Like Chuck and Oliver have done in their preceding posts, in this post we’ll continue to explore how Vervint is experimenting to find the Human-AI Partnership in the ways we’re working to bring products and services that bring value and delight to our clients and their customers. Our focus this week is on how the Human AI Partnership impacted our typical approach to human-centered design, as seen through the lens of our internal case study on building a staffing and resource management dashboard.
More specifically, we’ll explore how I experimented to find the right balance of AI and human partners to deliver that design process in under five days.
Human-Centricity
As a refresher, practicing human-centered design revolves around understanding the wants, needs, and pains of the people we’re building for, whether it’s customers, internal users, and/or stakeholders. More than any technology we use or method we apply, the success or failure of the things we build depends on how we well we understand and deliver on those things; without that understanding, we might be spending a lot of time and money to deliver something that is spectacularly wrong. Imagine having a friend who’s really into cooking that you give a bunch of steaks because hey, I’ve love to cook and to grill and I’ve got a half-cow in the freezer and who doesn’t love a good protein-rich meal, only to remember later that they’re vegetarian. Without understanding your friend’s needs, you risk delivering them a crummy outcome… no matter how nice those steaks might be.
We want to learn as much as we can about the people we’re building for so that we stand the best chance of being able to meet their needs with experiences that are delightful, valuable, viable, and effective, so we follow a process that considers people at every step:
- We empathize to better understand the people we’re building for;
- We define the problems and challenges clearly;
- We ideate and create concepts based on that definition;
- We prototype to explore and a single idea more thoroughly;
- and we test to determine how close our prototype gets us to success, by how far we may have missed the mark, where things work, and when they fail.
But Wait… You Said You Had Five Days?
I wasn’t willing to compromise our process just because I was working with limited time, so I wanted to explore how and at what points in our process I could best leverage a human-AI partnership to keep things moving without losing the integrity and quality of our work.
Just like on any other team, the best work comes from applying the right skillsets to the right situations. Where were the best places to apply my AI teammate, and where were people best suited to take up the charge? The key was in the type of thinking and working necessary across our process.
Our thinking and working styles change along each of the steps of our process. Divergent phases like empathize, ideate, and even test are characterized as “flares” where researchers gather as much information and designers create as many concepts and encouraging as many ideas as possible. The goal of this phase is to generate as much material as possible: quantity to enable the critical evaluation and inflection on quality characterized in convergent phases. When we define and prototype, we’re narrowing our attention and making decisions on the most relevant insights and the most promising or interesting concepts.
Divergence, AI, and Quantity
Generative AI excels at making things, so it was well-suited to assist in divergent thinking contexts. Because the emphasis here is on quantity over quality and my time and capacity to go deep was limited, I used AI capabilities in a variety of design tools as a very-fast-maker-of-lots-and-lots-of-things to get me past a blank page and into having something for myself and others to react to. Using the limited inputs I had on hand, I was able to accomplish each of the following:
- Generating insights from meeting recordings with our stakeholders
- Creating personas including tasks, goals, pain points, and potential opportunities
- Mapping journeys and jobs to be done based on those personas
- Prototyping multi-screen interactions based on the personas, journeys, and design system documentation
This usually takes days to accomplish, but I was able to generate multiple draft personas and journeys as well as almost two dozen prototypes in less than two hours. Thanks to my AI partners I had all of the trappings of an effective research and design effort: all of the digitized sticky-notes, journey maps, and clickable screens I could possibly want, seemingly ready to be handed over to the next phase in a fraction of the usual time required.
But was any of it any good?
Convergence, Humans, and Quality
The concepts I got from my AI partner weren’t bad, but they all had critical faults that made them undeliverable: they had incomplete workflows, or missed the mark on design library implementation, or they were functional but not earth-shattering or innovative. The work of convergent thinking requires subjectivity, expertise, and care that generative AI does not have—but experienced SMEs, skilled designers, and human users do. Since my divergent work efforts had moved quickly, I had time to pull people together for:
- Validating journeys and personas through peer review with human SMEs
- Reviewing concepts with designers and developers to ensure quality, feasibility, and accessibility
- Extending concepts to ensure effective integration into complex ecosystems
- Ensuring edge cases are anticipated and designed
- Refining designs prototypes using best practices, design systems, and innovative thinking
- Usability testing prototypes with human SMEs and users
Maybe it’s possible to prompt your way through this kind of work effort but when I tried it was fussy, time consuming, and resource-intensive. Engaging people to use their talents to guide, build, and refine with care, creativity, and intention is time and effort well-applied.
We Made It!
Our experiment with the Human-AI Partnership made it possible for me to accomplish a cycle of our design process in a very short time and demonstrated that, if anything, the value of maintaining humans in the loop becomes even more valuable as your timelines become tighter.
AI tools promote efficiency and speed, but their best work comes from intentional application in partnership with your human experts—your developers, your product managers, your designers, your users—are the driving force of quality and innovation because they do the one thing that AI will never do: and that’s care.
