Convo question: Should I learn programming?
As a conversation designer, should I know how to code? Some pros and cons.
I get a lot of questions from aspiring conversation designers. This time: should convo designers know code? Here are some arguments for and against!
Yes, it will make you a better designer
I often coach new conversation designers who come from a writing and language background. And while they usually fly through the copywriting part of their coaching, I often see them struggle a bit when it comes to the uber-structured approach that’s needed to make a consistent and well-functioning CUI.
Even the most no-code or lo-code platform will use some form of programming logic somewhere. Variables, conditional flows, slot filling, snippets of JSON, even the very happy paths and disambiguations that you design: you’re bound to encounter elements of code logic in your day-to-day conversation design.
Mind you, this doesn’t mean that you need to become a full-blown developer! But learning to read some basic JSON, HTML, CSS or — of course — SSML, experimenting and tweaking it…for sure!
No, don’t water down your skill
With the professionalisation of conversational AI, I see a growing tendency towards specialisation. A few years ago, a conversation designer could be a lot of different things. Right now, I see a diversification into content & copy, NLU and language, UX and product management and data analysis.
Programming could very well belong to a specialisation of its own. Like writing and designing, it’s not a skill that you learn in a few weeks, and true mastery of the craft may easily take years.
(Funny, isn’t it…I’ve never met a software developer whose job profile included copywriting for a corporate website. Go figure!)
Yes, coding will open up new design possibilities
Knowing what’s technically possible and being able to build some of that yourself, can be a great addition to your designs. Rather than limiting yourself to what a no-code platform offers you, you could open the wonderful world of platforms like Rasa, and build anything you’d imagine possible.
No, coding will draw you way too much into technical details
Something I do notice myself when I’m working with code and logic: it’s so easy to let your design possibilities be dictated by technical (im)possibilities and totally lose sight of your customer and their needs.
Too often do I wonder afterwards: did I really help my customers with this? Could I have solved this tech or NLU issue with conversation design, in a fraction of the time? And of course, the answer is almost always ‘No, and yes’.
Yes, it can make your analysis life SO much easier!
The one reason why I keep returning to coding: the absolute horror that’s called manual data analysis and classification! Plodding away through endless Excel lists, manually labelling false positives, conversational topics and whatnot.
Here is where I see possibly the most added value: handling large amounts of data, being able to clean them, sift through them and gain insights from them in a structured, repeatable way.
And yes, there are more and more commercial platforms becoming available that can cluster and classify your customer data in seconds. But I strongly believe in grasping the basic principles yourself, in really understanding text analysis and especially, in familiarising yourself with the intricacies of human language that make these questions so much more than just data.