AI case study: Powerswitch

When users enter their individual power usage data into the Powerswitch platform, they are twice as likely to make the switch, because the results are much more tailored to their circumstances.

Better data = better comparisons, more compelling information and, therefore, more switches.

Until recently, Powerswitch relied on consumers entering data from their bill – but many were unwilling or unable to do so.

We explored with the Powerswitch team the possibility of building a service that could read power bills, but there were technical challenges that made it impractical, including the fact that New Zealand power bills do not adhere to a standardised format (i.e. they all look different).

Availability of AI tools unlocked the opportunity to deliver a bill reader that could accurately read and understand consumers' power bills. Our key priorities for implementing this service were:

1. Maintaining a user experience that felt familiar – not tacked on.
2. Maintaining privacy and processing the user information securely.
3. Making it easy and seamless for users!

While still technically in a beta release (final draft, for non-technical people!), the results have been really strong.

Our beta AI bill reader has now been live for a month, and has exceeded our expectations. Despite minimal promotion, we’ve seen over 20,000 successful bill uploads in its first month alone! This has driven a steep change in site engagement. – Paul Fuge, Powerswitch Manager.

Over June, Powerswitch recorded around 9,000 initiated switches directly from the site. This was the highest number recorded in a single month and double the previous record.

The upward trend has continued, with 9,706 switches initiated in July, a 170% increase from the same month last year. These figures highlight the impact of the more user-friendly experience and more tailored comparisons the AI-powered bill reader provides.

Consumers using the bill-upload feature can easily provide more data than those who manually enter their bill information, which is building a bigger overall picture of power companies’ pricing, which in turn leads to improved recommendations and intelligence from the Powerswitch platform.

For a tool that has been in use for only a couple of months, and with minimal publicity, these results are awesome.

With so much interest in AI at the moment, it's worth sharing a couple of really basic reflections on the success of this project.

First, it demonstrates the ability for AI tools to unlock significant benefits when focussed on the right use cases. There's a lot of things that AI can do but, in this case, the Powerswitch team's ability to identify the right opportunity, based on a deep understanding of their audiences, has been critical.

Second, it has confirmed for us the importance of the 'last mile' of AI implementation. A notable feature of AI is that it appears to do a lot straight out of the box, which is (quite rightly) driving a lot of enthusiasm around its potential uses, but we think it can also be a little misleading.

Actually going live with AI tools and services that are safe to use and integrate effectively with existing systems will often require real technical expertise and care. This is especially true when connecting up existing (and often legacy) data or other services, which is where a lot of the untapped potential for AI still lies.

Check out the new Powerswitch bill reader