Unleash the Power of AI: Game-Changing Strategies for Product & UX Teams
An actionable list of techniques to maximise the augmenting power of OpenAI & other NLPs in your day-to-day work building products & user experiences
There is a tremendous amount of noise, excitement, fear-mongering and general buzz around AI tools right now. I won’t get into my opinions on the matter because frankly as with everyone else’s… they don’t matter.
What will matter is how we use this technology. So this article will cover ways you can implement OpenAI (or any other NLP model) to boost your output and process. There are example use cases under each point, and the purpose of this piece is to bring truly powerful additions to our existing workflows.
The following tools will be showcased:
Notion AI
ChatGPT (GPT4 & GPT3)
DALL-E 2
Otter.ai
The following use cases will be covered:
Optimise
Stretch
Conceptualise
Realize
Tweak
Assist
Encourage
Optimise
Focus on your most time-consuming tasks first
Whether it be the design process or a sales funnel, if you’re building a product or managing a team, think about what makes you the most time to complete. Bonus points for repetitive tasks, and tasks where you condense a large amount of data down into understandable/digestible content.
In a typical Product & Design role, one will spend a tremendous amount of time researching. Researching products, techniques, frameworks and so on.
Out of those tasks, researching products can often stand out on its own. All creators will take time to research, and I’ve found the more you do it, the better your output. So we can use ChatGPT to help us here.
Specifically, it can help us discover products that you otherwise would not have seen or been inspired by. It’s very common to see products in a certain market all look the same over time, and so…
You can combat this by looking for inspiration far and wide:
I first had this realisation when Parina Patel showed me experiences from Zambia and Australia while we were working on a project together at ZeroToDesign. Being based in the UK, it would have been highly unlikely for me to encounter those experiences.
There’s wonderful products being created all over the world, so don’t just research ones in your market, or rely on what you’ve heard about
Leverage the incredible work that’s being done everywhere.
Example use case: Finding unexpected/rare product experiences to enhance your inspiration list.
Stretch
Test its capabilities, and find its limitations
We all have meetings, and someone unfortunately has the very important responsibility to note-taking and summarising key points from a call. I used to do this job, mainly so that I had deep context into what went on and what needed to be done. But we can also leverage audio transcription to help us out here, by storing the words as we say them.
Audio transcribing has been around for quite a while now, and if you don’t use it already, you should. Both Zoom and Google Meet bring transcription in, and Zoom specifically is already powered by Otter.ai:
Now, once you have your meeting notes, you’ll want to strip it down to its most useful points. And that’s where a summary becomes essential.
These NLP models can be great at summarising things. I’ve sent documents to ChatGPT (and fairly long ones) and asked it to send hourly updates to my email while it processes them. I was unsure of its ability to create bespoke scenarios but found myself presently surprised:
While this felt like it would be successful, it did not end up being so. However, as AI has proliferated across so many tools, like the beneficial inverse of a locust swarm, I turned to Notion.
Notion’s AI was able to provide me with a summary of this large document and a list of action points too. All it took was an import of the document and an prompt to ‘Make longer’ the initial summary to get a fairly accurate analysis of the document:
Great success.
I tried to do a different kind of summary for another task. YouTube videos can be quite long sometimes, and there isn’t always someone kind enough to summarise in the comments. So I wanted to ask ChatGPT to do that for me.
I asked it to summarise the following video that showed up on my feed:
But, it didn’t go quite according to plan…
Originally I got a completely incorrect response to the video. Then I had to clarify and reask, and only then did I get some points. After watching the video and looking back at the comments, I found they weren’t very accurate.
The lesson here is to test what it can do, believing it can do a lot, and then see where it falls short. Good old-fashioned trial and error.
Example use case: Summarise long documents using Notion AI, but don’t ask ChatGPT for YouTube videos 😬
Conceptualise
Create ideas rapidly, then discard or proceed ahead equally rapidly
If you believe that DALL-E or Midjourney can automate the entire job of a UI designer then you are either misinformed (due to a lack of experience) or lacking confidence (probably also due to a lack of experience).
Building products is complex and requires tremendous amounts of context and information extraction. Context from those involved in building it, the existing market and potential customers.
So while things like Midjourney and DALL-E can create visuals quickly, they won’t be quite what you need.
The upside of that though, is the first point, the ability to create concepts incredibly quickly.
I used DALL-E 2 to generate images for a landing page for a music app, to evoke the emotion of someone engrossed in the act of listening to music, something which would’ve taken me an illustrator, time and money to even conceptualise.
