5th November 2026
1. Academic Concepts
Explicit and tacit knowledge management and preservation of traditional knowledge.
Artificial Intelligence, Generative AI, Retrieval-Augmented Generation (RAG), and knowledge graphs.
AI hallucinations, knowledge validation, and human–AI collaboration. Intellectual property, data ownership, cultural heritage, and ethical AI governance.
2. Learning Objectives
To understand the differences between explicit and tacit knowledge in traditional businesses.
To examine how AI technologies can capture, organize, preserve, and retrieve traditional knowledge.
To evaluate the limitations of AI, including hallucinations and its inability to fully replace human expertise.
To analyse ethical, legal, and strategic issues related to knowledge ownership, commercialization, and community participation.
3. Learning Outcomes
Students will be able to distinguish between explicit and tacit knowledge using real-world examples.
Students will be able to propose an AI-based knowledge preservation system using RAG and knowledge graphs.
Students will be able to assess AI-generated information for accuracy, reliability, and expert validation.
Students will be able to recommend an ethical and sustainable strategy for preserving traditional knowledge through AI.
4. Academic Relevance
PGDM, MCA, BCA, BBA courses
The Knowledge That Was Not Written Down
Raghavendra Acharya, 62, would open the wooden doors of his family’s workshop in Channapatna, Karnataka, each morning at 5:30 a.m. The workshop had been making traditional wooden toys and lacquerware for almost six decades. The workshop size was small. There were four artisans: Raghavendra, his younger brother and two apprentices.
Visitors were inclined to the belief that the brightly colored wooden toys were the product of a simple process: cut the wood, turn it on a lathe, color it, polish it and pack it. Raghavendra knew that this description was technically accurate, but practically useless
The real knowledge was in the details. He could look at a piece of wood and tell if it would crack as it turned. He could smell the right kind of wood, and weigh it. He knew that some colors behaved differently in different humidity. He knew how fast the lathe should run by the sound.
- He had learned these things from his
- His father had learned them from his
- Very little of it was
“Appa, why don’t you write all this down?” his 28-year-old son Arjun, a software engineer working in Bengaluru, asked him one afternoon. Someone else will have to learn it one day anyway.” Raghavendra laughed. How do I write down the sound of wood about to break?” Arjun had no ready answer. He remembered that conversation.
A vanishing business
The workshop faced a number of problems simultaneously. Raw material prices have gone up. The young people of the village were moving to the cities. The machine-made decorative goods cost less. Customers increasingly expected online ordering, personalized designs and faster delivery. Most importantly, the workshop was heavily dependent on Raghavendra.
With him gone, production slowed to a snail’s pace. If he retired, nobody knew if the workshop could go on producing the same quality. Nothing strange. Thousands of traditional businesses across India relied on tacit knowledge gained over decades.
Some examples are listed below:
- Handloom
- Traditional pottery
- Casting metals
- Wood carving
- Ayurvedic preparations practices
- Processing of traditional foods
- Making musical instruments
- Making jewellery
- Stone carving
- Traditional farming practices
It wasn’t just the fact that these businesses were losing customers. They were losing information. And once that knowledge was lost, it might not be possible to recover.
The Problem of Unwritten Knowledge
Arjun began to study the problem. He discovered that organizational knowledge can be grouped into two general categories.
Explicit knowledge: This can be recorded fairly easily. For instance:
- “Heat the material to 180°C.”
- “Drill with 0.2 mm drill
- “Apply two coats of the
This type of knowledge might be kept in manuals, spreadsheets, photographs, or videos.
Implicit knowledge This was a lot harder. For instance:
- “This wood’s a little too
- “The color is right but it doesn’t look
- “Take the material “
- “Listen to the sound of the
The decisions came from years of experience. Raghavendra felt he learned more than 40 years of practical knowledge, but never systematically put it down. Arjun began to interview him and to tape their conversations on his cell phone.
The first recording was a shocker. For nearly two hours Raghavendra spoke without a pause. He spoke about
- Wood choice
- Weather impact on production
- How different tools operated
- How defects were found by artisans
- How customers expressed preferences in non-technical terms
- Mistakes that are common and fixing them
- When to discard a product
- How to train apprentices
But when Arjun heard the recording later, he realized that it was not enough to only record conversations. Raghavendra used to say things like:
- “It’s about getting the
- “I can’t say “
- “Do it for ten years, you’ll “
The information was there but it was tough for another person to use.
