Building + Interview w/ Work & Co // BrXnd Dispatch vol. 028
A few fun side proejcts, plus a conversation with some folks from Work & Co about brands and AI.
You’re getting this email as a subscriber to the BrXnd Dispatch, a (roughly) bi-weekly email at the intersection of brands and AI. I am currently planning two corporate events, so if you’re an AI company or consultant and have an interest in connecting with brands, please be in touch.
Hi everyone, and welcome back. I’m going to dive into a few updates and thoughts, and then I’ve got a fun interview I did with some folks at the product development shop Work & Co about how they’re using AI with their clients.
First off, I’ve got a new little toy I’m experimenting with. I mentioned it last time, though, at that time, it was in prototype phase. Essentially, it subscribes to your calendar (Google only right now), looks up everyone you’re meeting with, and uses AI (mostly GPT 3.5) to do a bunch of cleaning, extracting, and summarizing of the people and companies. It mostly spits out a nice little daily briefing email that gives you your rundown for the day. I originally built it as a little thing for me, but now I’ve fully ported it over and am opening it up to a few folks to try out and give feedback. It’s called Dossify, and you can sign up for the waitlist. For no, it’s free. I’m not sure how long that will last, but it’s been a fun thing to build.
Another fun little AI thing I’ve been playing with is a “bot” to do some basic tasks with Linear, the project management app. As I try to get back in the swing of project management things (mainly I’ve been using much simpler methods—like my inbox—to manage my to-do list), I was a little annoyed that Linear couldn’t auto-file tasks to projects. So … I took an hour and built a little thing to do that. Basically, what it does is:
Catch a webhook from Liner every time an issue changes
Check to make sure it’s a new issue
If it’s a new issue, send the issue name + description along with a list of my projects + descriptions to GPT-4 (3.5 wasn’t doing the trick for this one) and get back which project it should be in (with a choice to return null if it’s not sure).
Update the task with the project using the Linear API
It’s all pretty straightforward and, again, maybe something I’d put out there if there’s interest. I actually have built a few other things into it, too, like the ability to email in tasks, which Linear doesn’t support.
But mostly, this kind of stuff is too perfect for AI. While many companies are adding various “magic” creative tasks to their AI (like this weird new Google Docs writer thing), these basic classification tasks are a much more fruitful and simple place to apply the technology. This is something you shouldn’t have to think about, and now, mostly, I don’t have to anymore.
Also, and I’ve mentioned this in the past, when you’re building things (or just prompting AI), one general rule to make sure you follow is offering an escape valve in the prompt in the form of something like “if you don’t know the answer return null.” Without that, the model will generally follow your prompt as strict guidance and return even if it doesn’t know the correct answer.
And while I’m on small random prompting things, I finally found a decent use for the ChatGPT custom instructions: I told it to always give me answers in Typescript and SQL for Postgres unless I specifically ask otherwise. That seems much more helpful than telling it about you or your job.
Okay, onto the interview. This is the first of what I hope is many interviews with folks who are applying this stuff for brands and marketing. As usual, I try to focus as much as possible on application over speculation, but of course, things sometimes have to veer there. I hope you enjoy it.
Let’s start by having you tell me a bit about who you are and what you do at Work & Co
Oliver Dore: I’m a partner of technology and oversee our engineering team here, but also am actively on client engagements, and have worked on technology strategy and development for clients such as Apple, IKEA, Pfizer, Embraer, Philz Coffee, and Virgin America. I’ve been at the company for 10 years.
Rachel Bogan: I lead our product management discipline. We form multidisciplinary teams of experts in design and engineering, led by a product manager to guide and shape the overall product experience. I am also a retail and travel practice leader, having led engagements for Target, ALDO, Vistaprint, Expa, and Royal Caribbean
What’s been your experience integrating AI into your work? What specific parts of the process have you found most fruitful for layering in AI?
Oliver: We think of AI as a broad ecosystem of tools that drive individual value for teams. That could be how it helps to facilitate a day-to-day workflow for individuals —and not just through ChatGPT or Midjourney— but through a range of emerging tools, the use of large language models, summarization, transcription, text-to-speech and more.
To the point around what’s been most fruitful, I’d say it’s simply gaining experience and developing perspectives through regular use of these tools. Every experiment and application of the technology, both internally for Work & Co and also for our clients, is uncovering value and ways for AI to drive operational efficiency for organizations and richer experiences for their customers.
The learnings are constant. This summer we convened a group of C-suite execs and senior digital leaders at retail organizations to talk about how they are exploring and testing AI within their organizations with the goal to streamline processes and gain insight into how it will impact their businesses. It was interesting to hear how they were doing this, from very simply encouraging the use of ChatGPT for first drafts of job descriptions to exploring inventory management systems to creating a workflow to produce content for product descriptions using AI platforms. These experiments created a jumping-off point to understand the power of the tool as well as the challenges and expand their teams’ personal comfort. A few months later, a workshop with one of the teams involved in that conversation dug into building an AI driven e-commerce vision for the brand.
Organizations have been taking different paths when it comes to organizing their teams around AI - but what we have found is a top-down approach isn’t effective. With learning and development around AI, it’s pretty constant and ongoing with lunch-and-learns and workshares happening on a nearly weekly basis with presenters from different disciplines - engineering, strategy, design. Some of the biggest takeaways have been being open-minded, having a willingness to test and share learnings across your team, being aware of the risks from a company perspective, and then positioning the opportunities to de-risk them.
