What designers need to know about agentive design

Computational design is the number-one trend in Invision’s 5 trends UX designers should know for 2020. One kind of computational design is what Chris Noessel calls “agentive design” (pronounced “agent-ive”). Noessel literally wrote the book on agentive design -- technology that assists humans in their goals without replacing human judgement or involvement. He is the Senior Lead Designer for the Sterling Supply Chain Business Assistant AI with IBM where he contributes to Watson innovation.

Clockwise discovered his work, and it inspired us. Now when we propose and evaluate different product plans we tend to bias towards the ones that satisfy agentive design best practices.

Clockwise recently interviewed Chris Noessel, along with Charles Martucci, Clockwise’s Head of Design, to discuss agentive design, Jobs to be Done, and lessons learned while trying to apply agentive design principles at a startup.

We talked about how agentive tech can help in a global recession, what your team should be doing to prepare for agentive tech, and more.

Below is a transcript of that interview, edited for brevity and clarity.

Moderator: Chris, could you give us a 30-second elevator pitch for agentive design for people who are new to the subject?

Chris Noessel: I believe that artificial intelligence (narrow artificial intelligence, not the kind that we have in science fiction) is at a point where it is sophisticated enough to not just help us do things (which software has been kind of doing for the last couple of decades anyway).

It stands to make our lives easier, buy us more time, which obviously, [laughs] I can use quite a bit of. And even, I think, help the species as a whole. How's that for a pitch?

Moderator: Would you say that Agentive Tech is for every company? Is this something that every company needs to be aware of and actually making moves on right now? And if not, what kinds of companies or maybe what industries is this going to impact first?

Chris Noessel: I first recognized this pattern out in the world around 2007 or 2008. Since then I have been looking for examples of where the technology can't be useful. I haven't found many.

Certainly from the consumer's point of view, there are things that agents can't do. I can't send it to go to the gym on my behalf. I need to send my body there. And I can't expect to get AI to learn something for me. So to learn a new language I'm not sure the Agentive Tech is going to be really useful.

Anytime you have a relationship with a customer or a client, I think that’s an opportunity for agentive solutions. So the short answer is I haven't found it yet.

Moderator: And what companies or kinds of companies are going to see the highest ROI first?

Chris Noessel: That's a hard one. And partially it's because I'm not a numbers guy. [laughs] I'm an interaction designer. But, if I had to make a call, I expect that ROI would come fastest to those who don't need to commit tons of resources or retool their business for it. That would include anyone who is in an information services capacity. Banks and insurance companies are information services. To an extent I think design is an information service and even Clockwise is an information service, where certain information goes in and it has to be pushed through a bunch of smart algorithms to be optimized well. And then what comes out the other side is information. And I think information service companies tend to act on agentive tech more quickly and then get a bigger return on investment.

Moderator: So information-based services, as opposed to physical goods or service-based services or just physical things?

Chris Noessel: As opposed to physical goods. If you sell socks or underwear, then it's sort of a weirder proposition to think, ‘How can an agent help?’ But then again, as soon as I say that out loud, I'm like, oh, well all of these products have an expected lifespan. There are opportunities to replace them and upsell. That's an opportunity for an agent as well.

Moderator: Switching over to Charles, I'd love to know what are some ways that Clockwise has been able to incorporate the learning from the book and teaching and talks into our core product.

Charles Martucci: Chris’s book was clarifying because after reading, I realized there were already a lot of agentive design principles baked into the DNA of our company and product. And I credit the cofounders and team there. Beyond validating our direction, the book gave us a helpful framework for identifying opportunities to improve the experience and share some best practices with the team.

A lot of the tasks that Clockwise helps people get done are things that humans often don’t want to do or aren’t capable of doing well. For example, we've done a lot of user research with executive assistants and other roles that do a lot of scheduling and a lot of time management coordination across multiple parties. And dealing with the scheduling constraints and the back-and-forth is generally a part of the role that they do not enjoy. Scheduling coordination across multiple parties is complicated, and is a huge pain point.

“One of the advantages of agentive tech that really jumped out to me when reading Chris’s book was the fact that agents do not get bored or tired.”

Charles Martucci

One of the advantages of agentive tech that really jumped out to me when reading Chris’s book was the fact that agents do not get bored or tired. [laughs] They can continually monitor complex data streams while simultaneously completing many parallel tasks and workflows — whereas humans are generally pretty focused on one thing or one task at a time, which is a huge limitation. Clockwise is able to broker time across hundreds or thousands of individuals inside an organization, across a complicated matrix of thousands, if not millions, of calendar events, in a very computationally efficient way. It’s really a great fit for agentive tech.

Ultimately, we're saving people time and we're able to do stuff that humans don't want to do and we're able to continually monitor multiple data streams and respond instantly to take action and help people get good outcomes. Working at this intersection made us feel like we were on the right track with our product and our approach.

Moderator: Charles, I know Jobs To Be Done is one of your passion projects and something that's really influential to you and your work. So I'd love to know how you've used both Agentive Tech best practices and Jobs To Be Done insights at Clockwise.

Charles Martucci: Agentive tech, especially when setting up the agent, is really about understanding people's goals and their preferences so that the agent can do the work on their behalf. And that is very aligned with Jobs To Be Done, which is a goal-oriented or outcome-driven innovation framework that emphasizes having a deep understanding of what people want.

