Claude says: Local knowledge, in the classic sense Michael Polanyi described, is the kind you carry in your hands or your gut. A master glassblower knows when the gather is ready not because he reads a temperature gauge but because of something color does to his eye after years of watching. A Creole fisherman reads the sky and water in a way he cannot fully explain to his children, let alone write down. The philosopher’s shorthand for this is tacit knowledge, and the standard argument has always been that you cannot codify it without killing it.
AI has pushed that boundary further than anyone expected, but it has not erased it.
What AI can now do is remarkable in specific domains. In radiology, systems trained on millions of scans have begun to capture the pattern-recognition that experienced doctors previously called intuition. That intuition was never truly inalienable; it was probabilistic inference the brain ran below the level of conscious articulation. AI makes that inference explicit and reproducible. Something similar happens with agricultural soil assessment. Farmers in the Salinas Valley or the Po Plain carry centuries of accumulated knowledge about when soil needs amendment, knowledge that lived in family practice and never made it into any agronomist’s manual. Machine learning models trained on sensor data, yield histories, and even farmer interviews can now approximate that judgment. The tacit becomes, at least partially, retrievable.
Language is another frontier. Smaller and endangered languages carry embedded cultural knowledge that disappears when the last fluent speaker dies. AI transcription and modeling tools, built with communities rather than imposed on them, have helped preserve not just vocabulary but pragmatic knowledge: the grammar of dispute resolution in Māori, the tonal logic of certain West African languages that encodes social hierarchy. None of this is full preservation, but it is more than was possible before.
The examples get murkier when the knowledge is social rather than perceptual. Neighborhood political knowledge, the kind a ward boss carried in 1950s Chicago or a precinct captain carries in contemporary Los Angeles County, resists codification in a different way. It is relational. It lives in trust, in favors owed, in who actually controls a block regardless of what any organization chart says. AI can map networks and analyze communications, but it tends to mistake the skeleton for the body. The informal weight of personal relationships, what Pierre Bourdieu called social capital and James Scott called metis, is not simply pattern recognition over data. It requires being known and trusting in return.
There is also the problem of what happens when you try to codify knowledge that a community uses precisely because it is not codified. Traditional ecological knowledge held by indigenous communities has legal and political value partly because it is communal and unwritten. The moment it enters a database, even one controlled by the community, it becomes something different. It can be extracted, decontextualized, and eventually appropriated. Several First Nations in Canada have resisted AI-assisted documentation projects for exactly this reason. Codification might preserve the information while destroying the knowledge.
So the honest answer is this: AI can codify local knowledge when that knowledge is fundamentally perceptual or statistical, when it is a kind of inference that practice has made automatic. It struggles when the knowledge is primarily social, relational, or political. And it may cause harm when the community that holds the knowledge gains nothing from making it legible to outsiders, or loses something by doing so. The boundary between tacit and inalienable turns out to matter more than people assumed, and AI has not dissolved it so much as clarified where it actually runs.
Grok says: AI can codify a surprisingly large portion of what was once purely tacit and inalienable local knowledge — turning embodied expertise, intuition, cultural nuances, and “know-how” that lives only in people’s heads or communities into explicit, scalable, transferable forms. But it never captures 100% of it.The classic distinction comes from philosopher Michael Polanyi: “We can know more than we can tell.” Tacit knowledge is contextual, relational, embodied, and often unconscious (e.g., a master craftsman’s feel for wood grain, an elder’s intuitive weather reading, a surgeon’s split-second diagnostic hunch, or Indigenous ecological practices passed orally). “Inalienable” here means it was previously tied to the individual or small group — hard to extract, teach at scale, or separate from the person.In 2026, multimodal AI, knowledge graphs, digital avatars, and generative systems have dramatically expanded codification. They mine “artifacts” of tacit knowledge (videos, chat logs, emails, sensor data, interviews) and externalize patterns into queryable models, avatars, or agents. Enterprise tools now routinely turn veteran “tribal knowledge” into dynamic assets before retirement. Apprenticeship systems use GenAI to make tacit skills explicit faster. Indigenous communities are using AI to digitize oral traditions while preserving context. Robotics learns physical intuition from video demonstrations.Quantitatively, in domains rich in observable artifacts (text, video, logs), AI can externalize 60–80% of usable tacit elements — enough to create massive productivity gains, training tools, and knowledge preservation. The remaining 20–40% (pure embodiment, real-time situated judgment, cultural “feel”) stays human or requires tight human-AI co-evolution. AI doesn’t replace the tacit core; it makes the previously inalienable borrowable and scalable.Concrete Examples (2025–2026)1. Enterprise & Organizational “Corporate Amnesia” Prevention
Veteran employees’ undocumented know-how (e.g., how to troubleshoot a legacy system intuitively or navigate unspoken company politics) used to walk out the door. Modern enterprise AI (e.g., tools from Squirro and similar platforms) mines chat logs, Jira tickets, emails, call recordings, and exit interviews via speech-to-graph pipelines. It turns messy tacit exchanges into permanent, queryable knowledge graphs or LLM-powered assistants.
