Q: How do you work with farmers?

A: We deliver robotic 'Harvestas' to allow farmers to automate the harvest of heads they sell to supermarkets, while delivering to us the stem and stalk that we then process.

Broccoli is a high value crop, but it is a low margin crop. A significant factor is the cost of harvest which is driven by current manual harvesting practice and the requirement for multiple 'passes' over the same field. The increase in casual labour rates are a major driver of farmgate inflation. Addressing this through automation will improve crop profitability for the farmer.

We also help farmers by overcoming a critical constraint i.e., constrained harvest capacity due to availability of casual labour. Broccoli may be left unharvested due to lack of casual labour, and Harvesta can overcome that.

In addition, broccoli may be left unharvested due to a glut, in which case we can take that crop (including head, stem and stalk) and turn it in to product, so let the farmer make something from crop that would otherwise be a total loss.

Our estimate is that, in total, there is up to 5x more broccoli biomass available than currently makes it to the supermarket shelf, and by working with us the farmer will share in this opportunity. This high waste ratio is because more than 75% of the aboveground parts of broccoli are left in the field, and the majority of these are edible, and farmers may plant 17 plants for every 10 that are actually harvested. In addition, there may be losses due to cutting heads down to florets and also for aesthetic reasons.

We expect capabilities to further improve as our 'Harvesta' technology evolves, part of which is the development of the data platform ('Intellia'). This will evolve to support reporting of historic field performance with HPI (Historic Potential Index), field performance with yield map processing, harvest date prediction models and yield prediction models. We intend to improve the sensing on Harvesta to allow better selection based on characterisation of colour, and identification and reporting of disease and pests, and potentially sensing of soil characteristics. We hope that this data will allow more efficient use of fertilisers, more targeted use of insecticide, and better use of water to support regenerative agriculture, specifically with respect to soil health.

Q: How do you expect to impact crop productivity, quality and total cost?

The analysis of our data indicates that yield rates from manual harvest are not 100% (i.e., harvest of heads within target size range), and there is room to improve on yield rates to a level that could impact crop profitability.

Harvesta can be operated in the dark so crop can be harvested at night/early morning when it is less metabolically active. In addition, farmers can harvest crops with long-stems and trim them just before shipment at the packhouse. Both of these could extend product life.

Obviously, the primary benefit of automation is in reducing harvest labour cost, a major driver of costs and cost inflation. In addition, a night/early-morning harvest means that the crop is cut at a lower ambient temperature reducing cooling costs at the packhouse, a major driver of power costs.

Q: Is 'Harvesta' needed to gain access to feedstock?

It is possible to use 'packhouse waste' as feedstock, but this the minority of the available feedstock, with the majority being left on the field. Without Harvesta, it is not possible to economically access the majority of the stalk for feedstock as harvest labour is expensive and in constraint, and the stalk is low value relative to the flower.

Manual harvest of majority of stalk would require the manual labour force to move ~5x more mass and strip the leaves by hand which would reduce harvest efficiency and compromise the product that is currently sold. It would also risk injury to the labour force in the form of RSI. Thus, to access the stem requires automation and our 'Harvesta' is the only automated selective harvester designed to harvest the stalk - others just take the head (flower).

While farmers could perform a slaughter harvest, especially as a final cut, they would risk damaging the head (which is what they currently sell); they would face the problem of removing the leaves (as leaves give any product a 'grassy' taste and have low nutritional content); and, they would have the issue of the large mound of biomass 'cooking-off' and becoming non-useable as food.

In summary, it is possible for us to produce our ingredient products without Harvesta, but the proprietary access to low-cost feedstock enabled by Harvesta transforms our business model. Together Harvesta and Bia have considerable synergy value.

Q: What are the advantages of 'Harvesta' compared to other selective harvesters?

When assessing the potential for robotic automation, the key issue is the ‘actuator’ - human or robotic, and the comparison of the difference between them in terms of economic performance to achieve the defined task.

