Our whole approach rests on one claim: automation designed without a human-cognition strategy makes organizations weaker over time. That claim is testable — and tested. Here are the five studies we build on, each in plain language, each linked to the source so you can check us.
MIT Media Lab · 2025
Writing with AI weakens the brain activity behind memory and ownership
Researchers put EEG headsets on people writing essays — some unaided, some with a search engine, some with an AI assistant. The AI group showed the weakest brain connectivity of any group, and minutes after finishing, most could not quote from their own essays. They also reported the lowest sense of ownership over their work.
A team that lets AI do its thinking stops encoding the work. The output exists; the understanding doesn't.
Societies (peer-reviewed) · 2025
Heavier AI use correlates with weaker critical thinking
A study of 666 participants found a significant negative correlation between frequent AI tool use and critical-thinking ability, mediated by cognitive offloading — handing your thinking to the tool. The effect was strongest in younger participants.
The people earliest in their careers — the ones who most need to build judgment — are the most vulnerable to losing it.
J. of the Assn. for Information Systems · 2023
An automated accounting firm quietly lost the ability to do accounting
A longitudinal case study documented an accounting firm whose staff, after years of relying on cognitive automation, could no longer perform core accounting tasks when the system was unavailable. Neither the workers nor their managers noticed the erosion while it was happening — that is what makes the cycle vicious.
Skill erosion is invisible until the day the system fails. By then the people who could have covered for it are gone or out of practice.
Frontiers in Psychology (NTNU) · 2024
Doing things by hand engages the brain networks that build memory
High-density EEG research from the Norwegian University of Science and Technology showed that handwriting engages far broader and more connected brain networks than typing — in the exact frequency bands associated with memory formation and learning. Active, effortful doing is how the brain encodes skill; passive, mechanical input is not.
This is the foundational principle: the work your team does by hand — judgment, review, decisions — is the work that keeps them capable.
CHI · Microsoft Research & Carnegie Mellon · 2025
Knowledge workers report thinking less critically when they trust AI more
A survey of knowledge workers using generative AI found that higher confidence in the AI was associated with less critical thinking, while higher self-confidence was associated with more. Workers reported shifting from doing the work to verifying it — and reported reduced cognitive effort on exactly the tasks they automated.
Verification is a skill too. If your team rubber-stamps the machine, the review gate is theater — which is why our deployments keep review load real and owned.
How we read this research
What we claim — and what we don't.
None of these studies say AI makes people less intelligent, and neither do we. The documented effects are on engagement, memory, critical thinking, and skill retention — the working machinery of a capable team. That's the defensible claim, and for an operating company it's the one that matters: a team can keep every IQ point it has and still lose the ability to run its own processes.
The conclusion we build every deployment on: automate the repetitive, protect the cognitive, and keep a human approval gate in the loop — not as a phase, but as the design.