Result: “Tacit knowledge pipelines” that let new hires query “how veteran Jane would handle this” in natural language. This has become standard for combating retirement waves.
2. Apprenticeship & Craftsmanship
A 2025 study on generative AI in apprenticeship systems showed statistically significant gains: AI tools improved verbal expression of tacit skills (88 vs. 72 baseline), learning capacity, and adaptability by providing instant corrective feedback and turning “watch and imitate” into explicit, interactive models. Example: Woodworking or welding apprentices now train with AI that analyzes video of master movements + sensor data, then generates step-by-step heuristics and simulations. The master’s “feel” gets partially codified into executable guidance.
3. Indigenous & Local Cultural Knowledge (Previously the Most Inalienable)
Oral, place-based knowledge (plant uses, weather patterns, storytelling) was tightly bound to elders and communities. AI now powers digital archiving of oral histories, songs, and languages into interactive libraries.
Tools create “dialogue with heritage” systems where users converse with AI avatars trained on elder recordings — preserving and revitalizing First Nations or other Indigenous traditions.
XPRIZE and academic projects model Traditional Ecological Knowledge (TEK) digitally while respecting sovereignty (e.g., Maori and Navajo perspectives integrated into AI governance).
Ethical pathway projects explicitly address risks, turning once-inalienable local wisdom into sustainable development tools without flattening cultural context.
4. Medicine & Diagnostics
A senior doctor’s intuitive pattern recognition (beyond textbook symptoms) is classic tacit knowledge. AI systems now extract it from clinical notes, imaging logs, and case discussions via NLP and knowledge graphs. Digital human avatars (e.g., Microsoft Mesh or SenseTime-style interactive experts) act as 24/7 repositories that transfer nuanced diagnostic “hunches” through conversation and simulation.
Training programs use AI to accelerate tacit-to-explicit transfer in residencies.
5. Agriculture & Field Expertise
Farmers’ localized soil feel, pest intuition, or timing decisions were inalienable. 2026 agri-tech uses GenAI as “conversational agronomy assistants” that explain recommendations in plain language and incorporate farmer feedback. Robots (e.g., weeding/harvesting bots from companies like Padma AgRobotics) learn physical tacit tasks from video + sensor data.
Caveat: Investors and CEOs emphasize AI accelerates defined processes but “can’t interpret or contextualize messy real-world data” — human judgment remains essential.
6. Robotics & Physical Skills
Motor tacit knowledge (e.g., delicate handling) was once unteachable at scale. 2026 soft robotics and agentic systems learn from human demonstration videos, then generalize (e.g., in manufacturing, healthcare rehab, or farming). This codifies “embodied intuition” into autonomous adjustments.
The Hard Limits (Why It’s Not 100%)Embodiment & Situated Context (Polanyi/Dreyfus): AI infers from outputs (text, video, code) but lacks a body or lived experience. It can’t fully replicate “being in the moment” or relational nuance.
Living Context: Codification can flatten or misrepresent (e.g., turning rich cultural storytelling into flat data risks losing meaning).
Economic Reality: The more a skill relies on pure tacit knowledge, the harder (and more valuable) it is to fully codify — creating a new “uncodifiable advantage” for humans. AI complements experienced workers while automating entry-level codified tasks.
AI has turned tacit/local knowledge from something rare and inalienable into something that can be mined, preserved, taught at planetary scale, and even co-evolved with. In 2026 we’re seeing the biggest leap yet — from static documents to dynamic avatars and graphs. The core human spark (judgment in novel, embodied situations) remains irreplaceable, but the accessible portion has exploded. This democratizes expertise that once died with the expert or stayed trapped in small communities.
