Joel Hellermark founded Sana at 19 to explore how AI could improve learning and access to knowledge. Sana later expanded into enterprise search and agents; Workday completed its acquisition of the company in November 2025. These lessons trace his thinking on learning, interdisciplinary teams, and AI at work. — Workday Acquisition Release.

Infographic for "Lessons from Joel Hellermark".

On the Future of Learning and AI

  1. Personalize the Path: Hellermark envisioned learning software that identifies each learner’s knowledge gaps and adapts the next exercise or explanation rather than giving everyone the same sequence. — 52 Insights Interview.
  2. Move Beyond One-Size-Fits-All: In his TEDx talk, he argued that education still often distributes identical material to students with different needs; he viewed personalization as the next shift. — TEDxStockholm Talk.
  3. Combine Data and Neural Networks: He described online education data and neural networks that learn from data as the combination that could make personalization possible at greater scale. — TEDxStockholm Talk.
  4. A Hypothetical Learning Multiplier: In his 2017 talk, Hellermark imagined how a reported one-year learning advantage might compound over twelve years; his 100-fold figure was a hypothetical extrapolation, not an observed outcome. — TEDxStockholm Talk.
  5. Use AI as a Learning Tool: He compared AI with a powerful calculator: students should learn to use the tool instead of spending all their time on tasks it can handle. — Forbes Education Interview.
  6. Generate Learning Content for Context: Hellermark proposed moving from static content to material generated for a learner’s knowledge, role, interests and context. — Forbes Education Interview.
  7. Build a Long-Lived Knowledge Assistant: His envisioned assistant would draw on a person’s writing, role, searches and courses to tailor help across time; this was a vision, not a completed universal product. — Forbes Education Interview.
  8. Progress from Search to Chat: Hellermark describes enterprise knowledge tools evolving from internal search to conversational answers over company information. — Computer Weekly Interview.
  9. Make Meetings Searchable: In Sana’s 2023 demonstration, Hellermark showed a meeting assistant recording summaries and returning users to the moment a topic was discussed. — Sana AI Summit 2023.
  10. Move from Answers to Actions: He demonstrated an assistant that could retrieve company information, query data and take software actions; his later description extends this toward teams of task-solving agents. — Sana AI Summit 2023.

On Building Sana and Entrepreneurship

  1. Connect Human and Machine Learning: Hellermark says Sana grew from his interest in applying machine learning to human learning; the initial question preceded a fixed product idea. — Scandinavian MIND Interview.
  2. Pursue Broad Learning Access: In 2018, he described an ambition for Sana’s technology to be used within learning products globally. This was a stated aim, not a claim that it had already happened. — 52 Insights Interview.
  3. Organize Around Missions: Hellermark says Sana looked for people energized by hard problems and periodically reorganized its teams around top missions. — NEA Founder Interview.
  4. Pragmatic Dreamers: Hellermark used “pragmatic dreamers” to describe people who combine generational ambition with close attention to product and execution. — NEA Founder Interview.
  5. Recruit Exceptional Teammates: Hellermark repeatedly treats recruiting as central to building Sana; he described patient, sometimes years-long efforts to persuade strong candidates to join. — Scandinavian MIND Interview.
  6. A Company for Long-Term Work: He said he designed Sana as his life’s work, something he hoped to pursue for decades rather than a short-lived venture. — Scandinavian MIND Interview.
  7. A Bell Labs-Inspired Environment: At Sana’s founding, he wanted scientists, designers, engineers and marketers working across disciplines to create tools for learning. — Scandinavian MIND Interview.
  8. Cross-Disciplinary Teams: In a 2023 interview, Hellermark described designers writing code and customer-success staff building dashboards, reflecting his preference for interdisciplinary talent. — Scandinavian MIND Interview.
  9. Make Chance Collaboration Easier: He arranged the office and shared routines to increase encounters between people in different functions, such as engineers and designers. — Scandinavian MIND Interview.
  10. Recruit Patiently: Hellermark said Sana might spend as long as two years convincing a promising person to join; he treated talent quality as worth the effort. — Scandinavian MIND Interview.
  11. Expand Learning Across Companies: Hellermark’s business-learning thesis was that organizations need better ways to create, share and personalize knowledge as they train and develop employees. — Scandinavian MIND Interview.
  12. A Large Knowledge Opportunity: Hellermark said a company that solves enterprise knowledge well could become very important; this was his ambition, not an established market-size result. — Scandinavian MIND Interview.
  13. Treat Learning as Operating Infrastructure: He framed organizational learning as a practical need for onboarding, training and developing employees rather than a peripheral content library. — Scandinavian MIND Interview.
  14. From Scattered Files to Answers: Hellermark compared company knowledge tools with successive internal Wikipedia, Google and ChatGPT stages, culminating in agents that can help with tasks. — Computer Weekly Interview.
  15. Workday Acquisition Completed: Workday announced that it completed the Sana acquisition on November 4, 2025. Hellermark said joining Workday could accelerate Sana’s vision for enterprise knowledge, automation and learning. — Workday Acquisition Release.

