September 10, 2018 5 things copywriters need to know about AI & conversation design

By Natalie Khoo

Dr Phil Cohen on stage at Summit

Anyone I’ve met knows I’m a geek for emerging technology. In particular, I’m very passionate about artificial intelligence (AI), and how natural language processing (NLP) will impact content strategists and copywriters.

With Avion’s experience writing for voice technology, chatbots and robots, I was incredibly excited to see what new stuff I’d learn at the 2018 Digital AI Summit, a headline event at this year’s Victorian Digital Innovation Festival. Here are 5 takeaways that every copywriter interested in conversation design should know.

1. Bots: The basics

If you’re yet to work on a chatbot project, it would help to know there are several types of bots. Each has its own purpose and human-bot interaction style. Agnes Panosian, AI and Data Lead at Microsoft, places bots in these 4 categories:

  1. Bots that work like IVR phone systems (i.e. to help you find a membership number)
  2. Transactional bots that help you complete a task (i.e. make a booking)
  3. Bots for advisory services (i.e. to ask what’s the best insurance policy for you)
  4. Conversational bots (i.e. like Siri from Apple, Xiaoice from Microsoft, or the scarily-realistic yet fictional Samantha from the 2013 movie ‘Her’ starring Joaquin Phoenix and Scarlett Johansson).

If your organisation or client would like to develop a bot, consider the problem you’re looking to solve; it should inform what type of bot is best before you start your project.

2. “People don’t give values, they give constraints”

Dr Phil Cohen, Professor at the Laboratory for Dialogue Research at Monash University, explained that we’re more likely to say, “any time after 2pm” as opposed to just “2pm”. Unfortunately, this poses challenges for AI. In such cases, it needs the ability to conduct research and make decisions, i.e. determine what times are available after 2pm, then offer suggestions.

This concept is what we call “plan recognition”. Although AI still has a long way to go in this regard, it demonstrates why copywriters should work closely with developers. As specialists in content and conversation design, we should know exactly how the AI will be built (e.g. are interactions based on machine learning or a decision tree?) and strategically suggest content that works within those parameters – not just responds to instructions.

Currently, Google Duplex has the incredible ability to decipher intent better than most. If you say “I’d like a table for 6”, it can usually tell whether your intention is “I’d like a table for 6pm” or “I’d like a table for 6 people”. Amazing, isn’t it?! If you haven’t come across it before, check it out in action below.

3. “Contextual awareness is really hard to unpack”

As stated above, AI has its limitations when it comes to functioning beyond specific tasks. What can we do about this? The key is creating a single NLP model that can handle multiple tasks, says Rob Wickham, Regional VP, Platform & Emerging Technologies at Salesforce Asia Pacific. That will get us one step closer to “general purpose AI capability”.

The Salesforce team has been working on DecaNLP, a model that processes 10 tasks simultaneously. To see how it works, check out this animation below that shows how context is cleverly and quickly reviewed before an outcome is delivered (image credit: Salesforce). As a copywriter, it helps to consider what information you’re providing to programmers, so they can fulfil this step in the build.

DecaNLP_Fig1

4. 20% of Gmail users are using quick response buttons

According to Dr Nathan Faggian, Cloud Consultant at Google, 1 in 5 Gmail users are using its auto-generated quick response buttons to send emails. That’s pretty impressive, considering Gmail has more than one billion users worldwide. It proves that AI can, in many cases, predict what we’re likely to say next.

Given this capability is a direct outcome of machine learning, there’s not much a copywriter can do. However, it’s good to understand the relationship between access to large amounts of data, the ability to identify patterns, and a positive user experience.Gmail quick response buttons

5. (Some) copywriters will be replaced by robots

If you’re a copywriter that writes product descriptions for a living, watch out. AI now has the ability to scan thousands of images at scale and automatically detect/describe product features, such as ‘round neckline’, ‘short sleeve’, ‘medium heel’ or ‘red’, just to name a few. Okkular, founded by Mahendra Harish, is one new start-up that is doing so.

This new AI capability will sting copywriters that lack creativity. Copywriters that will thrive, on the other hand, will be those that possess the ability to craft a compelling story or provide styling tips – things that machine learning (currently) can’t do.

Okkular screenshot

If your organisation would like advice from our copywriters to discuss chatbot scripts, writing for voice, conversation design or robot dialogue, get in touch. We’d love to talk.