On EQT Ventures’ investment in CallDesk and our approach to AI companies

Talking to customer support is often a last resort that most of us want to avoid at all costs. Long queues with horrible music, badly configured IVR systems that don’t understand your requests. And the constant agony of being reshuffled between agents that don’t have any history of your prior discussions, while in the end being disconnected and having to repeat the process all over again.

EQT Ventures recently announced its investment in CallDesk. A very exciting company that aims to solve the above issue, using AI done right.

Advancements in AI

A lot of IF cases does not equal AI.

At EQT Ventures, we’re very excited about the advancements in AI and firmly believe it will transform most industries in the next 5–10 years, in ways many can’t really foresee at this point. With that said, there’s also a lot of hype in this space. Pitch decks are suddenly filled with AI and machine learning slides, but when you start scratching the surface you very often realise that there’s no real or very little substance backing up the claims. A lot of IF cases does not equal AI. On the other hand, you have some very skilled and knowledgeable teams that are building groundbreaking tech. But unfortunately, in many cases, those services are solutions looking for problems.

Horizontal vs vertical AI

We like to break down AI companies into two different approaches, horizontal and vertical. Horizontal solutions are more general and solve a broader use case while a vertical is much more specific and tailored. If you look at the investments we have done to date most of them are in the latter category with Unomaly being a great example. It is full stack solutions that begin with proprietary data, models and ends with a usable end product that customers can plug directly into. There are a few main reasons why we like those businesses. To begin with, they require a founding team with deep domain expertise. Secondly, they almost always rely on proprietary data from the customers. This data improves the product over time and creates a strong moat that makes it harder for competitors to catch up.

“we have a lot of data”

Data, with emphasis on labeled data, is the key in those kinds of solutions. On a regular basis you hear from companies, mainly incumbents, “we have a lot of data”. This usually ends in disappointment, because they don’t really understand what a lot of data means. And even if they happen to have the data, it is rarely labelled nor structured.

If you look at horizontal solutions they are more generic and not full stack solutions. It could be, for example speech recognition or text understanding. Those problems require the best algorithms, a big amount of labelled training data and a lot of compute power. Here the big five (Google, Facebook, Amazon, Microsoft and Baidu) naturally have a clear advantage.

However, if you manage to build a horizontal solution better than the companies mentioned above, the upside is obviously very large.

CallDesk, a great example a vertical AI company

So what made CallDesk stand out amongst other interesting vertical AI companies?

First of all, the team has successfully developed algorithms that understand consumers’ intent and dialogues in real-time. Vincent Gire, the company’s co-founder and CEO, spent many years researching and solving this problem at his earlier company, which he sold to Solocal Group. A repeat entrepreneur, with strong domain knowledge and the technical skills to build the first MVP himself is obviously a huge plus in our book.

Secondly, the CallDesk team is solving a huge pain point as described in the introduction of this text. Everyone has a bad experience with call centers and, with more and more purchases and services moving online, this problem will just grow over time. It’s also a huge challenge for the businesses where customer satisfaction is such an important measurement. Bottomline this means those companies can increase NPS while keeping cost down, a typical win win situation.

Thirdly, taking a research project to production at scale is very hard. It begins with finding labeled data to use, how to ingest it and clean it up, to collaboratively work on models, and version control the models to finally integrate this into the customer’s system. You then have to be able to do all this at scale in real-time. Not only has the CallDesk team managed to do this, they have also managed to build a system that’s language agnostic, with a very small team of engineers.

Last but not least, we really believe voice and a more human interaction will be key in customer support going forward. It stands for the lion’s share of support channels today and it will just go up with a service like CallDesk. In terms of data backing up this argument, online search vocally for example stands for 20% today and is expected to grow to 50% in the next couple of years.

Want to make customer support great again? Join CallDesk today!