google Docs to Notion to a file share to a chat thread, they incur a cognitive cost.

Workflow fragmentation is the real productivity drag

It’s tempting to blame the individual tools themselves. But the real issue lies in the fragmentation across tools and contexts.

As one practitioner put it:

“The problem isn’t that workers don’t care about organization — it’s that their workflows evolved faster than their tools.”

When your workflow spans multiple formats (slide decks, spreadsheets, chat threads, PDFs, drafts, email exchanges) any single tool tends to come up short. Each new tool you adopt adds to the set of places you must check, the places where your “knowledge” lives.

That means your knowledge isn’t just in one searchable database; it’s scattered. It’s fragmented. It’s harder to connect the dots.

And when knowledge is fragmented, the ability to see patterns, across projects, across domains, gets damaged. One of the most important sources of insight for knowledge workers is recognizing those cross‐project patterns. But if those projects live in silos, those insights require extra effort to surface.

The cognitive cost of fragmented knowledge and AI over-reliance

Fragmentation doesn’t just waste time; it increases cognitive load. When you’re constantly switching contexts, you’re paying a mental tax. When you rely on multiple tools without integration, you’re creating hidden work. And when you layer on top of that the increasing reliance on AI tools, new risks emerge.

Recent research shows that frequent use of AI tools is correlated with weaker critical thinking skills, largely because of cognitive off-loading. One study with 666 participants found a significant negative correlation between AI tool usage and critical thinking abilities.[4] Another survey of 319 knowledge workers found that greater confidence in generative AI tools predicted less self‐reported critical thinking effort.[5]

Consider these reflections from practitioners:

“I feel like I’m losing my ability to think critically”
“The more I use AI, the less I feel the need to problem‐solve on my own. It’s like I’m losing my ability to think critically”

What this shows is that while AI can streamline certain tasks, it also risks reducing the engagement necessary for deep insight. When all we do is ask a tool for an answer rather than structuring our own frameworks, we gradually give up the connecting thinking that produces insights.

How knowledge management systems still let us down

Traditional knowledge management systems have long struggled with structural limitations: information silos, manual curation bottlenecks, disconnected data repositories, and inefficient search.

And when knowledge sharing remains local and fragmentary (within teams, tools, docs), we get less reuse, less pattern recognition, and more duplication.

Legacy IT and knowledge infrastructures often reinforce these silos: knowledge trapped in departments, locked away in tools, and not integrated into the broader workflow.

This means that even when tools claim to support knowledge management, the workflow reality remains fractured. If knowledge cannot move with the worker across tools, contexts, formats, you still lose the connectivity that powers insight.

Why “best AI tools for researchers” don’t fix the underlying issue

You’ll find dozens of lists of “best AI tools for researchers” or “AI research assistant tools”: summarisation bots, literature-review helpers, citation managers. These are great and have their place. But they address tasks, not the workflow fragment.

Research-oriented AI tools still assume that all your knowledge sits neatly in that one tool. They don’t always handle pull-in from disparate places, or surface insights across projects scattered across locations. A recent study of generative AI in knowledge work found major limitations: user isolation, overreliance, and poor integration with background knowledge. [6]

If you build your stack on “one tool per task,” you may reduce friction in that task, but you don’t reduce fragmentation. In fact, you might add to it by adding yet another silo.

The next wave: ambient AI as a knowledge assistant across contexts

What we need instead is a new kind of system: an ambient AI knowledge assistant, one that sits behind the scenes, knows where your knowledge lives, and surfaces the right information when contextually relevant.

In other words, knowledge management isn’t just about “capture and search.” It’s about “recall and relevance.” It’s about making the invisible connections visible.

Research on “ambient agents” - context-aware systems that quietly orchestrate across multiple platforms and sources - provides a useful framing. These systems are designed to monitor context, integrate background knowledge, and assist proactively.

When we talk about an AI knowledge assistant, this is what we mean: a tool that isn’t another silo, but instead operates above the workflow, stitching together information from wherever it resides, reducing context-switching, and lowering cognitive load.

Practical steps for organising research notes in the AI era

To move toward this ideal, here are some practical steps, grounded in research:

1. Capture with minimal friction: Notes should be easy to add, regardless of format or tool, so you don’t build yet another “downloads folder” tomb.
   
   
   

2. Enable cross‐tool integration: The tools you use must allow pulling in and linking across formats (docs, slides, chat, spreadsheets).
   
   
   

3. Surface connections proactively: Use AI to highlight patterns, similarities, recurring themes or across-project insights you might otherwise miss.
   
   
   

4. Maintain human critical thinking: Because research shows that over-use of AI without oversight can degrade critical thinking. Use AI as assistant, not replacement.
   
   
   

5. Measure time lost to search and switching: Given that many workers spend ~30% of their time searching or switching contexts, track how much time your system saves.
   
   
   

Closing: Teams that think faster, not just use more tools

To return to our opening anecdote: what these high-performing researchers, consultants and analysts were really saying wasn’t that they were lazy or disorganised. They were smart people who were forced into disorganised workflows because their tools didn’t work together. Their knowledge lived everywhere and nowhere at once.

The companies that figure out how to rebuild the knowledge layer - not another tool stack, but a true cross‐context, workflow-driven system - will win. Because the opportunity isn’t a faster note-app, or a better summariser. It’s teams that can think faster, deeper, and more insightfully.

In the end, the promise of AI in knowledge work isn’t simply automation of tasks, it’s the bridging of scattered knowledge, the restoration of cognitive flow, and the empowerment of human insight.

And that’s a future worth building.

If this vision resonates, it’s exactly what we’re building at Liminary: an ambient AI knowledge recall platform designed to work across your existing tools and surface the right insight at the moment you need it. Instead of forcing your work into yet another app, Liminary weaves your notes, sources, and ideas into a unified knowledge layer that stays with you wherever your workflow goes. Our goal is simple: help knowledge workers think faster and deeper by ensuring their best ideas never get lost.

—

Want to explore further?

The research that was curated for this blog post - beyond just the references below - can be explored here. Ask any questions to explore the topic yourself!

—

Sources:

[1] "Gartner Survey Reveals 47% of Digital Workers Struggle to Find the Information Needed to Effectively Perform Their Jobs". Gartner, May 10, 2023. https://www.gartner.com/en/newsroom/press-releases/2023-05-10-gartner-survey-reveals-47-percent-of-digital-workers-struggle-to-find-the-information-needed-to-effectively-perform-their-jobs

[2] "Do workers still waste time searching for information?" Xenit, May 22, 2018. https://xenit.eu/do-workers-still-waste-time-searching-for-information/

[3] "Digital hide-and-seek': Workers are wasting hundreds of hours a year sourcing the information they need to carry out their role" by George Fitzmaurice. IT Pro, March 12, 2025. https://www.itpro.com/business/business-strategy/digital-hide-and-seek-workers-are-wasting-hundreds-of-hours-a-year-sourcing-the-information-they-need-to-carry-out-their-role

[4] "AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking" by Michael Gerlich. Center for Strategic Corporate Foresight and Sustainability, October 14, 2024. https://www.mdpi.com/2075-4698/15/1/6

[5] "The impact of Generative AI on critical thinking: Self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers." Hao-Ping Lee, et al. Microsoft Research, 2025. https://www.microsoft.com/en-us/research/wp-content/uploads/2025/01/lee_2025_ai_critical_thinking_survey.pdf

[6] "Generative AI in knowledge work: Design implications for data navigation and decision-making." Bhada Yun, et al. Conference on Human Factors in Computing Systems. May 1, 2025. https://arxiv.org/abs/2503.18419