HAIP
AI-Governance-Platform
Putting enterprise AI in 360 employees' hands, without a single data leak
A 360-person manufacturer wanted to give staff access to large language models, but their IT lead was rightly afraid someone would paste a customer list into ChatGPT and watch it leave the building. I led the UX for a platform that made AI both safe and familiar.
The client manufactures consumer electronics and employs around 360 people. Their IT lead saw the risk before anyone else did, and came to us directly. Staff wanted AI; the danger was obvious.
A sales rep pastes a customer list into a public chatbot, and that data is now sitting on someone else's servers, outside the company's control.
The brief: enable AI across the organisation without exposing sensitive data, and without forcing employees to learn an alien new tool. Two goals that usually pull against each other: security and ease.
The core of the platform is a vertical workflow builder. A non-technical admin defines an AI workflow by adding nodes via drop-selects: a data source, a pre-processing step, the LLM, a post-processing step, and the output. The model itself is swappable (GPT via Azure, Claude, Mistral), chosen by whoever delivers the best result for that task. The LLM node also carries its own configurable system prompt.
The security mechanism is built into the flow, not bolted on. Before any data reaches the model, a pre-processing step encrypts and anonymises the sensitive fields; after generation, post-processing restores them. The model sees structure, never secrets.
This wasn't “we want AI to look cool.” It was “we need AI we can trust, before someone leaks something we can't take back.”
Henrike Helmer Head of IT, Hartmann Electronics HAIP left me with a conviction I bring to every AI project: in the enterprise, the hard part of AI design isn't the model or the prompt. It's governance and trust, making the safe path the easy path, so people choose it without being told to.
It's also the clearest proof of where I'm heading. Not “I added AI features to a screen,” but “I designed an AI system that solves a real security problem and shipped it.” That's the work I want more of.