One of the challenges with automating the harvest of crops is that the actuator may have a maximum speed it can operate at without damaging the crops (i.e., reducing yield through damage) or missing crops (i.e., reducing yield by omission).

It is this fact that means that the use of conventional industrial robots has limited application in broccoli harvest. By way of example, having 4 industrial robots costing (say a full cost of) $70,000 each within a system that is priced to the farmer at (say) ~$700,000 with an annual maintenance cost of ~$100,000 will have a poor pay-back even if the productivity is better than 8 human workers at $25/h x 2,000 hours per year.

Systems based on industrial robots tend to be heavy (which impacts deployability and soil compaction), power-hungry, frequently breakdown in an agricultural environment and may require an expensive support staff to oversee them (which may not be quickly available in rural areas). It is the breakdowns and remediation that is the critical issue as once a farmer has replaced manual labour with robotics solutions downtime means lost crop.

Our approach is different; the robotics do not require 6 degrees of freedom as they are essentially performing one task and do not need the cost and complexity that enables a range of tasks or infinite variability of movement. The ‘actuator’ is a pair of light-weight food safe plastic tongs that lift the cut plant if the sensor has decided that the plant is ready for harvest and triggered a cut, and do not damage the flower as they do not touch the flower.

Our actuators are placed ready to be triggered, and crucially there are many of them that rotate on a chain. They are a low-cost item that is designed to be sacrificial if there is a stone strike and are quickly and easily field replaceable in this event. As it is, none have been damaged so far. But it is better to plan to accommodate for likely failure modes.

This approach means that our system can run fast, currently ~ 4 km/h, and we hope that it will soon be capable of 5 km/h, with a likely realistic ultimate maximum of 8km/h. There is probably no point in exceeding this speed as the crop harvest ceases to the constraining factor, and instead becomes off-load handling, the logistic chain to the packhouse and the packhouse itself.

The data so far indicates that our ability to select is at least as good as humans (human harvest typically results in an ~80% yield in terms of completeness) and we believe we can get it to the high 90’s (it will not be 100% due to occlusion). In addition, there is very limited evidence of any crop damage. Importantly, we believe that the economics of use are very compelling versus both human and robotic solutions.

So, in summary, we are different as we use a multitude of low-cost dedicated-purpose actuators that are placed such that they do not move relative to the plant, and we use computer vision to make a ‘yes’/’no’ decision, which together allows us to run fast. The alternative approach is the use of a low number of high-complexity, high-cost actuators which require advanced sensing and move relative to the plant, and which thus move slowly. Our approach is far less costly, in terms of CAPEX and maintenance, and offers the possibility of further material cost reduction even at modest scale (tens of units). This is the unique competitive advantage of our approach, which allows faster crop harvest and intrinsically superior economics at all unit volumes.

In addition, by deploying our technology on 4-wheel drive/steer self-powered sprayer platforms we achieve a system with good ground clearance (to avoid damaging crop), an excellent turning circle which reduces issues in the field headers, that is comparatively lightweight so reduces soil compaction and is easier to use in mud than a trailed system. Most importantly, it makes maintenance easier and reduces development and operational risk.

Q: What is the commercial model behind 'Harvesta'?

A: Our commercial model is 'Harvesta-as-a-Service' ('HaaS'), requiring the planter (farmer) to grant us exclusive rights to side-stream. In our HaaS model the farmer will crew the Harvesta. Pricing is set to share the economic benefit of the harvest automation between us and the planter.

The estimated gross value per acre in the US is $6,800 for fresh market broccoli. The costs of production of broccoli vary depending the production location. It is labour-intensive, especially for harvest and post-harvest handling and packaging.

