The Friction Paradox

Organizations Need Resistance to Build Real Knowledge Systems

The MIT State of AI in Business 2025 report contains a counterintuitive finding, argues Forbes contributer Jason Snyder. This finding challenges conventional wisdom about technology adoption: friction isn’t the enemy of successful AI implementation — it’s the proof of it. Organizations that deliberately design for resistance, that embrace the difficulty of integration, are the ones crossing what the authors call the GenAI Divide. “Pilots that glide frictionless from demo to deployment never build the muscle to scale,” Jason writes, “They collapse the moment they hit real organizational texture, compliance, politics, data quality, and human judgment.” Success comes when enterprises embrace friction.

This insight reveals something fundamental about how knowledge actually develops in organizations. The same friction that makes implementation challenging is what forces adaptation, learning, and the emergence of genuine understanding. When we try to eliminate all resistance, we eliminate the very conditions that enable knowledge to grow.

Friction as a Design Principle

Snyder puts it bluntly: companies fail because they avoid friction. They want smooth demos, easy adoption, seamless integration. But as Snyder notes, friction is “what keeps your tires on the road.” Without it, you have motion but no traction.

This metaphor deserves deeper examination. In organizational knowledge systems, friction manifests as:

The effort required to translate between specialist domains – forcing people to find shared language
The resistance when new tools don’t match existing workflows – revealing how work actually happens
The discomfort of surfacing problems early – creating psychosocial pressure that must be addressed
The time needed for iterative learning – preventing premature optimization

Each of these forms of resistance, when properly channeled, becomes a site of knowledge creation. The question isn’t how to eliminate friction but how to design it productively.

The Shadow Economy as Productive Friction

The report’s discovery of a “shadow AI economy” – where 90% of employees use personal AI tools while official initiatives stall – represents friction at work. This isn’t dysfunction; it’s adaptation. Employees are creating the resistance necessary for real learning by:

Testing tools against actual workflows rather than theoretical use cases
Building personal understanding before organizational commitment
Creating informal feedback loops that official channels can’t provide
Developing contextual knowledge about what works and what doesn’t

This shadow usage creates what I call productive illegibility – spaces where knowledge can emerge without premature formalization. The friction between official and unofficial practices forces organizations to confront the gap between how they think work happens and how it actually occurs.

Dialogic Friction and Knowledge Emergence

Drawing from Anderson’s framework of dialogic organization development, we can understand friction as essential to the conversational processes through which organizations construct knowledge. When different perspectives collide – when marketing’s language meets engineering’s, when frontline experience challenges executive assumptions – the resulting friction isn’t noise to be eliminated but the sound of knowledge being forged.

The MIT report shows that successful organizations create what I’d call calibrated friction through:

Learning loops – The report emphasizes systems that learn from feedback, adapt to context, or improve over time. This isn’t smooth automation but deliberate resistance that forces reflection and adjustment.

Deep customization – Rather than accepting off-the-shelf solutions, successful organizations demand tools that fit their specific workflows. This customization process is friction-intensive but creates genuine integration.

Partnership over purchase – The finding that external partnerships succeed twice as often as internal builds (66% vs 33%) suggests that productive friction comes from negotiating between different organizational cultures and expectations.

The Psychosocial Dimension of Productive Friction

The report identifies “unwillingness to adopt new tools” as the top barrier to AI implementation. But this resistance often stems from legitimate psychosocial concerns that, when addressed, become sources of strength:

Fear of revealing knowledge gaps becomes an opportunity to normalize learning
Concern about job displacement drives conversations about human value and augmentation
Skepticism about AI reliability leads to better governance and validation processes
Resistance to changing workflows reveals tacit knowledge about why current processes exist

Organizations that acknowledge these sources of friction – rather than dismissing them as “change resistance” – create the psychological safety necessary for genuine knowledge sharing. The friction becomes productive when it’s recognized as legitimate rather than problematic.

Beyond Smoothness: Designing for Generative Resistance

The MIT findings challenge the dominant narrative of frictionless digital transformation. Instead of seeking the smoothest path, organizations should ask:

Where is friction revealing important information about our actual workflows?
How can we design resistance that forces deeper learning rather than surface compliance?
What shadow practices are showing us where official systems create unproductive friction?
How do we calibrate friction to be challenging but not paralyzing?

The Learning Organization, Reconsidered

The concept of the learning organization has often been reduced to smooth knowledge transfer and efficient information sharing. But the MIT report and Snyder suggests something different: learning organizations are those that productively engage with friction. They understand that:

Resistance reveals requirements – What seems like obstruction often highlights legitimate needs
Difficulty drives development – Easy adoption often means shallow integration
Conflict creates knowledge – Different perspectives must wrestle to produce new understanding
Slowness enables sustainability – Fast implementation without friction rarely creates lasting change

Practical Implications for Knowledge Leaders

For those of us working to improve organizational knowledge systems, this friction-positive framework suggests several strategies:

Map your friction points – Where is resistance occurring? What is it telling you about underlying knowledge needs?

Distinguish productive from destructive friction – Some resistance enables learning; some simply wastes energy. Learn to tell the difference.

Create “friction budgets” – Acknowledge that meaningful integration requires effort and allocate resources accordingly.

Celebrate productive struggle – When teams wrestle with integration challenges, recognize this as knowledge work, not failure.

Design for evolution, not implementation – Build systems that expect to encounter friction and adapt, rather than systems that assume smooth deployment.

The Paradox Resolved

The friction paradox – that resistance is necessary for progress – only seems contradictory if we mistake efficiency for effectiveness. The MIT report shows that the most successful AI implementations are those that embrace friction as a feature, not a bug.

This has implications for how we think about organizational knowledge. Instead of seeking frictionless information flow, we should design for generative resistance. Instead of eliminating all barriers, we should ask which barriers are actually bearing walls – supporting the structure of organizational learning.

The organizations crossing the GenAI Divide aren’t those that avoided friction. They’re the ones that recognized friction as a teacher, a signal, and ultimately, a tool for building knowledge systems that actually learn.

What friction in your organization might actually be telling you something important about how knowledge really develops?