We started Dialog because we recognized the role language models will play in the new vertical of customer research: automated interviews. While human-to-human interviews remain the gold standard, an expertly configured automated interview yields insight outcomes that far surpass what surveys or other traditional automated feedback tools can collect. So, our north star will always be conducting the highest quality interviews possible. The quality of the data collected sets the upper bounds on the quality of the analysis that can be done.
That's why we spend so much of our time focused on the engine that powers Dialog interviews. Building the Dialog Novalis 1.0 Interview Engine has been a long challenging road, but today we are excited to announce its 1.0 release!
On top of general interview quality, Novalis introduces some exciting new capabilities:
- In-conversation media support (with video coming in v1.1!)
- We can now support interview lengths of 45m+
- We have added deeper follow-up question logic
- Enabled finer-grained interview controls at topic level
- Upgraded the core NLP models to latest versions
A Picture is Worth a Thousand Words
Novalis 1.0 isn't just about text. Our new media integrations allow you to show product images, prototypes, or ads during interviews, capturing real-time reactions and insights that traditional methods miss. One of the largest challenges of traditional automated feedback (surveys, feedback components, etc) is to incorporate media. There's so much nuance with media, that trying to blend media into traditional automated feedback tools is nearly impossible. But as they say, the devil is in the details.
So, one of Novalis's most important features is in-conversation media support (video coming in v1.1!).
For a recent customer study we explored high level concepts in plain conversation first and then showed participants product screenshots to see if the product screenshots supported what the participants were wanting/envisioning. This ended up yielding fantastic results as we were able to explore the problem space around the concept to understand how participants reacted to the feature list before getting into the product screenshot granularities.
Better Real-time Response Evaluations
We are well aware of the quirkiness that can come from AI responses. It's something that the entire industry endures, and if you think otherwise I have a bananum to sell you.
So, part of Novalis feature set is deeper reasoning and real-time follow up logic that helps ensure that the follow up questions are relevant and situationally aware. Leveraging the speed increases of both our own architecture and the industry wide model speed increases, have allowed us to perform much more logic in parallel to the interview. We have positioned ourselves to scale interview quality as a function of model speed and quality as a result.
Novalis is topical
An interview protocol can take many different forms from a flexible, exploratory conversation to a strict list of questions. Each of these interview protocols has a unique consideration around what an optimal interview response looks like. Do you prioritize the flow of conversation or the interview's protocol? Do you press the participant for a deeper response or do you move to the next topic? These are a few of the many real-time considerations that need to be made during an interview. And Novalis handles that at a topic level.
A topics based approach allows us to bridge the big picture goals of the interview while delving to an appropriate depth for each of the key considerations of the interview. We wanted to provide a level of control that gave customers the ability to be very precise in what they want without introducing a proverbial footgun.
Please let us know how this topic based approach feels when setting up a Dialog.
What's coming in Novalis v1.X?
We are adding:
- Videos
- An improved interview quality scoring system
- New ways to set up your report
On top of interview related feature improvements, we are also focused on integrations. We have always seen Dialog as a workflow enrichment tool. What we mean is that Dialog is meant to bring critical data and insights where and when they are needed. The "where" part we are solving with integrations and that will be rolling out in the coming weeks.
Stay tuned!
A little bit about Dialog
Understanding your customers' needs takes valuable time and energy. Companies often invest massive resources into conducting this research, even at incredibly low volumes. Our mission is to make this valuable data more accessible to companies of all sizes and offerings