A 2020 study by the Charles H. Dyson School of Applied Economics and Management College of Agriculture and Life Sciences at Cornell University found that average harvest costs in New York were $1,010 per acre (of which $897 is labour), representing 22% of variable costs of $4,685 and 17% of total costs of $5,784. Similar figures are reported in a 2023 UC Davis study. An analysis by Bayer, a major producer of broccoli seeds, cites harvest costs per acre of $1,395 in New York (37% of total costs) to $3,624 in California (39% of total costs). Around 110,000 acres of land are farmed with broccoli in the US, implying that total harvest costs in the US exceed $110m annually.

Similar data are not available for the UK, however employment costs in UK fresh produce businesses account for a large portion of total production costs, sometimes up to 60% for certain crops. The employment cost increases announced in the October 2024 budget are expected to raise production costs by an additional 10 to 12%, and it is likely that UK costs per acre will exceed those in the US if they do not already. Around 7,500 acres are used in the UK to for broccoli, indicating total harvest costs of circa £7m annually.

Given the materiality of harvest labour cost and its impact on cost inflation, and potential for cost reduction of harvest through automation, we believe that a model based on share of saving aligns incentives between ourselves and the farmer. In addition, the farmer shares in the upside of the side-stream.

Q: Will you sell 'Harvesta'?

A: Possibly in time, but it will still require side-stream rights to be assigned to us.

Q: What is your position on autonomy?

A: Autonomy for Harvesta is on our product roadmap. We do not see autonomy for broccoli harvest as 'controlled from the farm office' but as being supervised by an on-field worker once the worker has driven the Harvesta there. Product off-load to an accompanying vehicle will likely still require workers, so autonomy adds value by further reducing the total size of a harvest crew. It will also allow use of the use of AB lines created during automated planting in order to improve precision.

We will not build the autonomy modules ourselves but will use 'commercial-off-the-shelf' ('CotS') systems as 'bolt-on' integrations to the CotS chassis on which we deploy our automated harvest modules. Our indicative time-line is: 2026, data gathering during normal operations, 2027, in-cab oversight of autonomy; 2028 on-field oversight of autonomy.

Because we intend to integrate other companies' technologies, our delivery risk of autonomy is limited and manageable. Our main focus will be on developing and deploying the protocols to ensure safety, for instance remote 'kill switches' to allow all on-field staff to stop operations if there is any risk of harm, and ensuring that autonomy delivers economic value to the farmers.

Q: Where do you intend to operate?

A: We are starting deploying Harvesta in the UK, then deploying in Spain in the UK off-season - producing across the UK and Spain is how supermarkets give 12 month access to brassicas for UK consumers. Doing this will allow us to supply food manufacturers on a 12-month-a-year basis. In 2023 63,000 tonnes of broccoli was grown in the UK (implying over 325,000 tonnes of potential feedstock), and in 2022 ~477,000 tonnes was grown in Spain (implying 2.4m tonnes of potential feedstock). At the European level, Spain represents 34.4% of total cauliflower and broccoli production with 677,280 tonnes. This production has been increasing over the last 10 years.

The US is our next market. In 2022 ~472,000 tonnes of broccoli were grown in the US (implying almost 2.4m tonnes of potential feedstock). In the US, broccoli cultivation is highly concentrated with circa 90% of broccoli being grown in California over a season of up to 10 months, with a significant amount of off-season crop being grown in Arizona. For this reason we intend to focus deployment in California and Arizona.

It is likely we will franchise in the EU and the rest of the world to allow us to focus on the US opportunity.

Q: Can you work with other feedstocks?

A: We can work with cauliflower as a feedstock although we do not have an automated harvest solution for it. While cauliflower has less edible biomass left on the field post-harvest, it suffers from large packhouse losses due to a tendency to discolour when bruised or exposed to UV. This does not affect taste or nutrition, but consumers (thus retailers) will not buy it. As a result, there is significant availability of cauliflower biomass. In 2023 71,000 tonnes of cauliflower was produced in the UK, implying available biomass of a similar value. In 2020 the US produced 1,003 million pounds of cauliflower (~455,000 tonnes) and Spain approximately 200,000 tonnes.

We expect other feedstocks to follow.