Skip to content
This is the second in a multi-part series about Generative AI, focused on how to set up your Generative AI project for success. Whether you’re new to GenAI, or have your own tactics to share, there’s more we can all learn about implementing this new technology.
Hopefully you’ve already read our Checklist for GenAI Readiness and you are following our best practices: from managing expectations, to cleaning up your data, to building in time for testing and embracing the whimsy of an exciting emerging technology. Let’s move on to the next step and talk about tips that can properly define an ideal design and project process for your GenAI product.
Here at Grand Studio, we tend to follow the classic double diamond design thinking process, defined by the British Design Council in 2005. We use this to guide our clients through a design process that focuses not just on making sure we’re designing something that works and is reliable, but something that people trust and are excited to use.
Let’s dive into some of the major phases that can define an ideal GenAI design process.
Phase 1: Definition + Discovery
Does a GenAI solution effectively solve user problems?
We’ve said it before, and we’ll say it again: While GenAI is a useful, exciting technology, it isn’t guaranteed to be the answer to every problem under the sun. Here are some examples of what it can do now:
- Automate repetitive, tedious tasks such as writing boilerplate emails, or scheduling meetings
- Create outlines, write basic copy, and provide inspiration and ideas into the process
- Create customized images – some of which can even be generated from any of your existing illustrations
And here are some examples of what GenAI is not quite suited for just yet:
- Replacing human decision-making processes, or making suggestions that require long-term, contextual strategic thinking
- Fact-check itself against hallucinations or bias due to the technology’s limitations in understanding the “truth”
- High degrees of nuance and problem-solving. Although many LLMs have passed crucial benchmarks set by different data sets and exams that we use to test people (such as the MKAT, SATs, Bar Exams), they’re not really capable of fully autonomous thinking (yet…).
As you and your team consider the possible applications for GenAI in your work, make sure you’re not replacing the difficult, nuanced, creative work that humans are uniquely capable of doing with a technology that is not yet well-suited for it. You will end up with unhappy employees, unhappy customers, and certainly unhappy technology.
Phase 2: UX/UI Design
What should the user interface and experience for your GenAI product look and feel like?
Now that you have identified and selected GenAI as the right technology for this project, it’s time to figure out how your users might interact with it. Understanding how GenAI will integrate into existing – or brand new – products is an emerging field of UX/UI design. We’re seeing everything from AI-enabled UI design engines that create UI from a user’s text or sketch input, to straightforward chatbots, to DJ players built on text inputs, to agentic co-pilots.
There are so many creative ways that users can interact with GenAI, and part of this initial building phase is to take the learnings from the Definition + Discovery Phase to understand what kinds of interaction will make the most sense to your users, solve their problems, and delight them. You and your team should ensure that the interface follows basic UX/UI heuristics, is user friendly, and intuitive. Don’t be afraid to use existing patterns, brand guidelines, voice, tone, and personality from familiar applications and products – but think carefully about what is unique about the interaction between your user and the AI and how it can become an extension of your brand that delights users. We’ve found that it’s best to implement incrementally and integrate into existing systems, as opposed to making something brand new. Even though this is an exciting, emerging space for UX/UI design, the same foundational principles still apply!
Phase 3: Beta Test
How do users interact with and understand the product, and what is required to build trust in the technology?
For anyone who has introduced a new tool or technology to users, you know that people don’t always react or understand the technology in the same ways, or even ways that you could have anticipated! Add the current range of attitudes regarding GenAI – and the complexity of emotions people have when using it – and this effort starts to invite risk upon launch.
Fortunately, to mitigate all of those risks we would recommend conducting a round of thorough beta testing to understand how your users will understand, interact with, and trust (or mistrust) your GenAI product. The goal is to learn more about your product and your users at the same time. Here are some things to monitor as you run a Beta test:
- Prompting: What are the prompts people are using to interact with the GenAI? Do they understand how to edit or adjust their prompts when they get an unexpected, or unhelpful response?
Consider: Think about creating a basic prompt library of pre-written prompts according to common use cases. These can help empower users and ease their comfort levels using GenAI. Another plus is that the product will work more reliably!
- Benchmarks: Speaking of reliability, it’s crucial that everyone is on the same page about how success for this product will be measured. Beta testing is a perfect opportunity to start seeing how well the product solves the problems you’ve designed it to solve.
Consider: You can set some pre-established benchmarks by checking your model against existing industry model benchmarks and comparing how your model measures up. Keep in mind that these benchmarks will change if you are using a customized model, and that the models themselves are constantly changing and improving.
- Change Management: People have lots of feelings about GenAI, and you can’t necessarily blame them. GenAI is constantly in the news as either the answer to the world’s problems, or signaling the end of people’s jobs. You’ll want to be sensitive to this when launching your product, and beta testing is a good opportunity to ask your users how they feel.
Consider: Plan to add interviews or an open-ended survey to your beta test to surface what level of experiences people already have with GenAI, what they think about it, how much (or little) they trust the technology and its outputs, along with anything else that you think will make a difference in your adoption strategy once you reach product rollout.
Phase 4: Refine + Deploy
How does the product work in the real world?
The final phase of your GenAI product is getting everything you’ve made, tested, and refined out into the real world. You will want to make sure you’ve incorporated all the research findings you’ve uncovered along the way.
Make sure you’ve built in methods of continuous improvement and measurement. The last thing you want is any ill confidence around your product’s ability to function correctly or – even worse – causing problems somewhere else. One thing we’ve done in the past at Grand Studio is build in short surveys at the end of GenAI-enabled chatbots that ask for feedback specifically about how the GenAI performed. We can then use that feedback to refine our model and measure reliability and user trust.
Finally, ensure you have your best testing-informed change management plan in place once the product is deployed. In our experience, complex matrixed organizations can face a lot of tool fatigue (when new internal products and tools are constantly being rolled out so fast that people aren’t sure of what they do, or how to use them). An effective change management process can help guide people toward trust and understanding once the tool is out there.
Need help with the design process for your GenAI project?
Grand Studio can help. We use a tailored approach to work with you to define and fully understand your problem space and how GenAI can be best utilized to solve problems for end users. Be sure to stay tuned to the final part of the series: Advocating for the Human in a GenAI World.