The AI concept
Arjun worked for a tech company and had some experience with generative AI. He made a proposal. What if they built a digital repository of knowledge for the workshop?
They might congregate:
- Interview(s)
- Pictures
- Video
- Audio recordings
- Product catalogues
- Old notebooks
- Customer conversations (convos)
- Process papers
Then an AI system could help structure the information and let apprentices ask questions. For instance:
Newbie:
How can you tell if this wood is good for turning? The system might respond:
Look to the grain direction, moisture condition and surface reaction as you start to turn as shown by Raghavendra in his recordings. See Visual Indicators in Demonstration 17. The system could also bring up the relevant part of the video.
Arjun had the idea of:
“CraftMemory” But there was a hitch.
Would AI really be able to preserve knowledge or would it be a very convincing clone of knowledge?
Memory Building Craft
Arjun pitched the idea to his father and the other artisans. The first reaction was mixed. Raghavendra was doubtful. “AI has never been given a piece of wood. How will it know. His brother was more outspoken. “Why would we instruct a computer in all our knowledge? The apprentices were enthusiastic though. One apprentice, Meena, said: “Sometimes I don’t want to disturb uncle when he is working. “If only I could ask the system first, it would help.
Arjun suggested a pilot project.
Step 1: Knowledge Acquisition
The team would film Raghavendra at work doing important work. Each session would involve:
- Video
- Sound
- Picture
- Oral explanations
- The environment conditions
- Tools used in the process
- Materials
- Result
Step 2: Transcribe the audio
Speech would be transformed into text.
Step 3: Knowledge mining
AI would recognize:
- Procedures
- Materials
- Tools
- Choices
- Exception
- Common mistakes
- Relationships between concepts
Step 4: Knowledge graph
The information would be linked. For instance:
Wood type → moisture → speed → tool → surface finish
This would enable the system to correlate seemingly unrelated information.
Step 5: Retrieval-Augmented Generation
Instead of using a general-purpose AI model to answer questions based on its own training data, CraftMemory would pull answers from the verified knowledge repository of the workshop. Then the AI would compose the answer.
Step 6: Validation by content expert
Raghavendra or some other experienced artisan was to verify every bit of important information.
The system would tag information as:
- Checked
- Review Needed
- Conventional practice
- Experimental (not within the scope of this work)
- Anonymous
The pilot was for a 3 month. The results were promising. Apprentices could access information way faster. Instead of having someone explain a 30-minute explanation, a new worker can watch a five-minute demonstration. The workshop also noted that some practices had not been formally recognized.
Like, Raghavendra had built a sort of informal process to spot bad wood before production starts. It was recorded with the help of the AI system. But new questions came up.
When Replication Is Preservation
One evening Arjun got an email from HeritageCraft Global, a big international lifestyle company. The company came across CraftMemory through a social media post. Its executive wrote: “We are interested in partnering with you to scale traditional craftsmanship globally.”
It was an attractive offer, proposed by HeritageCraft:
- Investment of ₹50 lakh
- Digital infrastructure
- Worldwide marketing
- E-Commerce distribution
- Training facilities
- International exposure
But there was one condition. HeritageCraft needed access to the CraftMemory knowledge system. Its purpose was to educate artisans in different places. At first this looked positive. The old craft could go world-wide. But Raghavendra had a concern. “If everybody can learn it from the computer, what happens to the people who spent forty years learning it?” The question was more complicated.
Now imagine that HeritageCraft used that knowledge to train 500 new artisans. Would this have kept the tradition alive? Or had the knowledge been commercialized and detached from the community from where it originated?
Ownership question
Who was the knowledge owner?? Potential answers are:
- Raghavendra as he had acquired the
- His family, as the knowledge had been passed down through the
- The community of artisans since many of the practices were developed
- The commercial use of the knowledge, which makes the
- The public, since traditional knowledge was part of cultural
- Or maybe it could not be owned exclusively by
Another problem is
- Some of the knowledge wasn’t all
- Raghavendra knew the traditional customers who would order specific
- He knew which artisans were considered experts in particular
- He was aware of the stories behind some
- He knew family-specific
- Should all of this stay in the AI system?
The Problem of Hallucination
Six months into the project, the system hit a serious snag.
