This is the third of a series of interviews I am conducting with members of the Machine Commons supplier Collective. Subscribe to the site to be alerted about future posts, or become a partner today!
Colin Dillon and Essa Jabang are the co-founders of Taybull Ltd, a ‘Low-Code Rapid Application Development Platform’ company based just outside London.
How’s life in the pandemic era? Have you felt much impact?
“They keep lifting restrictions - quarantine lifted, then putting them back on. In day to day life, you have to wear a mask in all public places. So no, it’s not ideal”
And business specifically?
“It’s been tricky for us as the main product we launched was for the airline industry. So yeah, wasn’t perfect."
"It wasn’t the best timing for a once-in-a-lifetime product geared at the airline industry!"
Oh dear. Tell me about the airline industry solution?
“We built an OCR (optical character recognition) solution; you may not know there’s an ‘airway bill’ – a piece of documentation – that goes along with every airway transportation. It holds information about where stuff is from, where it’s going to, the weight. It’s used by every airline.”
“Well, instead of keying in all the info on a piece of paper, you can just scan a sheet. The software we built reads the text and puts it into a digital format.”
"We’re building a [generic] document scanning solution off the back of that.”
How did you choose the airline industry?
“It wasn't one of the industries we thought we were going into!”
“We were talking about an analytics product we were going to sell to them, but then – while we were there – the manager mentioned this huge problem they had.”
“The manager asked: ‘can your tool do this functionality; can it extract this info and then send it to other systems?’. And we said, well no we don’t have that yet… but we can build it!”
Huh. Pretty random!
“Yeah we didn’t even know this need existed at all!They said if you can build this then this is where our interest would be.”
Open dialogue with your client helps lot, it seems.
Having talked to a few machine learning startups, this is definitely a common trend. Entrepreneurs find themselves in the right place at the right time, knowing the right people, win a project, then they realise they can apply the nuts and bolts of the project elsewhere. It’s basically paid R&D.
Colin and Essa worked together long enough that they thought they may as well start a business.
“Colin and I worked together at Yell in the R&D team.We were doing a lot of stuff around machine learning. Colin’s got the entrepreneurial spirit and I thought the same way, so we decided to start up a company.”
(There’s definitely a type of person who revels in the instability of enterprise!)
Any strategies for the Covid era?
“Covid has definitely put a spanner in everybody’s works. But we didn’t have offices. We didn’t have loads of monthly outputs, so we were very lean anyway. Thankfully we weren’t left in a situation where we had huge bills and suddenly no revenue.”
“We know plenty of companies that just fell off a cliff.”
Following the crash of the airline industry, what next?
“We have a legal tech solution, but a really relevant area we’re looking into is agri-tech.”
“Yes, yield estimation to enable farmers to sell to an end user. Usually the farmer sets their ‘best price’ then sells to a middle man, who then sells it on. Obviously, inaccuracies here mean underselling the value of their crop.”
“So, we’re using drones and neural network tech to take images of crops, using AI to make accurate estimations on the quality of produce, disease recognition, helping the farmer make price estimations.”
“We’ve already started building the model and working with farmers who are willing to pilot it.”
You need data to make a solution, but you need a solution to justify getting the data (data --> model --> client --> data).
Chicken or the egg. How do you beat that?
“Actually, getting the farmers to get the data is pretty simple for us. Through Essa’s connections –it’s so informal. Some of these relationships are literally just family and extended social networks.”
“We have a small office in Gambia to just do data gathering. A few farmers have just allowed us to start collecting data. The core engine is something that we’ve had for some time.”
What’s your expectation or hope for the agritech part of your business?
“We built it in contact with end users, so we want it to be really useful for them.”
“The idea is for these farmers to use smartphone tech to sell directly to their end customers (such as hotels and shops).”
“Some farmers are seriously unable to estimate what their yield is worth.”
You seem to have your fingers in a few pies?
“Yeah, exactly, we don’t want to be pigeon holed with one solution. The aviation sector fault has really reinforced that view.”
“One bad thing happens to a market segment and you can be in big trouble very quickly. So the agri-tech industry was where we have to be.”
I mention vertical-specific knowledge.
How do you find the challenge of these varying industries?
“In terms of data gathering, it’s fairly similar in that basically you can’t have a quality output without a quality input. The reason we picked agriculture is because my business partner has a strong interest in it. The research that was done in this space is really exciting.”
How do you choose what to focus on?
“Covid has made it difficult. Yes, we start with an analysis, but we also just look at the resources in our team.”
“Legal tech was a market that just presented itself, one of our partners happened to know the a Gambian Legal Practitioner. We spoke to him, performed an analysis looking at the UK legal tech, and thought ‘yes there’s definitely an opportunity here’.”
“Some of the requirements he was talking about we already had the components for.”
“Originally it could have just been a consulting operation, but in this case because he doesn’t have a technical team let’s build it together and manage the product going forward.”
So, a consulting and product mix is your deliberate business plan?
“At first, we reasoned there are two things we could do as a business. We could focus on consulting, or we could just build products.”
“But, combining these approaches means we gain new market knowledge (earned from the days consulting in the industry) and then we capitalize on that new market knowledge by building a product.”
“Most of our revenue right now comes from consulting, but the reason we’re expanding is to have products in different domains.”
“We’re aware that one of these products could just hit.”
What are the most important aspects of ML technologies?
“Our strength is in tech. So when we go into these industries, we don’t lose sight of what we bring in – and what we don’t know. We have to talk to a lot of farmers.”
“For tech to work you need to understand your domain.”
“Also, keeping components as broad as possible. We always think about ‘usability’, it has to be able to be repurposed for future ventures.”
(Colin chimes in: “I think the word is pivot!”)
What have you learned?
“We had to learn fast and understand exactly what the market was all about.”
(Essa chimes in: “Yeah! Very fast!”)
“The errors you make come from not knowing about industries. Blindly going in and making mistakes.”
“When we started, we concentrated a lot more on the technology than the business. We were thinking product, product, product, but that’s something we’re correcting now. We didn’t quite get that balance right until now.
“When trying to understand new market, what the client is telling you they need isn’t exactly right! When you go in and learn more about the industry you realise it’s not as simple as they’ve suggested!”
What are you watching out for next in the ML world?
“I follow what Google, Facebook and Microsoft are doing very closely. What they’re doing makes stuff so much easier for us."
"You spend so much time building libraries but the libraries they’re building make it much easier.”
“Any algorithm you make you have to score it. To get from 90 to 91%, you’re trying to get it to be as good as a human."
"With a startup like ours we’d be spending years improving like this. But these big companies are doing all of this with their frameworks and they’re getting better and better by the day.”
“Things will continue to become more main stream–easier to use. Like autoML, it still takes some technical knowledge but the world already has a lot of software engineers. You’ll see many more AI-driven products released, as they’re being enabled by these large platforms.”
“Two years ago if you’d said where we’d be now I would have not believed it.”