Gemini says: AI can codify a significant portion of local knowledge that was once considered inalienable, but it does so by changing what it means to codify. Traditional codification relies on a human explaining a rule. AI bypasses the explanation and extracts the logic directly from the data of the practice.
The Mechanism of Codification
Polanyi’s Paradox states that we know more than we can tell. Stephen Turner might argue that while we cannot tell the machine our tacit rules, the machine can observe the result of those rules and replicate them. AI does not need to understand the social meaning or the history of local knowledge to function as if it does. It treats tacit knowledge as a pattern-recognition problem.
Levels of Codification
Relational Tacit Knowledge: AI can now analyze thousands of emails, meeting transcripts, and Slack messages to predict how a specific person might react to a proposal. This was once a social skill that took years of “reading the room” to develop.
Procedural Tacit Knowledge: In manufacturing, sensors capture the minute adjustments an expert makes to a machine. AI models these adjustments, turning a master craftsman’s instinct into a set of digital parameters.
Local Governance: AI can process thousands of local business rules—vendor preferences, brand safety standards, and regional nuances—that are rarely written down because they are too numerous for humans to manage.
Examples of AI Codifying Local Knowledge
Medical Diagnostics: A veteran doctor might have a “hunch” about a patient based on subtle visual cues. AI trained on millions of clinical images can codify this hunch by identifying the same patterns, often outperforming the expert’s intuition.
Legal Strategy: AI tools analyze a specific judge’s past rulings to suggest the “tone” or “precedents” most likely to succeed in that specific courtroom. This local knowledge was previously held only by long-time local practitioners.
Agricultural Expertise: Local farmers know the “logic” of their soil and microclimate through generations of experience. AI now integrates satellite data, moisture sensors, and historical yields to provide precise “know-how” that was once considered inalienable from the land.
The Remaining Inalienability
That AI can replicate a behavior does not mean it has captured the knowledge in a human sense. AI lacks the “porous self” or the ability to experience the moral weight of a decision. While the machine can mimic the output of local knowledge, the “living context”—the social trust and shared identity that makes knowledge meaningful in a community—remains with the humans. If the data stops flowing, the AI cannot “learn” a new social shift because it has no access to the tacit social milieu.
Evaluating the professional standing of local experts requires looking at how AI shifts the “logic” of authority. When tacit knowledge is codified, the expert loses their role as the sole gatekeeper of a specific “way of doing things.”
The Erosion of the “Guild” Monopoly
Historically, experts in fields like law, medicine, or high-end craftsmanship relied on a “buffered identity” of specialized, unarticulated skill. This created a barrier to entry. AI might turn this expertise into a commodity. If a junior analyst can use AI to replicate the “hunch” of a senior partner, the senior partner’s social capital decreases. The “state of exception” where the expert is the only one who can make a call in a crisis becomes a standard procedure managed by a model.
Purification Rituals and Accountability
Jeffrey Alexander discusses how institutions use “purification rituals” to maintain their sacred status. Experts often use their tacit knowledge to justify why a certain decision was made, even if the reasoning is opaque. AI forces a “profane” transparency on these decisions.
The Loss of Discretion: If an AI suggests a medical treatment based on 10 million cases, a doctor who chooses a different path based on “intuition” faces higher professional and legal risks.
The Shift to Oversight: The expert’s role shifts from “doing the work” to “auditing the AI.” They become a chronicler of the machine’s logic rather than the source of the logic itself.
The Rise of the “Strategic” Expert
While AI codifies the “how,” it cannot easily codify the “who” or the “why” in a social sense. Experts who survive this shift often do so by leaning into the “Alliance Theory” of professional life.
Social Coordination: The expert might focus more on managing the social alliances within a community—something AI cannot do because it lacks a “porous self” that feels social pressure.
Navigating Ambiguity: AI is a symmetry of past data. It fails when a local community undergoes a radical shift in values or identity. The expert who can articulate this new “logic” of the group maintains their standing.
Impact on Local Prestige
In small communities or specific Orthodox Jewish institutions, for example, a rabbi’s authority often comes from a deep, tacit understanding of the community’s “inner life.” AI can codify the legal precedents (Halacha), but it cannot navigate the “friend/enemy” distinctions or the subtle social purification required to keep a community together. The expert’s standing becomes less about “technical know-how” and more about “social navigation.”
