One of the more famous papers about artificial intelligence last year came from METR, a nonprofit that evaluates frontier AI models. In July, it published results of a randomized controlled trial studying experienced open-source developers. It found that when they use AI tools, completing tasks takes them 19 percent longer than when they go without. That was surprising enough. But the real twist is that when these same developers were asked what AI had done for them, they reported that it had sped them up by 20 percent.
This was a fascinating dive into the professional productivity that’s promised by our AI overlords. We’re starting to learn that much of this productivty is perceived. I’ve felt this in my own professional life, and have drastically reduced my use of Copilot at work because I found myself spending far too much time reviewing and correcting incorrect output from the model.
I also found many of the outputs, when accurate and correct, were just OK and simply not up to my professional standards. So much of my daily work requires communicating effectively through writing – explaining value and impact to leadership; acting as a translation layer between engineering, design and the business; and aligning stakeholders to broad, complex initiatives – all of which need to be buttoned-up to my highest standard. I’m simply not getting that quality from any AI model I’ve tried.
In a meeting earlier today, a colleague used the term user error. My ears perked up. After nearly twenty years building digital products, I’ve heard this phrase countless times and it never sits right.
This is how I’ve found it usually plays out: Someone uses our product in an unexpected way. Something breaks or doesn’t work like they thought it would. A product team member then asserts:
“They’re using it wrong.”
“The app isn’t designed to work that way.”
“They just need to tap this button first.”
“The software is functioning as designed.”
Next comes the verdict: user error. The person is making the mistake, not us. They need to adapt to our design, not the other way around.
This mindset is poison for great product work.
Every unexpected behavior, every “misuse,” every bug report from someone navigating our product differently than we imagined aren’t errors. They’re signals. They’re opportunities to learn about actual usage patterns, understand what people are really trying to accomplish, and iterate toward something better.
The phrase user error lets us off the hook too easily. It places blame on the person trying to get something done rather than on the system we designed. It assumes our mental model of how the product should work is the only valid one.
Users don’t care about our mental models. They care about their goals. If our product makes it easy to take a wrong turn, that’s a product problem, not a user problem.
There’s no such thing as user error. There’s only unwillingness to acknowledge that we might have gotten things wrong and a refusal to recognize opportunity to make things better.
I’m looking forward to being back in New York City next week for NRF 2026. The big things on my radar this year are learning about ways to create progressive clienteling experiences for our customers, flexible & unified checkout technologies, and task management tools for store teams.
Also, I’m extremely keen to meet up with other people working in store tech product management at other retailers. If you’ll be at NRF and you’re interested in similar things or offer these types of services/products, let’s connect.
Also, also! The unofficial NRF Pittsburgh Steelers wild card watch party will be at Printer’s Alley on W 40th in Midtown on Monday night. I hate that I won’t be in the Steel City for this game, but the next best place to watch is at NYC’s Steeler Nation headquarters. Pack your Terrible Towel and swing by.
New to me: the News Product Alliance, a nonprofit that supports product professionals in newsrooms with the goal of elevating the product practice and expanding the diversity of product thinkers in decision-making roles at news organizations.
Our vision is to empower the news industry with a new generation of diverse leaders – News Product Thinkers – who have the empathy and know-how to build resilient news organizations, deliver quantifiable business results, and rebuild trust by ensuring we truly serve our communities.
This is a worthwhile cause. I hadn’t thought about how product thinking might play a role in the future of journalism, but it makes complete sense. Grounding media strategy in product foundations and understanding desirability, viability, feasibility and usability (through a constant lens of journalistic integrity) may just be a winning playbook to breathe much-needed life into the industry.
I’ve been heads down at work for the past several weeks rolling out a significant operational change (supported by new technology) to a segment of pilot stores in our retail fleet. This rollout required me to visit stores on both coasts listening for feedback, observing things that need to be improved and iterating quickly to deliver value in near-realtime. It’s been invigorating and has brought back to the surface all the elements I love about product management as a practice!
Ultimately, the problem space we’re operating in is this: How might we make the products in our stores easier to find for our employees and our customers?
The big change we are attempting to deliver is the introduction of product location data in-store. It’s a big, gnarly problem to solve for a legacy co-op like REI, mainly due to the extreme variability in our store layouts. If you’ve been to more than one REI, you know that some locations are in historical buildings – an old train station in Denver, for example – while others inherit a simpler big-box & strip mall retail feel. This variability in warehouse size & organization, combined with diverse floor sets across markets, create complexities that make standardizing a process and delivering tech that works across all locations extremely difficult.