On His Personal Philosophy and Influences

  1. Learn to Code from Open Courses: Hellermark says he began programming at 13 after finding Stanford courses online, then put the skills into practice at a digital agency. — Scandinavian MIND Interview.
  2. Study Both Kinds of Learning: He says he was interested in machine learning and human learning together, and designed Sana to explore where those disciplines intersect. — Scandinavian MIND Interview.
  3. Aim for More Than Incremental Change: In a 2018 interview, Hellermark said the ideas that attracted him as a teenager were those with the potential to change the world, not merely make small improvements. — 52 Insights Interview.
  4. Learn from Polymaths: He cites Leonardo da Vinci, Michelangelo and Edwin Land as models for combining insights from different fields, an influence on Sana’s interdisciplinary approach. — 52 Insights Interview.
  5. Work on Meta-Problems: Hellermark recounts a Buckminster Fuller–Einstein story to explain his interest in problems whose solutions could unlock others; treat the walk as his anecdote, not independently verified history. — NEA Founder Interview.
  6. Improve Access to Knowledge: Hellermark believes changes in how people access knowledge have broad downstream effects and made that idea central to Sana’s purpose. — Scandinavian MIND Interview.
  7. Augment Human Capability: He was drawn to tools that expand what people can do; Sana’s original aspiration was to use software and machine learning to augment learning and work. — Scandinavian MIND Interview.
  8. A Computational View of Intelligence: Hellermark has argued that language models raise questions about what aspects of human intelligence can be modeled computationally; this remains a perspective, not a settled scientific conclusion. — Scandinavian MIND Interview.
  9. Learn Nuance from Examples: He contrasted hand-coded expert rules with models that learn nuanced language patterns from large data sets. — Scandinavian MIND Interview.
  10. Start Building Young: The official TEDx Talks description records that Hellermark founded a video-recommendation company at 16, after learning to code at 13. — TEDxStockholm Talk.

On the Broader Impact of AI

  1. Distribution Matters for AI: In a Workday interview, Hellermark framed enterprise AI as a distribution challenge: connect capable foundation models to users and the context needed for useful work. — Workday Sana Interview.
  2. AI as a New Work Interface: Hellermark says AI can generate an interface around a task, drawing on the necessary data and tools instead of making users navigate many separate applications. — Computer Weekly Interview.
  3. Early Productivity Studies, Not a Universal Rate: In his 2023 Sana keynote, Hellermark cited early task studies reporting substantial gains for AI-assisted workers. His 30–50% summary was an illustration, not a demonstrated productivity rate for all work. — Sana AI Summit 2023.
  4. Make Company Knowledge Accessible: Hellermark described Sana’s goal as indexing company knowledge so employees could use it to pursue their work. — NEA Founder Interview.
  5. Unify Scattered Institutional Knowledge: He identified knowledge dispersed across documents, presentations, messages and employees’ heads as two related barriers to organizational learning. — Scandinavian MIND Interview.
  6. Context Matters More Than the Wrapper Label: Asked whether Sana was a wrapper around language models, Hellermark said customers care about speed and accuracy and that models need relevant company context. — NEA Founder Interview.
  7. Build Multi-Step Reasoning into Agents: He described an approach where an assistant plans, uses tools and relevant knowledge, evaluates its answer, and can iterate instead of responding in one pass. — NEA Founder Interview.
  8. Prioritize Adoption Over Immediate ROI: In a Workday conversation, Hellermark urged companies to help employees adopt AI before demanding a precise return from day one. — Workday Sana Interview.
  9. Connect Company Knowledge to Models: Hellermark sees enterprise assistants as useful when language models can draw on an organization’s own knowledge and task context. — Computer Weekly Interview.
  10. Watch for Unrepresentative Training Data: Hellermark argued that model bias becomes a problem when training data fails to reflect the people who will use the system. — 52 Insights Interview.
  11. Human Learning Remains Hard to Model: He cautioned that neuroscience still explains only part of how people learn, so educational AI cannot assume it has a complete brain model. — 52 Insights Interview.
  12. Learn Rather Than Hand-Code Rules: He contrasted manually encoded heuristics with neural-network approaches that learn relationships directly from data, particularly for complex learning behavior. — 52 Insights Interview.
  13. A New Workday Entry Point: Workday says Sana’s search, agents and learning tools are intended to connect enterprise data and workflows in a single employee experience. — Workday Acquisition Release.

Learn more:

  1. Workday — Workday Completes Acquisition of Sana
  2. Forbes — The Future of Education Will Tap AI, Not Be Replaced by It, This Founder Says
  3. TEDx Talks — How Artificial Intelligence Will Radically Transform Education, Joel Hellermark
  4. Scandinavian MIND — Joel Hellermark’s Decades-Long Speculation on Artificial Intelligence
  5. NEA — Sana: Harnessing the Power of AI to Curate Your Company’s Internal Knowledge
  6. Computer Weekly — Interview: Workday’s Joel Hellermark on Making Everyone a Leonardo
  7. 52 Insights — Interview: Joel Hellermark, The Prodigy Revolutionising Education
  8. Sana — On Knowledge, Joel Hellermark, Sana AI Summit 2023
  9. Workday — Workday Signs Definitive Agreement to Acquire Sana
  10. Workday — Workday + Sana: Unlocking the New Front Door for Work