A superheroine felt like an appropriate character to evoke the powerful feeling of listening to great music. So I asked for some images with the following prompt:
With these visualisations, I’m able to easily decide whether or not to proceed with something that was an idea in my head. There will always be ideas, but many suck in reality, so it’s better to find out sooner rather than later.
You can put these directly into the designs you're thinking of, without fear of plagiarism, as they’re all unique and don’t need attribution:
We eventually went for the Pink option as it felt like the best illustration for the emotion behind a deep music-listening session 🎧.
To have been able to do this pre-2022 would have taken not only a tremendous amount of time but probably another role, in the form of an illustrator. Now, one person can do it, before handing it over. Saving everyone time, and increasing their ability to dive deeper into their roles.
An experienced illustrator could now take this image and adapt it to be even more like what we might need, rather than needing to create the 5 conceptual pieces in the first place.
Example use case: Generate illustrations for visuals to plug directly into your experiences on Figma.
Real-ize
Use real data to bring your mockups to reality
Tell me if this sounds familiar.
You’ve created a great mockup with fantastic components, and now want to add real copy, real names and actual content inside your relevant components.
Insert Lorem Ipsum…
NO!
Not anymore.
Being able to bring real data from existing customer datasets was one of my favourite requests when working with product teams.
And now with ChatGPT, you can very easily create content & copy for your designs:
This can be a huge value add to bring your designs into reality. Mockups are good, real data + mockups are fantastic.
You can even use the ChatGPT for Figma plugin directly in app to:
This will help to create more appropriate, understandable and relatable designs when sharing with stakeholders or running usability tests.
Example use case: Create real filler content to make your designs realistic, instead of using Lorem Ipsum
Tweak
Alter the outputs, don’t take them off the shelf
As I’ve mentioned in the points above, these models may not always give you exactly what you need, and therefore some amount of tweaking and tinkering will be necessary.
The benefit though, is that you can create a starting point for whatever you’re solving, and then cut this down, adapt it, be inspired by it, or generally twist the very malleable content you receive.
AI will not have full context over what you need. The reason for that is simply that you cannot download all the knowledge you have in your brain and transcribe it onto the NLP for it to process, for your particular problem or scenario.
Therefore there will be some gaps. So don’t take what it gives you completely as gospel. Instead, use it as a foundation and tweak it to what you need.
I did this recently for a review I needed to send to a stakeholder.
The tone and exact needs will be hard to share with AI for now, so you’ll want to tweak this to sound a bit more relevant to the individual you’re sharing with.
So I can take the initial summary, but then tweak it based on what I believe to be most appropriate.
Example use case: Use AI to summarise things as a starting point and cut down and refine through proofreading exactly what’s appropriate for you.
Assist
Use it as a sounding board, impersonating different roles
Many compare ChatGPT to a powerful co-pilot, right by your side. And just like a co-pilot, it presents an excellent opportunity to soundboard against ideas/thoughts/solutions you come up with.
This can be amplified when you ask AI to take the stakeholder role, for example. Or a customer you might be trying to sell something to, interview or anything else.
Having different types of sounding boards is incredibly powerful, as it saves you time and effort in finding those people for those high-level insights.
You’re also able to prepare for meetings and feedback better, as you can use AI to spot holes in your work or prepare for discussion points.
Ask the questions you’re afraid to answer real humans first.
Example use case: Find potential challenges, issues and discussion points in your work, to prepare better for meetings with stakeholders & customers.
Encourage
Bring others to leverage the power too
If you’re reading this, you’re probably already leveraging AI somehow. That puts you in a powerful position, to influence and encourage others to do the same.
AI is undoubtedly an augmented. There’s debate about it being a replacer, but no one can doubt it enhances our abilities. Therefore, I encourage you to encourage your teammates and partners to leverage AI and start testing out prompts, to see what it can do for you.
The way you will actually get replaced is if you don’t try this at all.
Closing note
These techniques should give you tremendous excitement and hope in the new age of building products.
Focus on augmenting, not replacing. If you worry about what AI can take over from you, you need to do one thing:
Adapt
Quite simply it's Darwinian.
You can always attain new skills, and you can now leverage AI in ways that you could only dream of before.
Be a dreamer, not a reader.
Thanks for reading!
Sakky B
Co-founder, ZeroToDesign
Make sure to subscribe to my Substack, The Sketch, to stay up to date with the latest UX frameworks, Product Processes & AI augmentations.
If you found this article useful, you may also be interested in:
