Adapting chatbots to handle COVID-19
How we helped the Australian Tax Office support millions of Australians.
Since the beginning of 2020, the world went into lockdown. The Australian Tax Office became suddenly responsible with providing millions of Australians with tax relief, tax information and income support.
As a result, the ATO were inundated with phone, email and live chat requests from Australians with complex support issues.
How might we use the ATO’s existing infrastructure to support their contact center
We worked closely with the ATO to improve the effectiveness of their chatbot, Alex, by improving both the way you interact with the chatbot, and the way it interacts with you.
Diagnosing the problem
Through our research (usability testing, heuristic review, analytics review) we identified the following challenges:
- Low Engagement
While the Alex is proactively opening on high traffic pages, it has a low engagement rate of only 5% as the upfront information presented wasn’t seen as useful.
2. Confusing responses
The content provided by Alex is structured haphazardly and doesn’t present you with a next best action.
3. Subpar interaction design
You can only interact with the chatbot by typing in the text entry field, leading to choice paralysis among customers not knowing what to type.
4. Text heavy
The chatbot relies heavily on long text responses, with little to no visual aids to support the content.
Recommendations
In line with our views on best-in-class chatbot experiences, we were driven to apply what we knew to help improve Alex.
Overall we delivered over 40 recommendations for improvements to the interface, website and conversations — the following is a summary of the major improvements we provided, and learnings for your own chatbots.
- Give customers more ways to interact.
Giving you the flexibility to type to the chatbot, or choose from suggested options allows us to interact with the bot in a way that feels natural to them. It also presents “next best actions” allowing a conversation to keep moving in a natural flow and introducing follow up questions.
2. Speak to people in a natural way
One of our main challenges was helping the ATO break down complicated Tax information into something that can be easily understood by anyone.
Some of the guiding principles we employed were:
- Clarifying the intent — making sure that we’re going to deliver you a specific response rather than something generic by offering a follow up question.
- Directly answering the question — if there’s a short answer, make sure that it’s at the beginning of your response.
- Explain like I’m 5 — if there’s a technical term that’s been used, or a concept that can be interpreted multiple ways, it’s best to follow it up with an example or definition.
- Keep your terms and conditions off the chatbot — by this we mean, if there’s any extra information that isn’t mission critical, keep it in your website content and provide a link to learn more, rather than pushing it all through the chatbot.
- Capture feedback — for us, good work is not set and forget. It’s a constant cycle of delivering, testing and adapting that allows us to improve what we’ve made.
We created a new set of conversation flows specific to COVID-19 content, and introduced post-conversation flows to track effectiveness.
If you’re interested in learning more about what makes a great chatbot, check out our piece below on 5 pillars that make a chatbot best-in-class.