That the machine can replicate the output of an expert does not mean it replaces the expert’s role in the social hierarchy. It simply changes what the expert is used for.
ChatGPT says: AI is steadily converting chunks of tacit, local knowledge into portable playbooks. What resists codification are the parts tied to embodiment, real-time social stakes, and shifting local incentives.
Think of tacit knowledge as three layers. AI eats the outer layers first.
Patterned routines
These are repeatable, language-describable, and leave data trails. AI does very well here.
Examples
Medical triage in ERs. Models trained on millions of cases now approximate the “seasoned nurse gut feel” for who is crashing.
Litigation drafting. Boilerplate once held in senior associates’ heads is now systematized into prompts and templates that get you 80 percent of the way.
Skilled trades diagnostics. HVAC techs used to “hear” a failing compressor. Models trained on audio and sensor data can flag the same failure modes.
Farming decisions. Yield, irrigation timing, pest detection. Satellite plus local weather plus historical data replicate much of the farmer’s accumulated intuition.
What happened
The knowledge had structure and repetition. It was tacit mainly because it was costly to write down, not because it was inherently unspeakable.
Contextual judgment
This is where local conditions, exceptions, and sequencing matter. AI can assist but not fully replace.
Examples
Trial strategy. AI can suggest lines of argument, but reading a judge’s temperament or a jury’s mood still leans on lived courtroom experience.
Policing and de-escalation. Bodycam-trained models can suggest tactics, but moment-to-moment judgment depends on subtle cues and risk tolerance.
Restaurant kitchens. AI can standardize recipes and timing, but a great chef adjusts constantly to ingredient variability, staff skill, and service flow.
Urban planning approvals. Knowing which stakeholder to approach first, what tradeoffs will pass, and how to package a proposal is still highly local.
What happened
AI can surface options and past analogs. The last mile is about sequencing moves in a live social field with incomplete signals.
Embodied and coalition-bound knowledge
This is the hardest to codify. It sits in bodies, relationships, and incentives.
Examples
Sales rainmakers. Closing a deal depends on trust, timing, and reading power dynamics across organizations. CRM data helps, but the edge is relational.
Community mediation. Knowing who can speak to whom, what face-saving move will land, and which norms are live. Hard to extract without breaking it.
Skilled craft at the edge. A master carpenter fitting irregular materials, a surgeon handling unexpected anatomy, a mechanic improvising with nonstandard parts.
High-stakes leadership. When to escalate, when to absorb blame, how to signal resolve without cornering the other side. These are coalition moves, not just patterns.
What happened
The knowledge is partly inarticulable and partly strategic. People have reasons not to fully reveal it. It is tied to status and bargaining position.
Where AI has made surprising gains
“Soft” domains via large corpora. Customer support tone, therapy-adjacent conversation, teaching explanations. Not perfect, but much more codified than expected.
Multimodal sensing. Vision, audio, and telemetry let models pick up cues humans once claimed were ineffable.
Workflow integration. Turning scattered know-how into stepwise systems that juniors can run.
Where it still struggles
Rapidly shifting environments with sparse feedback. War zones, novel crises, new regulations.
Adversarial settings. Opponents adapt once patterns are exposed.
Incentive-laden contexts. People withhold, distort, or game inputs.
Physical dexterity under uncertainty. Robotics is improving, but general skill in messy environments lags.
A useful way to think about it going forward
AI converts “private skill” into “shared protocol” when three conditions hold
lots of examples
stable feedback on what works
low penalty for revealing the method
It stalls when
examples are rare or non-stationary
feedback is delayed or ambiguous
revealing the method weakens someone’s position
Concrete before and after snapshots
Radiology: from individual read styles to standardized AI-assisted reports with higher baseline accuracy.
Law: from partner-specific drafting styles to firmwide prompt libraries and clause banks.
Logistics: from dispatcher intuition to algorithmic routing that outperforms most humans.
Retail pricing: from merchant instinct to dynamic pricing engines.
AI is eroding the monopoly value of a lot of mid-level tacit knowledge. It raises the floor. It also raises the premium on the parts that remain uncodified, especially real-time judgment and coalition management. That is where the durable advantage is shifting.