Takeaways from Beakbook’s ‘Agriculture’s AI Revolution’ webinar

Beakbook is a tech startup from London that works with poultry companies to explore how they can best use AI, including by building tailored AI models and sensor systems. On November 23, 2023, they held an online webinar with the UK poultry integrator Avara, who rear 4 million chickens per week for human consumption.

Here were our key takeaways from the webinar:

1) AI’s predictive capabilities could help the chicken farming industry remain profitable.

Currently, producers gather a huge amount of data but don’t have the resources or skills to actually do that much with it. Beakbook’s main goal is to help poultry producers predict the ‘harvest weights’ of their birds as this can help producers determine bird health, make decisions around stocking density, and ensure that their birds’ weights match the requirements of their customers. They use AI to process all the necessary data and make those predictions more accurately than producers are currently able to.

2) AI’s gains are expected to be significant, but marginal.

The poultry industry is already incredibly efficient. (As you’d expect, the two companies gave a very sanitised overview of what this ‘efficiency’ entails, skating over e.g. the welfare issues surrounding the use of fast-growing chicken breeds). Given this efficiency, AI isn’t expected to bring huge gains, but rather to bring marginal gains across a variety of processes. In this vein, AI is expected to complement human farming for the short to medium-term, rather than fully replacing human farmers any time soon. Beakbook were very keen to manage expectations on this front.

3) Getting farmers on board with AI tech can be challenging.

AI can be prohibitively expensive for farmers, especially when it comes to sensors and other hardware. There’s also general distrust of complex tech solutions among older and more traditionally minded farmers. AI tools will need to be made as easy as possible to incorporate into farmers’ daily routines while tangibly saving them time and money.

4) There could be real opportunities for improving animal welfare - but only to the extent that these align with productivity.

Welfare improvements from AI could include using sensors to: track behavioural traits and correlate these with different diseases, to allow earlier detection, prevention, treatment, and culling; monitor the sound frequencies of chickens’ vocalisations to detect signs of distress; and measure levels of volatile organic compounds and pick out correlations between rates of those compounds in the air and levels of disease. All of this would require a huge amount of data to do well, which the industry can’t currently collect.

Predictably, animal welfare was only considered in the framing of ‘animals need to be healthy to be productive, so high productivity inevitably means high welfare.’ When asked about potential trade-offs between productivity and welfare, the companies waved this off by a) reassuring attendees that legal animal welfare standards would prevent major welfare concerns and b) putting the burden on consumers to find out whether companies are using AI responsibly.