Product

Behind the scenes of Helix with Rob Mackenzie

Behind the scenes of Helix with Rob Mackenzie
Behind the scenes of Helix with Rob Mackenzie
6:29

Now the cat is finally out of the bag about Helix, our new AI skills intelligence solution, we wanted to give you a peek behind the curtain to find out how it’s all come together over the last few months.

The design and build of Helix has been a real team effort, featuring input from over 100 learning and talent leaders and a lot of rapid upskilling from 5app’s developers. As 5app’s Head of Technology, the journey to Helix hasn’t always been easy, but we’re all very pleased with the product you see today!

So, let’s get stuck in and take a backstage tour around Helix!

Multi-mock up 2 (2)

 

It all starts with an idea…

We at 5app have long recognised the challenge of ensuring lasting behavioural change after workplace training. Typically, someone will attend a workshop or complete an elearning module, and they might try to implement what they learned for a week or so, but before long they’ll slip back into their old habits and forget everything from the training.

But with the rapid adoption of AI, we saw an opportunity to take advantage of the new technological capability to analyse real behaviour change in the flow of work. We knew that L&D professionals were missing an efficient way to track and measure skills in real time, and we knew that with 5app’s small-but-mighty team, we had a way to make it happen.

 

Exploring our options

Traditional assessments and quizzes can be disruptive and unrealistic. They present learners with a list of questions, which may be answered accurately in the moment, but they don’t show that the knowledge, skills or behaviours have carried through into real day-to-day work. We knew L&D teams needed a less intrusive, distracting way to continuously track real skills in the real workplace – and that meant looking at real work activities and channels.

We looked at a variety of data sources, including:

  • Emails

  • Instant messages

  • Online meetings

  • Face-to-face conversations

While Helix gives us plenty of scope to expand in the future, it quickly became clear that online meetings were the best place to start. The rise of remote and hybrid work means that most of us have at least a few online meetings a week, giving us plenty of rich opportunities to track skills, and the widespread use of AI notetakers means that we’re used to seeing AI bots appearing in our calls without it feeling too intrusive or Big Brother-esque.

image (7)

The basic premise is that a Helix bot attends and records your meeting in order to create a transcript. AI then analyses meeting contributions against our soft skills framework and proprietary analysis engine spearheaded by our Head of Product James Cranwell.

 

Designing robust skills analysis

Multi-mock up 2

Collecting the data is the easy part! While it’s easy to use tools like ChatGPT to get a basic analysis of a transcript, the challenge lies in obtaining high-quality, consistent results which demonstrate growth over time.

That’s why we developed our own proprietary 5app 3Q Engine, which stands for qualitative, quantitative and quotient. It’s essentially the secret recipe that powers Helix, and it’s what allows us to measure soft skills against a tried-and-tested skills framework which is free from bias, as we set the parameters ourselves.

Designing the 5app 3Q Engine was a challenge in itself! We built our methodology on ChatGPT, which required a huge amount of research and iteration around prompt engineering for efficiently handling large amounts of data, as well as a lot of trial and error.

With a lot of hard work (and some frustrations along the way), we finally struck gold with our solution, and we were able to start refining the way the software analyses transcripts for more insightful results.

 

Testing, 1 2 3…

Multi-mock up 2 (1)

Of course, a huge amount of testing has gone into the Helix design process. This is a brand-new product, built on fast-changing technology by a lean team. 

We first used a set of manually scored transcripts to compare against the model’s output, allowing us to validate the quality of Helix’s analysis. Once we were happy with this, we asked the 5app team to start inviting Helix to our internal meetings. Based on the results and feedback from the team, we refined the prompts and skills summary emails from Helix. For instance, for a short period, the team was receiving scores of 8712% or 9254% for their skills – we’re a talented group, but we’re not quite that talented!

 

Small teams can do big things

It’s no secret that 5app is a small team. We may not have hundreds of employees, but what we do have is a small, committed core of extremely talented people who make magic happen.

We went from 0-100mph with our understanding of AI. There was a bit of online training, but mostly we relied on peer-to-peer learning and trial and error. We also worked with brilliant data scientists in our network, who provided us with valuable insights on cost optimisation and building our admin dashboards.

 

What’s next for Helix?

I don’t want to give away all of our secrets, but rest assured that we’re already thinking about how we can evolve Helix to keep up with the rapid pace of skills today.

It goes without saying that we’ll be increasing the number of skills available to track in the coming months, adding to the original Core 9, and we’ll be testing them all on the 5app team first to make sure they’re in the best possible shape for our customers. The beauty of Helix is that it’s highly scalable, so we should be able to expand the skills available for analysis relatively quickly based on customer demand.

We’re also interested in exploring some of the other potential data sources, such as recordings of in-person meetings or analysing written communications, as well as looking at offering Helix in other languages. 

But there’s so much we want to do with Helix! Version 2 will include linking suggested development areas to specific learning content assets and coaching tools (like VeeCoach), integrating with other LMS providers, creating team manager dashboards, building more detailed admin analytics and more. User feedback will play a crucial role in the development of Helix, so don’t be shy about sharing your ideas, thoughts and suggestions!

See Helix in action for the first time!

Want to see how Helix works? Join CEO Philip Huthwaite on 17th July for the very first public reveal of our new AI skills intelligence platform.

SECURE YOUR SPOT



Similar posts