Interview with Erik Riddervold Sandberg, CEO of Hubert.ai

Erik Riddervold Sandberg is the CEO of Hubert.ai, a company that uses Artificial Intelligence to improve the teacher feedback process and eliminate the time spent on tedious surveys.

In this interview, Mr. Riddervold Sandberg shares insights on how teaching functions such as grading and giving/receiving feedback to and from the students can be drastically improved with AI.

This interview has been featured in the AI for Education Initiative 2019.


What is your background and how was Hubert.ai started?

I have a Bachelor in Mechanical Engineering and a Master in Industrial Management and Innovation, combined with half a year at The School of Entrepreneurship at Uppsala University, Sweden.

Hubert.ai is the result of an earlier idea my co-founder Viktor Nordmark had, which was to provide clever incentives to make people fill out surveys. We were exploring this together, and still studying at the university, it was natural for us to start out with course evaluations to validate the concept. When we had conducted a few tests at the university and were in the middle of building our incentive system, we realized that it’s not the lack of incentive who’s to blame for low response rates and quality of survey results. People really want to share their opinions (just look at online forums or the comment sections on Youtube), but a survey can feel tiring and even limiting due to it’s static, mathematical nature (it’s very hard to discover new improvement areas from 1-5 scale answers in a matrix). Turns out many students even “Christmas tree” (ticking random boxes) their way through questions they don’t care about.

We set out to change this and make it possible for everyone to make their voice heard, highlighting what they think is important instead of forcing opinions that risque skewing the feedback. After talking to educators at the university and doing a good amount of research we learned that the best way to do so is to have one-on-one conversations with everyone in a class. With any class larger than 20 students, that quickly becomes unrealistic given the already heavy workload our educators have. That was when we realized that we could use the new AI-based chatbot technology to help the educator talking to everyone, and Hubert was born.

What is Hubert.ai’s mission?

His mission is to put an end to all surveys so that your voice can be heard.

Who are Hubert.ai’s customers and how do you create value for them? What challenges do they face, or will they face in the future, that Hubert.ai can help them overcome?

Our main customers are organizations, both universities and private companies, that want to improve their understanding of their students, employees or customers so that they can provide even better services and products. The common denominator is that they don’t get good enough data through traditional surveys and that their respondents are tired of filling them out. This “survey fatigue” is becoming more prominent with every day that passes, in virtually every audience. They want a more natural, relaxed and fun way to give feedback and our customers don’t want to go through results chart looking for what’s important. That’s when Hubert is making their life easier, as in this case study with Siemens (PDF): https://drive.google.com/uc?export=view&id=1eZw0jqa8TYisyvYakLtDIbfGgpKp4qwK.

How is AI leveraged in your product? 

Almost everything in Hubert, from how he is behaving to how he presents the insights, is based on several layers of AI. As an example: first he has to decide what type of answer he is receiving, then decide if it contains useful information and, if it doesn’t, come up with a clever follow-up, to last but not least compare it to every other answer he has received to be able to tell how important it is in its context. These are also the key things that make Hubert unique; no other feedback service today is able to ask good, relevant follow-up questions to answers who don’t contain information that is useful to the feedback receiver.

How do you measure the performance of your technology?

We have several numerical performance indicators in different parts of our system (response rates, completion rates, answer quality, etc.), but the most important ones are actually what our users are saying about interacting with Hubert and the usefulness of his findings. At the end of every chat with a new person, Hubert asks about the experience of chatting with him compared to filling out a traditional survey. We combine the answers to that question with our customers experience of the usefulness of the information they’re getting, and this is our way of measuring how well Hubert performs.

We also have a numeric performance indicator that really showcases the effectiveness of collecting feedback in this way; in a normal survey you may have text fields that are required before the respondent can move on to the next question, which initially feels like a clever way of “forcing” the respondent to provide qualitative, written feedback. However, we’ve found out that these fields often contain expressions like “.”, “I don’t know”, “everything”, and “nothing”, which the feedback receiver usually finds too uninformative to derive actions from. Hubert solves this issue by asking follow-ups like “Surely there must be something that could be improved with X?”, and in most cases, he receives good, informative answers to those questions. In comparison with reference surveys, Hubert has been able to reduce these uninformative answers during his conversations by over 90%, turning them into actionable feedback.

Tell us about the Hubert.ai team.

Team Hubert is still at a startup-stage with around 10 employees right now, but we’re rapidly growing and will be moving to a new office this fall. We’re hanging out in the university town of Uppsala, Sweden, and if you drop by you’ll meet a variety of Swedish, German, Korean, Taiwanese and Indian persons happily nerding away with the latest AI-technologies and NLP-research combined with as weird tasks as how and when to use a particular smiley. We’ve been developing Hubert since three years back and most of us have studied at Uppsala University. We’re lucky to have one of Swedens best investors as a co-founder and investor, so that we also have expertise and experience with building companies from scratch.