One store workflow this new product location data will help improve is the sales floor restocking process. For the first time at REI, we know how many units of a SKU are on the sales floor, how many units are in the warehouse, and where those units are in the store. My team is also ingesting several data elements from our Visual Merchandising team and we wrote a machine learning algorithm to forecast a Target Sales Floor Quantity for every SKU in the store.
So with the raw location data and the algorithm telling us what should be on the sales floor, we were able to develop a new restocking tool in REI’s employee mobile app (Ascent) that is centered around one key hero metric: Sales Floor Percent Stocked. Store employees can now get a real-time snapshot of their sales floor stocked rate, along with a prioritized list of products that need to be restocked, on their mobile devices. This is a big step forward for our store teams.
One of the things I’m most proud of related to this pilot rollout is the feedback we’re getting from users about the Ascent features. Because we lean into co-creation mindset, the product team was able to deliver an initial version that delighted store teams out of the gate and we continue to iterate as we learn more about usage. The Ascent team obsesses over quality and store employee experience, and I think that’s evident in the product we deliver. I mean, tell me this is not one of the most elegant employee-facing app interfaces you’ve ever seen.
Employee-facing apps don’t need to suck. The Ascent team is lean: 1 front-end engineer, 1 back-end engineer, 1 QA analyst, shared product designer, shared product manager (me). The lean-ness of the team presents some hurdles, but it also affords us the ability to take both an an agile approach that prioritizes speed-to-market and an artisanal approach that prioritizes craft. I believe this mode of operating is our sweet spot.
As we enter REI’s holiday code freeze, we’ll be hands off on production changes but we’ll be working hard behind the scenes on the next version of Ascent (ETA January) that will power location data enablement across the enterprise.
I fed Claude some examples of bi-weekly stakeholder updates for products I previously managed. I then asked it to learn the format, understand the tone of the writing, and help me draft a first installment for a new initiative I’m leading. We chatted for a few minutes about the voice I desired, recent progress by the team, and the health of the project. After I provided adequate context, Claude generated a draft for me to review. The initial version was very good and only required a few copy and formatting edits. I was happy with the result and it saved me about an hour this morning.
Note: This post is the first in an ongoing series called How I Used AI Today, inspired by friend and former colleague Beck Tench who does something similar over on LinkedIn. I’m starting to believe the thinking and narrative around generative AI is becoming too binary. The intent of this series is to keep me publicly honest and intellectually responsible with my use of this emerging technology.
Few things in life are actually urgent. True emergencies do happen, but hopefully they are rare. The urgency I’m referring to is fabricated. A modern myth.
Our culture has evolved to value instant gratification, instant response and instant turnaround for most things. The faster your synapses get feedback, the better.
It doesn’t need to be this way. In fact, this faux urgency creates conflict with several of the personal pillars I hold dear: care, craft and quality.
Caring about something requires that you get to know it over time. A relationship is necessary for care to exist, and relationships don’t take shape instantly. They’re built on connection, trust and empathy – all elements difficult to nurture quickly.
Likewise, craft requires practice. And by definition, practice is working toward perfection over time. A craft is not developed overnight, but over years. Sometimes decades.
I think quality is the summation of care and craft. A thing of quality can only be the result of time spent caring about an outcome and crafting a response to that care.
All of this requires that we slow down. Turn off the firehose. Preference the signals that matter. Notice the details. Ask nuanced questions. Make space for diverse perspectives. Take on difficult conversations. Become intentional about our actions. By living this way, we’ll be able to center the care and craft required to deliver the quality the world deserves.
I envy product thinkers who operate within the context of a lean and born-digital startup. Product strategy is never easy, but building technology in this environment becomes fairly straightforward. Write code, test, deploy. Rinse and repeat. Or some variation of this. But try bringing that same approach into a complex physical environment like a retail store, and suddenly you’re not just a product manager – you’re part ringmaster, part therapist, and part exorcist for technology that seems possessed by real-world demons.
I’ve spent the last two decades building tech designed to be used in physical space – first in museums, then in retail organizations – and if there’s one thing I’ve learned, it’s that the gap between the digital roadmap and the reality of the floor is extremely wide.
First, there’s the idealism divide. Most technologists think about users as disembodied entities who interact with software in predictable, often ideal, ways. This is the happy path mentality. Meanwhile, most retail associates are often juggling many scenarios at once: a customer who’s trying to return a swimsuit they bought 6 months ago, a thief trying to steal an expensive piece of outerwear, a random question about product specs, or a manager who’s just informed them they need to be cross-trained on a new area of the store – all while attempting to use enterprise systems on six-year-old hardware.
