AI and environmental monitoring

An AI Story

A short while ago, our Sentience System went through one of its toughest challenges: We had just won a tender in one of Europe’s largest capitals. In the next phase, our solution was benchmarked against competitive hardware products from two traditional hardware manufacturers. Both are big players in the industry with a combined existence of nearly 250 years, around 400.000 employees and multi-billion dollars in revenue.

Technically, the challenge compared an AI-enabled solution with air quality monitoring with pure hardware or – in other words – it was a performance evaluation of an innovative solution developed by a young company against decades of experience in the traditional way of measuring.

This would have been hard enough, but in addition, there were two more constringent facts:

  • The client knowing we use AI, did not allow us to train the AI any further after the deployment of our measurement device. So, we could not use one crucial element of our AI-based, self-learning solution. We had to rely on the initial algorithm without the advantages of continuous improvements,
  • While preparing for delivery, we lost our training station at a public measurement station in Munich due to the local authority’s decision. So, even for the initial calibration of our Sentience device, we had to rely on the algorithm as developed so far – not applying our machine learning-based process.

So, it seems like fighting with one hand tight behind the back. Would we have any chance? However, we were confident in our solution, firmly believing that we might not be able to deliver the best possible but still acceptable results.

After 15 weeks, the evaluation was closed, and the results were shared:

Not only that we achieved «acceptable» results, we even clearly outperformed the competitors in two of three accuracy-related categories.

So, what’s the conclusion?

  • AI-based air quality monitoring can replace traditional, costly and bulky, hardware-centric measurement methods without compromising accuracy.
  • The AI we developed for our products is robust and can deliver acceptable results even under unfavourable conditions.
  • There is strong evidence that if the AI-based models are fully leveraged – including continuous improvements by machine learning – this innovative solution will deliver accuracy levels which are also challenging traditional monitoring hardware.

The results of this comparison are public. Drop us a line (, and we will share them with you.

Read more about why AI is favourable.