Join our weekly newsletter to receive:

  1. Latest articles & interviews
  2. AI events: updates, free passes and discount codes
  3. Opportunities to join AI Time Journal initiatives

What are the major challenges in the educational system today? How can AI technology help solve or mitigate them?

I think that the largest challenges are steadily increasing class sizes and that educators are getting less and less time to talk to their students one-on-one. Some educators never even meet their students due to whole courses being taken online. It is a good thing that we are educating more people, don’t get me wrong, but it brings challenges where an educator needs to spend time with each student in one way or another. The two largest areas where I see an urgent need for AI-assistance is in grading/giving feedback on assignments, and in connecting and collecting feedback from the students. Hubert significantly increases engagement while reducing the time needed for the latter.

What are the major opportunities brought by AI in Education today?

For all areas where an educator can be relieved in performing time-consuming and repetitive tasks, there is an opportunity for AI to help. Tasks such as grading, giving feedback and checking attendance are being heavily researched by AI-companies today, but I think we’re still just sneak peeking through the keyhole to the door that AI can help us open in both the educational and corporate world.

How can teachers prepare for an AI-powered education?

I think that they need to get used to the thought of being assisted in different areas and start trusting that AI’s are doing some parts of their jobs good enough for them so that they can focus on the activities that are creating the most value for their students and themselves. There is no need to be afraid that AI is going to replace educators in a foreseeable future, so I encourage all educators to relax and let the AI take care of the stuff that they feel is tiring and time-consuming. Be open to trying out the new AI services that are already out there and I assure you that you’ll be positively surprised, and probably save a lot of time.

Which AI technology do you think will have the biggest impact in Education in the coming years, and why?

We are already seeing a lot of Adaptive Learning Environments, and I think that is important to make sure that students are learning as effectively as possible. However, where I see the largest potential and research activity is in using Deep Learning and Natural Language Processing. It has ha huge potential in helping educators give feedback on assignments, grade exams and ensure originality, as well as answering questions (see Jill Watson) and identifying who may need special attention.

How should educational system management software adapt for the advent of AI?

Prepare for API connections, make sure that a lot of data is readily available for AI plugins and make it easy for established companies and startups to connect and set up environments to help push the technology forward. Give educators the possibility to set up their own small testing environments with simple integrations on an individual level so that they can evaluate AI software without having to engage administration instances. We’ve been faced with a lot of obstacles in connecting to different LMS systems, some of which we still haven’t found a way in to, and it is a big time-killer for everyone who is trying to improve education. Having said that, some LMS systems and educational software are doing an awesome job in this.

Do you see threats of inequality in the access to AI-powered education?

Of course. Access to technological equipment that is up-to-date and the internet is still varying a lot around the globe and will limit the access to new AI services, but that’s an obvious limiter who is becoming less and less of an obstacle every day. What will make AI-powered education a global win is that it can be a huge help for those who have very limited resources, e.g. can an AI assistant help handling more students with fewer educators and therefore make it possible to provide better education to more people to a lower cost. With e.g. SpaceX’s Starlink project and other efforts to provide internet to every corner of our planet, all you’ll need is a computer, tablet or mobile phone to have access to AI and deliver world-class education.

What are the current technological limitations of AI which, once overcome, could bring massive improvements to the educational system?

I think that Natural Language Processing and Emotional Intelligence are still in their cradles, but once they become more mature AI will be able to help us out in a lot broader sense. Right now, most AI’s is still somewhat stiff and un-emotional (not Hubert of course, he’s great), and is, therefore, more suited for specific tasks within math, natural sciences, and fact-heavy subjects, but future AI’s will have a deeper human intuition and therefore be able to broaden their use a lot. That’s where we’re going with Hubert, and it would be great to see even more AI startups going in this direction.

Where do you see AI for education in the next 5-10 years?

In 10 years, AI will be an as natural a part of education as online learning platforms are today. You will write an exam, have it instantly graded by an AI, ask questions about it to your bot assistant and get a tailored learning plan from your AI education coach that knows what subjects will be most important for you to learn given your interests, personality, and future job opportunities.

What’s in the near future for Hubert.ai? 

Very soon we’ll be releasing a completely new version of Hubert, where we have taught him to collect feedback in both Education, Customer Experience, and Human Resources, with a lot more customizability. He’ll also be tri-lingual, speaking English, German and Swedish. Keep an eye out at hubert.ai, where you can also sign up for our newsletter to get notified about the release, and if you’re interested in reading more about how Hubert is disrupting the world of feedback, read our blog at blog.hubert.ai.

About AI Time Journal Editorial Staff

The mission of AI Time Journal is to divulge information and knowledge about Artificial Intelligence, the changes that are coming and new opportunities to use AI technology to benefit humanity.

View all posts by AI Time Journal Editorial Staff →