Polish AI harness: a new series on how to harness AI for work in a company
Series · episode 1 of 3
An AI model has the strength of a horse, but it is the harness that decides where it will run. In three parts, every Friday, we show how to build one in a company: from technology through people to law. We start with technology under control
The “Polish AI harness” series follows one example from start to finish: a salesperson asks an AI assistant to prepare an offer for a client. This single task shows everything that determines the value of AI in a company: where the model gets its knowledge, where the data goes, what the assistant can do on its own and when a human decides
The author of the series is Andrzej Okoński, Head of Software Development at allclouds.pl. Day to day, he and his team develop our SAIE ecosystem: TWIN:DESK, PROXY:AI and CDF Platform
Three parts, every Friday
Part 1: technology under control
A good harness does five things: it delivers the right knowledge, controls the data path, matches permissions to the task, makes it possible to stop and explain actions, and measures the quality of the work. In the post, we show how TWIN:DESK, PROXY:AI and the CDF methodology do this
What you will find in part 1
- why the same model can produce completely different results in two companies
- how the assistant tells the current price list from an archived one
- what happens when the permitted model is unavailable
- why the offer is sent by a human, not by an agent
A good Polish AI harness should increase the organisation’s capabilities and strengthen people’s ability to steer technology
The full post is on our blog