Twitter, Human Technology FoundationLinkedIn, Human Technology Foundation
Shedding light on “Shadow AI”: HTF Workshop

AI in the workplace: from field to framework

On 17 September 2025, partners and practitioners joined the Human Technology Foundation for a working session on “Shadow AI” , i.e., the generative AI uses that are emerging outside official channels.

The workshop drew on the Inria-datacraft study conducted with 14 organisations, and on a framing presentation by Yann Ferguson, with a clear objective: turn informal practices into a shared strategy that is both useful and safe.

What we’re seeing

  • In the organisations observed, AI is used first and foremost to augment work (writing, summarising, research) rather than to automate it.
  • A significant share of usage remains largely invisible, making it harder to ensure quality, compliance and skills development.
  • The discussion helped to outline a clear trajectory: make practices visible, share them and secure them, without stifling initiative.

What we’re learning

  • Making practices legible speeds up collective learning and reduces duplication.
  • Putting quality of work at the centre is essential: clarifying the expected output, the conditions under which AI can be used and the checks required helps avoid unwanted side effects.
  • Training must go beyond prompts: core foundations, evaluated use cases, legal and ethical aspects, and the role of managers as architects of collective intelligence.
  • A co-designed framework (authorised and prohibited uses, illustrated by role or profession) protects data without suffocating innovation.

What we’re putting into practice

Shadow AI is not an anomaly to be corrected, but a laboratory for innovation.

With the right governance, it accelerates AI adoption and reveals untapped sources of value. Without adequate guardrails, it undermines quality of work, compliance and organisational learning, and exposes teams to unnecessary risks.

The real question then becomes: which activities should we enhance with AI, which should we not delegate, and how do we preserve and secure human expertise over the long term?

Next steps

Many thanks to our speakers and participants for the depth and quality of the discussion. This second workshop is part of a broader series structured around three key workstreams:

  1. Mapping risks and governance frameworks
  2. Quality of work, skills development and the place of human creativity alongside tools
  3. Organisational transformation, sustainable operational performance and social dialogue

Each session combines an assessment of current uses, a sharing of practices and the co-design of operational tools.

To register for these workshops or receive session summaries, contact our partnerships team.

Related Articles