Frode Opdahl, CEO of Keenious explains the origins of an AI-driven start-up and how its success depends on library buyers keeping the sector attractive to start-ups that wish to innovate the research experience.

Keenious is a new type of search engine that helps students and researchers discover relevant material.

It uses artificial intelligence to analyse documents and retrieve a list of relevant research articles. It is accessible through its website and users can add Keenious to Microsoft Word and Google Docs for efficient use. The tool uses AI to search the Open Alex data set – a free index of the world’s research.

In this interview for CILIP's Buyers' Guide 2024 , Frode said: "What we are developing is an AI tool that makes it easier to discover information that you want to read, a tool that helps you discover things that you could be really interested in."

“We are not curators of the data set,” Frode says. “Think of us as the engine that figures out what is relevant from this data set. So, we don’t have the full text itself, but we provide links to interesting articles and if you’re at a university you can check if you have access.”

Until recently the tool was available free online but it has been modified as sales to university libraries picked up: “Our new basic version has some limits on results and document size. This is to make sure it’s fair for universities that pay for our service. We still want to make sure people who can’t afford it, especially in less wealthy areas, can use Keenious too.”

Keenious has 30 university clients, their first customers mostly being based in the nordic countries, but now includes Carnegie Mellon in the US which, Frode says, “is seen by many as the birthplace of AI, so selling an AI tool to them is very exciting” (Keenious is mentioned in this article about the university library’s role in developing AI) It is also in trials with Trinity College Dublin and in talks with multiple others.

Origin

“I never worked in a library, I’m a computer science student and I wasn’t a big user of the library myself,” Frode says. “I was struggling to find relevant articles for my master’s thesis as a student. Ironically, I was working on a recommendation system for articles while I was struggling to find relevant articles. That was when I thought ‘what if I analysed a lot of the research articles and could compare them with my text?’ Then I could put the results in a sidebar next to my document and it sounded super useful and no one else had made it, and that’s basically the origin story. It was later that we saw it as a tool for libraries.”

Asked if Keenious poses a threat to the library profession, Frode points out that direction and the development of Keenious is carried out in collaboration with librarians. He adds: “They are happy because this is another tool in their tool kit to help their users.”

Referring back to his own experience as a student who didn’t use the library, he says: “A big selling point is that Keenious can be added into the workflow of all the students and faculties without them having to install it. That makes it a new avenue for the library to get exposure and a new place where the library wasn’t available before. When students or researchers open it up we make it very obvious that this is something that the library is providing.”

Less is more

Frode says Keenious is academia-friendly AI because it delivers value to its users differently to most other AI models. “ChatGPT and many other AI tools are very focussed on giving you an answer. You ask a question, it gives you an answer. All the value is in the response. We don’t focus on giving you that answer, partly because the answer might be wrong, but mainly because giving you the answer means you didn’t go through any critical thinking of your own.

“What we are developing is an AI tool that makes it easier to discover information that you want to read, a tool that helps you discover things that you could be really interested in, and the end product is not just a simple one paragraph answer, but a user that knows something.

“Not many people in AI are thinking like this because it’s so tempting to just give that answer. And in many sectors, giving a right answer immediately might mean saving a lot of time and money while allowing you to just go onto the next thing.”

He said there were already citation tools that give immediate answers: “You can just select a piece of text and say ‘create a citation’ and it just automatically adds one in at that point. There’s no process of the user making, writing or thinking on their own. We could make that too, but we don’t want to do it, because you’re not left with anything as a person. What is the purpose of a thesis? It’s not actually the thesis that, in most cases, is just going to go on a shelf. The purpose was the process of writing it and what you become, who you are after writing it.”

This approach has many positives from a business point of view. One is that Keenious isn’t competing with ChatGPT. “If we’re trying to be a better ChatGPT we’re going to lose, there are thousands of companies trying to do that.”