“If the surface gets little cracks when you are polishing, what do I do?” asked an apprentice. CraftMemory replied with a detailed procedure. The apprentice followed suit. The product was dissected. Raghavendra looked at the piece. He knew it was coming, and it came fast. This was the first time he had taught the procedure.
The AI had confused information from two different recordings and produced a plausible-sounding recommendation. The system had created it. Fortunately, only one product was affected. But the incident posed a basic question:
If an AI system cannot reliably tell the difference between what an expert actually said and what the AI inferred, can it preserve knowledge?
Arjun rewrote the system. Now every answer had to contain:
- Source
- Video/audio reference
- Level of confidence
- Status of expert validation
If there was no verified source, the system would reply:
The knowledge repository does not contain sufficient verified information to answer this question.” This reduced the hallucinations. But it also made the system less chatty.
Now the team had a trade-off:
- More useful AI = higher risk of hallucination
- Restricted AI → Less risk but less flexibility
The Problem of Human Knowledge
Another surprising result came out of the project. The apprentices were learning quicker. But something troubling struck Raghavendra. Before, apprentices would observe him, ask questions and try things out. Now they questioned more and more:
What does the AI have to say?
One apprentice refused Raghavendra’s suggestion, because “the system says the other way is better” Raghavendra felt angry. “That system learned from me! The situation had created a paradox. The AI meant to help preserve human expertise could one day lessen the need for human experts.
The team did an informal experiment. There were two sets of apprentices trained. Group A: Training with the help of AI.
Group B: Received the traditional apprenticeship training.
- Group A performed routine tasks faster after six
- However, Group B performed better in uncommon situations that were not available in the knowledge repository.
“This was an unexpected finding. AI was good at pulling up stuff that was already known. Humans were still better at interpreting novel situations. So, Arjun proposed to change the purpose of CraftMemory. It would no longer try to supercede the master artist.
It would be a: “Digital apprentice assistant” instead. Its job would to be:
- Obtain knowledge
- Describe documented practices
- Demonstration show
- Identify similar historical examples
- Remind apprentices about safety issues
- Suggest questions to ask the master
But the human expert would have the final word.
The Decision
Three years into the project, Raghavendra was close to retirement. There were three possible futures for the workshop.
• Choice A — Sell the CraftMemory
HeritageCraft Global remained interested. The company now offered Rs 1.2 crore for sole commercial rights over the knowledge system. The money would secure the family’s finances. But this knowledge would be primarily owned by a private company.
- Option B – Maintain CraftMemory Privacy
The family could retain full control. The system could only be used in their workshop. This would safeguard the family’s knowledge, but limit the project’s potential reach. Other artisans with similar problems would have to come up with their own systems.
• Option C – Establish a Community Knowledge Trust
Arjun proposed a third way. The family might set up a Traditional Craft Knowledge Trust. Under model,
- Knowledge is owned by artisans
- Contributors receive credits
- Access to sensitive information limited
- Knowledge verified can be shared for education
- For commercial use you need permission
- content produced by AI should identify its source
- Community representatives engage in governance
- Revenue from commercial applications is shared with the knowledge
CraftMemory may be the platform to save knowledge from many traditional industries. The first phase could be:
- Woodworking
- Handloom
- Ceramics pottery
- Metals craft
- Conventional food processing
But that’s an option that would need substantial investment and governance. There was no guaranty that it would ever be financially sustainable.
The Final Conversation
On the last day of the pilot Arjun played a video to his father. It showed Meena, an experienced artisan now, teaching a new apprentice.
The apprentice said this:
How will I know when the wood is ready?” Meena grinned.
” she said Ask CraftMemory first. Then come and show me.
Raghavendra watched in silence. A few seconds later he said, “Maybe that’s what we’ve been trying to do all along. Arjun turned to him. “Preserve the knowledge?” Raghavendra shook his head. “No.” “Keep it paying it forward.” The family had a decision to make now.
Will CraftMemory become a private business, a family archive, or a community-controlled AI platform to preserve disappearing traditional knowledge?
The decision would not only affect the future of one workshop, but potentially how society thinks about AI, cultural heritage and human knowledge.
Discussion Questions
- What is the difference between explicit knowledge and tacit knowledge in the case?
- Which types of knowledge can AI preserve effectively?
- Which types of knowledge are difficult for AI to capture?
- Why did the initial AI system hallucinate?
- How would you validate AI-generated knowledge?
