Right now, artificial intelligence is ‘having a moment’. AI and machine learning may have been around for years in one form or another, but with the late 2022 launch of OpenAI’s ChatGPT, we’ve reached some kind of tipping point. No matter the type of business, sector, or field of endeavour everyone seems to be talking about AI and what it can do for them.
Reactions range from “we’re all doomed”, to “it’s too much/I’m too old”, to gung-ho enthusiasm. But what about the field of product design and development? What can AI do for us?
For a recent Product Unleashed event, we invited Avi Ashkenazi, Head of Product and Design, Store Localisation at Shopify to answer that very question, looking at the different stages of product development: shape, develop, analyse, optimise, and maintain. But first…
Artificial intelligence is…
Put simply, AI is the field of using computer systems (software, algorithms, etc.) to carry out tasks or functions that usually require human intelligence. Right now, the best-known example is probably ChatGPT, a generative large language model (LLM) that can engage in human-seeming conversation and assistance. Users input ‘prompts’, from simple questions to detailed instructions, and in response ChatGPT produces answers in the form of written content. That written content can range from a simple reply to an essay, social media post, programming code, etc. Other examples of AI might deal in images (e.g. Midjourney) or act as a kind of personal assistant (e.g. Pi).
The outputs take the form of familiar media, such as text, audio, video, images, and potentially 3D models. And the really good stuff happens when you combine AIs.
“Interesting things happen when you use one to feed another. For example, if you want a good Midjourney image, you ask ChatGPT to create a prompt for you based on a sentence – it creates 200 words for you, you feed that to Midjourney, and you get a beautiful picture.”
Avi Ashkenazi
Back to product development…
Shape
In a product’s design and development, there’s a lot of data to sort through – from user interviews, feedback forms, online comments, competitor content, and so on. At this early stage, AI is ideal for parsing this quantity of information to produce useful summaries that can then direct the first design attempts.
Once you have a product version, however, the challenge can be getting stakeholders to understand it. Again, AI such as ChatGPT can serve as a great explainer. It can even explain lines of code in a prototype or MVP app, saving developers from simple questions (simple to them, at least!)
One point Avi was clear on is that for reasons of proprietary data and security, you probably want to set up a private GPT for this kind of usage.
(NOTE: Unless you turn off ‘chat history’ in your account settings, ChatGPT will save your data and potentially share it with third parties. For more info, check out OpenAI’s data privacy policy).
Develop
As you work on a product, you can also use an LLM like ChatGPT as a sounding board or devil’s advocate. As Avi says, “Usually in a project, you have a bunch of assumptions, you have a bunch of risks in mind, and running it through an LLM can really help you understand alternative opinions or views.”
Maybe you ask your AI what potential questions people might raise about a product or presentation, or ask it to highlight the gaps in a proposal. AI can be a useful ‘critical voice’ when working on your creation and, as such, can speed your progress towards product-market fit.
That said, as with any new tool, there’s a learning curve. Simply throwing a few keywords at ChatGPT will not get you the best results; a conversational approach is much more effective: “It works better if you talk to it like a human.”
Analyse
Thinking of the product development process in terms of funnel, again, AI can help you refine that process and get the most from it. You can take the data from your development funnel, feed it to an AI, tell that AI what you’re aiming to achieve, and receive a variety of proposals for improving that funnel.
When it comes to making ideas a reality, Avi suggests that you can even feed it images, such as a product sketch, and AI could potentially turn the back of a napkin into usable HTML code.
Optimise
Back to dealing with stakeholders during product development, AI can make it, “…easy to convert something into a variety of different outputs, enabling you to communicate with a variety of different people.” Avi tells of a time he used AI when faced with the need to get a key decision from a hard-to-reach (and hard-to-impress) senior stakeholder. Taking a transcript of a previous meeting in which all the issues had been debated without a firm conclusion, Avi asked ChatGPT to summarise the pros and cons of the situation. He then asked for a two-minute scripted conversation based on those pros and cons, with two characters taking the two sides. That became the basis of short video presentation (with Avi playing both characters!) that could then be sent to the key stakeholder. AI in action.
This transformation of informational content also has implications down the line. A marketing team, for instance, could use AI to produce a blog post; then turn that blog post into a LinkedIn article, then a series of Tweets, then an entire ad campaign, then… you get the idea.
Maintain
AI’s ability to handle information beyond human capacity comes in handy for maintaining or updating your product or information. Avi offers the example of a Shopify store. Let’s say the client – the ‘store owner’ – wants a new theme for the Black Friday sales. While the ‘new’ store is being created, the existing store is changing (and its data with it) as customer interactions and transactions continue. AI can track every element of the store(s), highlighting only those updates necessary to ensure continuity when the new store theme is applied. With AI, updating is quicker, easier, and smoother
AI & Product Development – key takeaways
It’s clear that AI solutions already offer a varied and highly useful toolkit for the product designer or developer. We’re still in the relatively early stages of widespread adoption, but already AI can manage (and add to) user and stakeholder input, identify issues that the development process must tackle, produce and maximise content, and help produce creative solutions to a variety of product-related challenges.
However, one last caveat, in case you’re thinking AI is the quick fix to all your product design and development worries: you can’t expect to throw a basic prompt at ChatGPT to get perfect results.
“You need to invest time in using it, build expertise in how to use it, AI is not something that delivers the final output.”
Avi Ashkenazi
Yes, there is still plenty of room for human input and creativity in product design and development!