Then there’s the physical environment itself. That sleek tablet kiosk we designed? It’s now positioned directly under an HVAC vent that drips condensation like a leaky faucet. That in-aisle digital display meant to guide customers? It’s been commandeered as a support pole for seasonal decoration. And the once-white customer-facing payment terminal now bears the fingerprint smudges of a thousand customers.
Let’s not forget connectivity. In the product requirements, the system requires a stable internet connection. In reality, we’re dealing with large-scale Faraday cages that create spotty Wi-Fi at best.
The gap between digital intention and physical implementation creates a special kind of cognitive dissonance. Most product managers are trained to think in terms of user journeys and personas, only to watch customers use the self-checkout as a surface to scratch off a lottery ticket. I’ve grown to love this dichotomy over the years.
Within this chaos lies a peculiar beauty. Unlike purely digital products, retail tech exists in a messy, human world – one where success is measured by metrics, of course, but also the absence of complaints. The most elegant product isn’t the one with the cleanest code or the most impressive AI; it’s the one that works when the Wi-Fi doesn’t, when the user hasn’t slept, and when reality refuses to conform to a carefully plotted customer journey.
I’ve found the best retail tech product managers develop a kind of zen-like mindset. We learn to let go of digital perfection and embrace analog reality. We don’t build for the ideal conditions of the demo environment. We build for the beautiful disaster that is actual retail.
So the next time you’re struggling to operate a seemingly simple piece of technology in a store, know that somewhere a product manager is observing, taking notes, and going back to the drawing board to try once again to bridge the gap between the binary code and the bricks-and-mortar – one humbling iteration at a time.
Things have been quiet on the site lately and the reason is due to some professional news. A couple weeks ago I transitioned into into a new role at REI: Manager of Product Management for Front of House Technology. That’s a mouthful, but basically this newly-created role expands my responsibility beyond my current domain to include leading the product strategy for almost all the technology employees use to keep stores running smoothly and customers use to shop with us. Think point of sale & checkout systems, bike and ski shop services, apps that power sales floor workflows and our growing Re/Supply business.
The new role is a “player-coach” position and it feels like a natural extension of my work over the past few years. I’ll continue to have hands-on product management responsibilities, while taking on the added responsibility that comes with being a people leader. I now lead a team – and while I’ve done this in previous roles – it’s a new realm for me at REI.
I’m approaching this shift with open eyes and open ears. Good leadership isn’t about having all the answers, but about understanding what you don’t know, asking the right questions, listening with empathy, and creating space for honest conversations. It’s with this spirit that I plan to lean into leadership.
We have a lot of work ahead of us. What really gets me hyped is how these interconnected front of house systems — from browsing to checkout to services — have the potential to create a cohesive, amazing experience for our customers, members and employees. We’re not there right now. But we will get there eventually.
The retail technology landscape shifts quickly, and the boundaries between digital and physical experiences continue to blur. At REI, we’re embracing this head-on while staying true to our core mission of connecting people with the outdoors. In many ways, this parallels my own journey of finding balance between technological innovation and mindful living – my personal and professional sweet spot!
As with any new challenge, I’m approaching this role with both excitement and vulnerability. There’s so much to learn, systems to understand, and relationships to build. Things may be quiet around here for a while longer, but I’m super grateful for this opportunity, eager to continue my work with REI, and committed to leading with wholeheartedness. And as capacity allows I’m looking forward to sharing this journey as it unfolds.
I’ve always found capacity planning for UX design and research efforts to be difficult, primarily because sizing tends to be an engineering-focused exercise. This framework by Jeremy Bird looks like it has a lot of potential. Can’t wait to try it out with my team.
Come work with me at REI! If you are passionate about the outdoors & product strategy, our Customer Experience team is hiring a Principal Product Manager who will set the course for key customer journeys, including our foundational offering Co-Op Membership. LMK if you have questions!
Jared White articulating quite clearly the biggest obstacle facing the Open Web:
…never before has The Indie Web been such a glorious platform for building anything you might dream of and sharing it with anyone you like, yet never before has The Corporate Web been so awful and damaging to the body politic. I wish I knew how to deal with this cognitive dissonance, and how to convey to mere mortals out there that the The Indie Web is alive and kicking, and that The Corporate Web doesn’t have to define their experience of being online.