But there are ethical motives too: “Personally I want to be able to sleep at night. I want people to be better off using Keenious and to know I didn’t help plagiarism or help people cheat. When AI can answer every question you wonder what’s the point of getting an education. But why doesn’t a teacher give you the right answer right away? It’s because you want a teacher who helps you discover the answers for yourself. I’d be happy if the AI is enhancing education and not contributing to the breakdown of education.”

Angel librarian?

Frode says Librarians have played a vital role in making his start-up viable and keeping the sector attractive to new technology investment.

“We come under the Ed Tech umbrella and there aren’t so many investors in this space. It’s not like fintech – which is financial tech with banking – which is huge and has a lot of investors. Ed Tech has some very different characteristics: it may be more conservative, moving less fast, and there are a lot of decisions that potential customers (universities) have to make. These can be seen as dangers by investors.

“But if you get a good reputation and have happy customers, they are not afraid of sharing this with other similar universities. This is something that is very specific to this sector. It means our customers recommend us to others. In a competitive space you wouldn’t often get this help from your customers. Purchasers are hesitant because it’s such a big user base and they don’t want to buy it until they are sure. I’d say it’s tricky to get into this space, but if you build relationships and keep your customers happy then it has a lot of benefits. It’s a lot more sticky. It’s not as volatile. You can’t just be ‘fail fast’ which is normal in a lot of companies. But if you’re an investor who is not used to this you may be not aware of these benefits, and from the outside it can seem a bit slow.”

Library role in start-up journey

Frode explained how library buyers played a vital part by investing time and effort in his start-up, and then alleviating the risk for other potential investors by being transparent and supportive.

“The first organisation to take the risk with Keenious was The Research Council of Norway,” Frode says. “They had a great programme where you could get a million Norwegian Kroner (about €100,000) as a student just out of university, it was a very specific programme to start up something based on what you had been working on during your master’s thesis.”

Keenious got the funding after building a prototype and doing market research by “going up to students sitting at cafes at LSE and in California and asking ‘what do you think about this?’”

The funding enabled Frode and his Co-founder Anders Rapp to build a better prototype that got the attention of the University of Tromso, where Frode and his co-founder are alumni.

“The library said ‘this is really cool, but it’s not ready to be commercial, but we would love to help you guys out and give you feedback’, so we started a project with the university and the library there took us under its wing. We really worked hard with them to understand the need of the library and how we could make something that they want. The assumption was that university libraries across the world probably have the same challenges, so if it works for one it’s probably going to work for others.

“Then we got investments from mostly angel investors but also some bigger ones.”

Frode said the way the contract worked minimised risks for all parties, adding: “We had to get all three onboard, the universities with Innovation Norway (a national development bank) with the investors. They didn’t have to do anything until everyone was aligned, which reduced the risk.”

Library opportunities

“I think the reason they invested is because what we were doing clicked for them. They could see themselves wanting to use Keenious, or that they would have needed it when they were students. And I think that’s the investment side, that if you’re going into Ed Tech there are a lot of people that can relate to the problem that we are trying to solve.”

Frode said that because Keenious is selling into academia, it makes sense to use academic methods to develop it and sell it.

“We have been taking the academic approach and doing research. We have a dedicated User Researcher on the team, and we speak at conferences about what we find out. Selling Ed Tech to Higher Education, taking an academic approach shows our seriousness. We have a lot of customers doing independent studies on Keenious and its effects. We want people to test it out, write about it, and publish papers on it, that’s good for us.”

He said that Keenious still depends heavily on finding supporters in universities, often in their libraries. He said this was particularly true when start-ups move into new jurisdictions.

“Figuring out all the purchasing processes and rules in different countries is just something you have to painfully get through. The first customer in each country is super difficult. Many times it is the first time that they’re buying a Norwegian product – so the first customer is a big hurdle and they’ve got to really want to buy the product, they’ve got to be absolutely convinced to go through this uncertain process.

“For the first one you really have to find a champion, someone who really believes in the product and is willing to sell it internally, because they really want it and they have decision makers they need to convince. That’s relationship-building and just being honest about it and as a start-up we need that help since we are not a big company with thousands of sales people.”