.jpg)
October 4, 2025
Workshop recap: From launch to scale: Building knowledge bases that grow with your product
On October 3, we hosted a live workshop with Nivedha Venkatesh, CEO and co-founder of Pageloop, on how startups can design knowledge bases that evolve alongside their products. Our conversation explored why documentation is no longer just an afterthought, but the connective tissue between teams, customers, and increasingly, AI systems.
Why knowledge bases matter early
Nivedha emphasized that even in the earliest stages of a company, documentation prevents silos and reduces repetitive questions. At Pageloop, engineers are expected to update docs every time they ship: not only to keep teammates aligned, but also to build empathy for customers who rely on those docs to understand new features.
Writing for agents and humans
A recurring theme was the balance between human readability and AI usability. Documentation today isn’t just read by customers or teammates, it’s ingested by LLMs powering support chatbots and search. That means headings, subheadings, and alt text aren’t just nice-to-haves, they’re essential for both accessibility and machine parsing.
- Overlap: Clear structure, concise explanations, and alt text improve the experience for both people and AI.
- Non-overlap: Formats like LLM.txt make docs more digestible for agents, even if humans never see them.
The role of visuals
While screenshots are increasingly well-handled by multimodal models, GIFs and videos remain problematic. Nivedha cautioned that relying solely on visuals without accompanying text risks losing meaning for AI systems. The best approach is to pair visuals with good alt-text captions and clear explanations. That way, humans can skim the images, but LLMs still have structured context to pull from.
Self-improving knowledge bases
Audience questions sparked discussion about AI-generated docs. While LLMs can accelerate drafting, Nivedha stressed the need for human review. Just as developers use plan mode in Cursor or Claude Code to check code before execution, documentation teams must validate AI-written content to ensure accuracy and trust.
Beyond customer support: internal alignment
Good documentation isn’t just external-facing. It’s also a cornerstone for onboarding and training. Instead of bogging down senior employees with repetitive Slack questions, teams can invest in knowledge bases that scale culture and processes. With LLM-assisted tools like voice-to-doc workflows, internal documentation is easier to produce than ever.
Who owns documentation?
One challenge is ownership. Historically, docs have floated between product, support, and growth. Nivedha noted a shift: support teams and CEOs are increasingly taking the lead, driven by two forces:
- Operational cost: Outdated docs immediately increase ticket volume and support costs.
- Generative engine optimization (GEO): Regularly updated docs are crawled more often by LLMs, making products more discoverable when customers ask AI tools about solutions.
What’s next for documentation
Looking forward, Nivedha highlighted three big shifts:
- Discovery is moving from Google to LLMs, making documentation a critical surface for growth.
- Form beats format: less energy on custom themes, more focus on content quality.
- Developers demand freshness: tools like Cursor and Context-7 already pull live docs into coding environments, making outdated documentation a blocker to adoption.
A huge thank you to Nivedha for sharing her insights, and to everyone who joined us live.
