Practice : AI Knowledge Sharing and Demos

Purpose and Strategic Importance

AI knowledge depreciates quickly. New techniques, tools, and research emerge rapidly; production experience produces learnings that are not captured in any course or publication; and the gap between what the AI field knows and what any individual team knows tends to widen unless there is deliberate investment in knowledge sharing. Teams that share knowledge actively — through demos, show-and-tells, internal talks, and written documentation — build collective capability faster than teams where knowledge remains siloed in individuals or sub-groups.

Knowledge sharing also builds the psychological safety and organisational literacy that effective AI work requires. When teams share what they are working on — including experiments that failed, models that behaved unexpectedly, and concerns about AI system behaviour — they create the transparency that enables better decision-making at every level of the organisation and builds the trust that makes it easier to raise concerns when AI systems pose risks.

Description of the Practice

  • Runs regular AI demo sessions where teams showcase work in progress, completed AI systems, experiment results, and learnings — including failures and partial successes.
  • Facilitates internal knowledge-sharing talks or workshops on techniques, tools, and domain topics that are relevant to the team's AI work.
  • Maintains accessible, well-curated internal documentation of AI practices, standards, reusable components, and lessons learned, making institutional knowledge findable and useful.
  • Creates learning communities — reading groups, communities of practice, special interest groups — that connect people across team boundaries around shared AI interests and challenges.
  • Explicitly recognises and celebrates the knowledge contributions of individuals and teams, building the cultural conditions for open sharing rather than knowledge hoarding.

How to Practise It (Playbook)

1. Getting Started

  • Establish a regular AI demo or show-and-tell session — fortnightly is a good starting cadence — with a simple format that is low-effort to prepare for and high-value to attend.
  • Identify two or three internal AI experts whose knowledge would be valuable if shared more broadly, and create opportunities for them to present to wider audiences.
  • Build a lightweight internal wiki or documentation space for AI practices and learnings, seeded with a few high-value documents, and establish a norm of contributing to it.
  • Celebrate the first knowledge-sharing contributions publicly, setting the tone that sharing is valued and recognised.

2. Scaling and Maturing

  • Develop a structured internal AI conference or learning day that brings together AI practitioners from across the organisation to share work, debate approaches, and build connections.
  • Build a curated internal AI learning pathway — covering key techniques, tools, and responsible AI topics — that helps new team members and people transitioning to AI work build foundational knowledge quickly.
  • Create a mechanism for teams to share reusable AI components, datasets, and tools — whether through a shared repository, a model registry, or a feature store — and invest in making them discoverable and well-documented.
  • Measure knowledge sharing activity and its outcomes — not just participation in events but evidence that shared knowledge is being applied — to understand what formats and topics are most valuable.

3. Team Behaviours to Encourage

  • Treat sharing a failed experiment as equally valuable as sharing a successful one — the learning from failure is often more actionable than the confirmation of success.
  • Prepare for knowledge-sharing sessions with care — a well-structured 15-minute demo teaches more than an unprepared 45-minute ramble, and respect for the audience's time is respect for the practice.
  • Ask questions actively in knowledge-sharing sessions and follow up on learnings in your own work — passive attendance does not build capability, but engaged participation does.
  • Write up significant learnings — from incidents, experiments, production observations — in a documented form that others can find and use, not just share verbally in a session that disappears from organisational memory.

4. Watch Out For…

  • Knowledge-sharing sessions that become dominated by polished success stories, creating a culture where only confident, completed work is shared and the most valuable early-stage or failure learnings are withheld.
  • Documentation that is created but not maintained, becoming outdated and untrustworthy — a documentation space full of stale, inaccurate content is worse than no documentation.
  • Knowledge sharing that is additive overhead on top of already overloaded teams, rather than integrated into the normal rhythm of work — if sharing requires significant extra effort, it will be the first thing cut when pressure builds.
  • Knowledge silos that form around individuals who are the only people with knowledge of critical systems or techniques, creating both a bus factor risk and a constraint on the team's collective development.

5. Signals of Success

  • Knowledge-sharing sessions are well-attended and actively engaged, with participants reporting that they learn things of practical value that they apply in their work.
  • Examples of cross-team knowledge transfer are visible — approaches, tools, or techniques shared in one team's demo being adopted by another — demonstrating that sharing creates real value, not just goodwill.
  • New team members can access internal documentation that helps them understand the team's AI practices, tools, and past learnings without needing to rely on individual knowledge transfer.
  • Individual knowledge contributions — written documentation, demos, talks — are recognised and valued in performance and contribution conversations, reinforcing the cultural signal.
  • The organisation's collective AI capability is growing faster than any individual hiring or training programme could explain, reflecting the compound effect of effective knowledge sharing.