It’s almost a year since we launched Atlas, our AI coach. This week we are revisiting our commitment to providing regular updates on its outcomes.
We’re on a mission to provide equitable access to economic opportunity for everyone. Here, we want to set out how Atlas is helping us to do that, because when combined with our industry-leading human coaches, it allows us to offer 1:1 tutoring - which is proven to dramatically improve learning outcomes - at scale. Together our coaches combine with Atlas to deliver highly personalized, effective, guidance 24/7, whenever our learners need it
We also want to explore what we’re doing to improve - because no good tech organization only celebrates its wins. We’re constantly iterating and learning ourselves. We'll touch on what’s new and what's next for Atlas at the end.
We often say that AI is useless unless people can use it. And at Multiverse, we also want to understand how people are using it. So let’s dive in.
Equitable access
First, let’s look at users - and what it could tell us about AI adoption.
We’ve seen tremendous growth in usage, with a roughly four-fold increase in the number of daily users in the past six months. Likewise, the number of messages per day has skyrocketed - we’re now seeing 5x the volume compared with six months ago. We also hit a new peak ‘helpfulness’ rating of 97% in January - which has remained consistently above 95% since August 2024.
But who are those users? We previously reported greater uptake among the over 40s and those with additional learning needs. That still holds true, with 46% of over 40s using it vs. 42% of the 25-39s and 34% of those 24 and under. This stands counter to the trend illustrated in some of the latest UK government AI usage data(opens new window), which suggests that 18-34 year olds are most likely to use AI tools like LLM-based chatbots.
Meanwhile we have proportionally equal of male and female usage from the past six months, with 42% of each group using Atlas. Again, this bucks the trend of AI usage skewing more male, according to that same government data(opens new window).
As for the racial demographic of our users, it’s broadly similar across ethnicities, with 40% Asian, 35% Black, 42% Multi-Racial, and 44% White learners using the tool. These proportions have grown steadily among each group, though adoption among Black learners remains slightly lower than average.
Economic opportunity
I also want to explain a little bit about outcomes, and in particular how people are using the tool. We’ve found that some of the top ways that Atlas is being used include:
- As a learning partner: helping learners to unpick how they’ve arrived at incorrect answers within their own work.
- As a curriculum guide: augmenting program material with advice for how to apply it to their particular role, or explaining specific concepts in a new way.
- As an impact evaluator: helping to uncover the impact of a user’s learning, by teasing out insights on what they’ve achieved, and helping calculate outcomes
While we always intended for Atlas to act as a technical and learning guide, we hadn’t anticipated the popularity of using Atlas to understand impact.
This hints at how Atlas could be used to improve individuals’ economic outcomes. What better way to talk to your manager about a promotion or a pay rise, than armed with data and insights on the impact you’ve been able to drive for your team? Could AI be your next career coach?
This also has perhaps even greater implications. It hints at how organizations can start to use AI tools to understand the impact of their wider workforce. We’re really just at the beginning of realizing the full potential for this tool.
Public vs private sector uptake
Of course, not all customers or learners are the same.
Working closely with more than 1,500 partners from every sector means we understand the unique differences in how different organizations learn and apply their skills. As you can imagine, the needs of a financial services company are vastly different from the needs of an NHS trust. The same holds true for what they require from Atlas - naturally, applying skills in a healthcare or local government setting differs from a commercial organization.
To investigate this, we compared behaviours between the public and private sectors. And found something remarkable.
There is an unfair cliché that in the public sector, technology is treated with suspicion or underutilized - that public sector organizations invest into technology without having the requisite skills to make it work.
According to our data, though, that assumption isn’t borne out. Comparing a sample of our biggest private sector customers with over 80 NHS and 50 local council partners, we found that proportionally more people are using Atlas in the public sector (66% vs 57%). This suggests that in the right environment, public sector workers are just as curious about and capable of using AI tools as their private sector counterparts. We just need to enable them in the right way.
What’s more, we know that public sector organizations are under pressure to deliver more with greater efficiency. AI has a huge potential to help with that, and Atlas is just the beginning.
What’s new and next for Atlas?
Finally, what can we do better?
We’re continually iterating to make Atlas more user-friendly and useful to the learners on our programs. Recent areas of exploration for us have included:
- Atlas as a learning co-pilot – we’ve made changes to Atlas to make it more contextually-aware, responsive and discoverable within the Multiverse platform. So Atlas now recognises the content learners are looking at whilst they are learning on the Multiverse platform. This means it can offer more relevant guidance, as well as displaying dynamic prompts based on what learners might want to know more about.
- AI-human handover - we know there are some circumstances where only chatting to a human will help. Atlas can now help smooth that handover by summarising a learner’s interaction when it hands the conversation over to a human coach. This makes it easier for the coaches to provide timely and thoughtful responses rather than spending a lot of time catching up on the chat.
Looking ahead, we want to embed Atlas more fully into the learner experience throughout their program, and find new ways to make it an even more helpful and intuitive learning partner.
And as always, we won’t take ourselves too seriously. You may have spotted SantaAtlas – you can bet there’ll be more surprises up our sleeve when it comes to Atlas’s sense of personal style. Any suggestions? Let us know…(opens new window)
