Guangzhou skyline

The story behind AutoMagic (and the most dramatic change in SaySomethingin so far)

In 2019, four of us went to visit a company in Guangzhou in the south of China that we had been told might be interested in an English course for Mandarin speakers. We had a fascinating time, and we really liked some of the people in the company that we met – but they told us, in the most polite terms possible, that they wouldn’t be interested in working with us if our course was based on static mp3 files. ‘It would be too easy to copy them,’ they said, and then, with a wink, ‘In fact, it would make more sense for us just to copy them ourselves!’

They were interested in testing our approach, though, if we could move the method into a streaming environment that would be more difficult to copy. By the time we were sitting in Hong Kong airport for the flight home, we knew there was a serious challenge in front of us. 

It’s a long old flight from Hong Kong to Caernarfon (and it’s not even direct yet!). I remember sitting on that Emirates plane and thinking about how streaming could work. It was 10 years since we’d shared our first few Welsh lessons, and those early 100s of users had become thousands and then tens of thousands. 

How wise was it to change what we were doing?

But the more I thought about it, the more I started to realise something exciting. Moving into streaming wasn’t a gamble – it was a way to make enormously valuable improvements to our entire approach. I started thinking about all the reasons people didn’t carry on with SaySomethinginWelsh, and I began to see ways to solve the underlying problems in a streaming environment.

Then it dawned on me that the way we would be able to monitor progress in a streaming environment would map perfectly to the kind of data collection that allows for AI-driven improvements – and the seed of the idea for our AutoMagic Tutor had been planted.

It’s taken us 3 years to get to a working implementation, and we’ve still got work to do. As you might remember, early in 2020 we all found ourselves in a global pandemic, and then in various flavours of lockdown. It was difficult for everyone: challenging for mental health, challenging for communications, challenging for working patterns.

But we kept going.

And as of now, our AutoMagic Tutor can already allow learners to go at different speeds (which has enormous implications for how many people are successful with the method) and to choose how long they spend learning – and the basis is there for increasing personalisation and gamification over the coming years. 

We’ve been test driving the AutoMagic Tutor with the National Centre for Learning Welsh, and with our Welsh government-backed schools pilot project, and with our work on the current series of the S4C show ‘Iaith ar Daith’. The results so far are genuinely exciting – I’ve seen people struggle with the static files and then achieve huge success with the AutoMagic Tutor. I’m confident it’s going to let us help tens of thousands more people to become confident Welsh speakers, and I can’t wait to see it being used in more schools.

And yes, we’re still in touch with that company in Guangzhou. We’ve been testing our approach to English with a number of projects in Sri Lanka (including some enormously worthwhile charity projects) and the results so far are very promising. We’re hoping to test an AutoMagic Tutor for English for Mandarin speakers in Shanghai before the end of this year, and the first company we spoke to are going to look at the results with a view to scaling if the trials go well.

Sometimes, persistence doesn’t just get you through the difficult times, it opens up entirely new possibilities. Entirely new ways to try and help make the world a little better, which is what we’re lucky enough to be trying to do.

Aran Jones