I know that Jobs To Be Done can be a little bit controversial in the design and research community. I’ll skip over that and just say that the thing that's great about Jobs To Be Done is it really unifies everyone behind a common language, kind of like a Rosetta stone, that all functions across an organization can sort of relate to. That alignment is the powerful thing about Jobs To Be Done — from Marketing and Sales, to Design and Research, to Product and leadership. Everyone's speaking the same language and you're all talking about customer needs, scenarios, and the outcomes that people want.

For example, a high-level job that we talk about at Clockwise is ‘Help me complete tasks that require deep concentration.’ And then related to that job, a desired outcome would be increasing the consecutive hours of uninterrupted time. And this is an outcome that you can design for and that the team can measure, analyze, and track. And our product efforts go into maximizing this. And as a Clockwise user, I can evaluate this week-over-week because there is a correlation between how much uninterrupted time I have and how much deep work I can get done.

Whereas an undesired outcome would be, ‘I'd like to minimize the amount of interruptions that prevent me from completing those tasks.’ And similarly, what experiences can we envision to help people avoid being interrupted?

As you dive in deeper and you start to really get much more granular with these outcomes, we're able to design the Clockwise brain in partnership with Engineering. A while ago I worked really closely with Mike, one of the co-founders, to map out the brain. We made a huge mind map together and what's really cool, if you look at the first-order nodes of that mind map, each one is a job to be done. And then the second-order nodes are all outcomes, both desired and undesired. And then the third-order nodes are the conditions the system is monitoring. And then when those conditions are true, here are dozens of solutions that can really help people get the outcome that they care about. It's kind of nice when you see this all mapped out.

When designing at Clockwise, almost every workshop or activity we run, we're starting with a job to be done, identifying outcomes, and then driving all the way through to what specific solutions can deliver these outcomes.

Moderator: Chris, if you could snap your fingers and make any product or service more agentive or incorporate agentive design differently, what would it be and how would it be different?

Chris Noessel: Well, it's funny. We're having this interview virtually because of the COVID-19 pandemic. It's on my mind, of course. I just published a giant blog post taking a look at spreading pathogen maps.

“Part of my reading up on epidemiology has taught me about contact tracing. My thought was, we could do it manually. But it seems to me like that would be a super easy thing for an agent to do if you could get the sensor network in place.”

Chris Noessel

Part of my reading up on epidemiology has taught me about contact tracing. My thought was, we could do it manually. But it seems to me like that would be a super easy thing for an agent to do if you could get the sensor network in place.

So I had a design in my head. You could do it in several ways using public cell phone location records, etc. That is the thing that I think we have to do. Because field contact tracing takes field epidemiologists three days per contact. Given the exponential nature of this disease, we don't have enough field epidemiologists to do it.

So I thought, God, there's a perfect solution for agents to take. Just today before our call I read an announcement that Apple and Google are both launching their trials for an opt-in contact tracing solution. Hopefully by summer it's going to be baked into the iOS. It's still going to be opt-in to respect people's privacy. But this is a system whereby two phones sort of use low-power Bluetooth in order to identify when they are near one another. And they do it through a pretty clever system of having a temporary ID that swaps out over time so that nobody can pull this data and identify the person from it.

But if you find out that you were sick, you can go into this app and say, ‘Yup, I'm sick.’ It goes back to its own records, kind of like bit-Torrent can go back to the network and not from a top-down perspective, but from a bottom-up, reach out and say, ‘Hey, somebody that you've been in contact with has contacted the virus, here's what you need to do.’

It is an agentive solution for contact tracing. And I am super excited that it's going to go out there in the world because we need to solve this problem.

Moderator: That's super interesting. Are you familiar with test and trace, which is kind of what they're doing in South Korea and in China? Is that kind of what you're talking about?

Chris Noessel: Yep. The same thing. In fact, I believe there's a solution called NextTrace, if I'm remembering the name, who is trying to do it. That plus the bigger names that have a lot more resources and control over the platform being on board gives me much greater hope that it's going to become a reality and get in a lot of people's hands.

Moderator: In the short time we have left, I just have one more question. It's in light of the macroeconomic challenges that we're facing, a potential global recession, does that change anything about your advice to teams on agentive tech?

Chris Noessel: We know at this moment in time in the pandemic that we've lost about a quarter of the momentum of our economy and it looks like it may be more. The typical relationship between the user and the AI requires the user's attention to get value out of that thing. But Agentive Tech says, no, no, no, no. How can we provide that value by bothering the user the least?

I was intrigued that you asked [in our email correspondence] what a team should be doing to prepare for agentive tech. I'm not a salesman, but I would say that I did write this book because I couldn't find this book. I was designing a robo investor for a company and I was like, that's really weird. The portfolio does its work outside of my attention. How do I run a usability test? Since I looked for best practices but couldn’t find those best practices, I wrote that book.

So I think certainly getting that book and understanding this line in the sand is one thing that you can do. Another is, certainly reach out to people who have found this idea and are excited about it. I have a Slack channel of people who've read the book and are super excited about the concepts and while it's not a very active one, I have a network of people who are eager to work on these things so I can hook you up. And I guess the third one would be to start playing with some of the publicly available tools for doing similar agentive work. My favorite example is if this is ifttt.com. It lets you play with all sorts of agents that can watch data streams for triggers and then you can do lots of stuff with it afterwards. And I think it's really easy. The interaction design, I just love it. It’s a really easy, lightweight way to get used to thinking this way. Even if you're just playing with silly things like a calendar or an agentive tweet, it's a way to get your brain and get your hands kinda messy, playing with the ideas.