AI: Could Your Next Intern be a Chatbot?

By July 10, 2019June 4th, 2024No Comments

Picture this. You’ve got a task or a problem at work. You’re pretty sure someone has dealt with it before…but you don’t know who. Or where. Or when. It might be that weird guy sitting three metres away from you. Or it might have been that lady who quit working in the Abidjan office three years ago. Since you don’t know where to start asking and time is pressing, you do the whole trial-and-error thing from scratch.

If you work in any company larger than a startup, you probably know this situation well. Welcome to the wonderfully challenging world of Knowledge Management.

For all the clever technology and systems we use in our work, some crucial nugget of knowledge of how to get something done is usually stored in people. And there’s a major disconnect there, because people come and go from companies. People don’t know what each other are doing. People don’t always like to share information. Even when they write down information to help their colleagues, these colleagues forget that the document exists at the crucial moment, or can’t find it. In short, people make for a terrible filing system.

And in a world of dubious or overstated use cases for chatbots, here we find a compelling one. The idea of an artificial intelligence that doesn’t hide information from colleagues or leave the company, but rather gets wiser all the time and shares its wisdom on demand, appears to be a natural fit here. AI only really makes sense when it can negate human shortcomings in some way. And in theory this would do exactly that.

Developing such a chatbot is exactly what Pioneers ’19 finalist startup Zapiens have been doing since 2016. Building and selling such a thing is a huge task, but one the Spanish company is executing very well already.

“We’re not looking for money,” is the first unusual thing Co-Founder and Chief Business Officer Aurelio Jimenez says in an exclusive interview with Pioneers. “We now have around 30 customers, all of them companies with at least 50 employees, and this base is growing every month. This already finances our day to day operations.”

That leaves the company free to put its existing investment (around €1m) into making its ‘Zap’ chatbot as good as it can be. We’re talking real proprietary tech, says Jimenez.

“Many companies doing AI, analytics and so on are using engines from IBM, Microsoft, Google or whatever. We are building our own engine and our own IP.”

As with any machine learning, the thing with Zap is that he needs input before he can start being useful. The machine underlying him needs to suck as much knowledge as it can from the company’s employees.

“We tell them ‘congrats, you have hired a new intern! Now he needs to learn from you.’ We’re always working to improve the system itself, but every company has its own silo of knowledge that is empty when we start with them.”

Will busy employees really want to spend their valuable Instagram time answering work questions in order to train a chatbot? Thanks to elements of gamification in the Zapiens system and (sometimes) companies offering bonus incentives for participating employees, it appears that people do want to contribute.

“Lots of companies say, ‘We see what you want to improve here, but everything depends on people sharing what they know, instead of trying to lock the knowledge into them for competitive advantage.’ We have this conversation over and over with our clients. And a lot of it is about their culture.

“But companies have been surprised by the level of user acceptance.
People are more proud to share their knowledge than companies think – that is usually a surprise. And this is the best thing for our customer retention!”

Those customers are not small fry, by the way: think Audi, Deutsche Bahn, Unilever and Vodafone. And if you’re thinking ‘Ah, but those are just piffling pilot projects!’ then Jimenez would have you thinking again.

“This year we changed the strategy: we don’t accept Proofs of Concepts any more. Because what we do only makes sense with a long term view. We will define our work with any new client as a long-term project, but of course we won’t lock them into Zapiens if they aren’t happy.”

Might the startup from Oviedo just be on their way to major success here? Big companies are notably terrible at even knowing how bad they are at Knowledge Management (or hiring people, as The Economist points out here) and many are only just starting to wake up to the preposterous inefficiencies going on in their midst. There are vast corporations in which nobody has the job of aligning all the in-house knowledge.

“We won’t start working with a company that doesn’t believe in what we are trying to achieve for them,” says Jimenez. “If they actually have someone in the role of Talent Manager or Knowledge Manager, this is a KPI for us that says this company is thinking about the things we do. And every day more and more companies have this role!

“The bigger a company is, the more difficult it is for them to update and manage their knowledge. The more rotation they have in products, services or people, the more important something like this becomes.”

So far the Zapiens system has found most application when it comes to product, internal or sales processes. ‘What’s the price of this? Where can I get this? How should I do this? How did you solve that problem?’ These are the kinds of questions a well-trained Zap is designed to answer.

If you’re wondering whether ‘Is there a document about this’ and ‘What document should I read about that?’ are also on the list, then the answer is yes – as long as an employee has shared that information. The system does not, in other words, scan through your folder tree and its files in the looking for search words (such hit-and-miss systems already exist). It targets only human knowledge – in effect it asks your question of choice to everyone in the company without you having to leave your desk. And if the bot has been well-trained by your past and present colleagues, then it will already have the answer for you – or at least tell you whom to ask.

Ambitious? Hugely. Have we tried it? Sadly not yet. What we can say here is that the early traction with clients suggests Zapiens is addressing a problem companies really want to solve. Horizon 2020 grants and a current participation in the Alchemist Accelerator in the USA suggest we’re not the only ones to see it. Above all, the company has a genuine confidence about it – they seem to know where they’re heading. We trust the destination is a good place!

For the full story of Pioneers ’19 click here.