What are clients asking you about most?
Rachel:
I think we’re seeing a divide with clients - those who aren’t quite ready to experiment with AI and those who are. Often the ones who are ready to take the leap are also the ones who invested early in digital transformation and are now focused on taking a leading position in integrating AI into existing experiences or building new ones given the opportunity AI allows. Our clients tend to be strategic and think deeply about AI beyond marketing applications.- and if they aren’t ready for a public-facing play, they are working through other ways to experiment with AI.
One area we are seeing this is back of house inventory and logistic management. Things that might not feel shiny but are critical for an AI-driven future. Those jumping in and focusing on those areas and approaching the adoption of AI in this way will be the first to feel the impact of efficiencies that help them realize growth for their companies in new areas.
Many clients also ask me how they should operationalize their teams. That conversation might start with “I think we’re going to need a prompt engineer.” or “How do we structure our entire engineering and product team around AI?” I think the short answer to that is we don’t know.
Using the example of “prompt engineer,” - maybe the future of prompt engineering is actually the future of a UX designer, a front-end developer, or a role you have today. That just changes as you think about tomorrow. It’s similar to the world of product – at Work & Co, we don’t separate UX designers and visual designers as disciplines – we simply have product designers. So I think one thing all companies should be planning for is not replacement but rather re-operationalizing around new roles because it is possible that these are roles that people will evolve into.
The last theme I’m seeing is clients who are eager to bring to market an AI-powered feature for consumers within one of their existing digital touchpoints. What I’m excited about is getting to the point where we are thinking about how it completely revolutionizes those experiences and how consumers interact with brands through reimagined user interfaces.
I saw your AI-driven commerce vision and one place I very much agree with you on is that there are big opportunities for conversational interfaces and AI in shopping. How do you see that playing out?
Rachel: Shopping and discovery powered by AI will be a fundamentally more curated experience for consumers as we move towards new commerce experiences. Everything that we call personalization today tends to be generic, and the lack of relevance often harms how consumers perceive brands. In ecommerce there has always been a desire to innovate to give the consumer a rich personalized and high touch experience. But a big factor hampering what’s possible is the limited technology we’ve had in this space. So much of what has been created is based on standard recommendation engines, or slowed down by the limited capacity that brands have to generate content including product descriptions or relevant recommendations. And so with AI, I think the digital commerce experience will become as rich as what a shopper would expect from an in-person curatorial experience.
But I also strongly believe that the beauty, especially as we shop, is going to be in what you can’t see. To a shopper these incredibly evolved experiences will feel like magic and they will have very little knowledge of the role AI is playing. It will play out in the most personalized and accurate lists or recommendations and elevated experiences that we have yet to imagine.
Where do you see all this going from a user experience perspective? There seems to be widespread agreement it’s not just going to be chat, but I haven’t seen a lot of good answers for what it will be.
Oliver: One significant change we will see emerge is multi-modal interaction. Yes, conversation through text, but also the ability to reference images, video, voice, 3D models and more. OpenAI recently released GPT4V with some of these multi-modal capabilities. I think it’s going to be a big component of how things are going to change.
Context and long-term memory are going to be topics that we’ll be hearing a lot more about in the future as well - beyond profile and history of buying behavior. What if you could go to a website or a mobile application and it not only knows what you bought and your profile, your demographic information, but also the conversations that you’ve had in the past and how it changes the way that it might interact with you or serve content to you. I think the idea of long-term memory is something that’s going to become a common topic, one that is core to helping to build more personalized, longer-term relationships with users in these experiences.
Considering the rapid advancements around all things AI, how have you fostered a culture of experimentation? And how are you working with clients to help them do the same?
Oliver: This comes up in almost every AI conversation, and it always comes back to the same point: AI is a big topic with a lotof hype right now. And everyone’s coming at it from slightly different places based on what they’ve seen and read in media. I think we have a responsibility for our own team and our clients to move past the hype and get down to education.
When it comes to innovation and experimentation, none of that’s really possible until you have a team that’s open-minded, maybe even excited about what the prospects could be. To get there, you need to start to unpack the subject so that they can move past the initial anxiety to embrace and understand more tactically what it could mean for them. You realize that the point of entry for working with AI is much more accessible than you may have realized and as soon as you know that the doors open. Experimenting like that requires a team with the latitude to explore with open-ended goals, and a bit of trial and error. AI has always been a point of interest for a curious team of designers and developers at Work & Co but as ChatGPT took off we expanded the dialogue to every team in the organization. We created forums, chats, internal learning sessions all which allowed us to get insight into what the teams were excited about - where their fears were - what they thought might be possible.
Lastly, jumping in will help shape your perspective as a team. And not only what you can achieve with the technology, but what your stances on the technology are —which comes back to ethics and values and what responsible AI means for your company. None of that’s achievable without getting started.
Rachel: To echo Oliver, you just have to get started. We recommend starting small and being nimble: focusing on a specific area of opportunity and experimenting with commercially available tools.
What we’ve learned through this process is that it’s about the journey and not necessarily the destination. The technology and tooling is changing rapidly. Building an underlying understanding of how the technologies work and the use cases in which it can drive value is the real unlock. That is what makes our team and our clients able to separate the real from the hype. The way specific features or functionality will come to market is going to evolve, so a strong foundation of understanding and creative thinking is what will set you up for success months and years out.
That’s it for this week. I hope you enjoyed it. As always, feel free to be in touch if there’s anything you want to talk about or I can help with.
— Noah