The Open Web has a messaging and onramp problem. There’s no shortage of brilliant technical and engineering minds working on it, but where are the designers and product strategists who might craft the ‘easy enough’ onramps for those who don’t really give a shit about ActivityPub and just want a healthy, constructive and friendly place to share online? Who is communicating the value of the Open Web in compelling ways and using language non-nerds can comprehend?
The recent growth of Bluesky is proof of a collective appetite for something more. Full disclosure, I don’t think Bluesky is the answer, but they definitely understand the onboarding assignment of making the experience easy without introducing dark patterns (yet).
The foundation for a scalable Open Web is here, thanks to the dedication and great work of the developer community that’s gotten us to this point. But to truly realize the potential and impact a universal open web, we need to augment the engineering focus with two additional legs of the stool: design and product. Only then will we be able to understand the problems and needs of the users who aren’t yet here and build the open, accessible and welcoming web of the future.
At work, we started implementing server-driven UI patterns in the iOS app used by REI store employees. This has allowed us to move faster and respond to employee feedback in near-realtime, without shipping app updates. My colleague and engineer-extraordinaire David Allison breaks it down in detail.
As a product manager on REI’s store technology team, I spend a lot of time thinking about how we can make life easier for our employees. Our team builds Ascent, an iOS application that helps more than 10,000 REI store employees access product information and accomplish tasks on sales floors across 180+ locations in the US. Our mission is simple but vital: provide store employees with the tools and information they need, in the moments they need them, so they can do their job with ease, confidence and joy.
Last week, during REI’s internal hackathon, my colleague Seth Daetwiler and I had the opportunity to explore how we might thoughtfully apply AI to support this mission. The result was CarsonAI, a friendly Raccoon process assistant named after environmentalist Rachel Carson (who, like Carson the bot, was also born here in Pittsburgh).
The Problem
Store employees face a common challenge: when they need to understand a process or find best practices, they have to step away from the task at hand, search through dense documentation scattered across various internal systems, hope they find the specific information they need (sometimes it can be one sentence in a 12-page PDF), and then return to their task. This constant context-switching isn’t just frustrating for our employees - it takes valuable time away from why they work at REI in the first place: serving customers, sharing their vast expertise and geeking out over gear.
Can AI Be Thoughtful, Responsible & Humane?
I’ll be honest - I have reservations about AI and I’m extremely conflicted. Like many, I’ve watched its rapid emergence with a mix of fascination and concern. But I’ve come to realize that rather than resist its inevitability, we should focus our energy on finding thoughtful, humane ways to apply this technology where it can create genuine benefits in people’s lives.
I think Carson represents this philosophy in action. Instead of using AI to replace human judgment or save a quick buck, we’re using it to remove pain points that employees have told us exist. The goal isn’t to automate tasks or increase throughput - it’s to give employees better, easier access to the knowledge they need to do their jobs well.
How It Works
Carson integrates directly into Ascent’s interface in two ways:
On the home screen for general process questions
As a contextual “Need Help” button within specific workflows
This means employees can verbally ask Carson questions about workflows and get immediate answers to process questions without leaving their current task. No more hunting through documentation or switching between systems - just quick, relevant guidance exactly when it’s needed.
Building During Hack Days
Working with Seth on Carson was a highlight of the hack days experience. His killer chops in iOS development and user experience design brought the concept to life in ways I hadn’t imagined. When I first floated the idea, his enthusiasm was immediate, and he ran with it, creating an elegant and intuitive interface that makes complex processes more accessible.
Looking Forward
While Carson began as a hackathon project, we’re excited about its potential. Obviously, we aren’t planning to immediately ship a project we threw together over a couple days to production. There are optimizations and tweaks we need to make before opening access to our fleet. Our next steps include:
Testing the prototype with store employees to gather feedback and better understand their needs
Making UX improvements based on feedback
Benchmarking performance and load testing
Exploring a beta release in Ascent during Q1 2025
Learning more about how employees use Carson in real-world situations once it’s live
A Reflection
The retail environment is complex and constantly changing. Our store employees navigate this complexity daily while working hard to provide the best possible experience for our customers. I think tools like Carson represent an opportunity to use emerging technology in a way that genuinely supports our employees - not by replacing their expertise, but by making it easier for them to access and apply it.
At REI, we live and work by a set of values called The Co-Op Way. Three of these values are:
We courageously embrace change
We go further, together
We start from a place of respect
I think Carson is in alignment with these principles. As we continue to explore the possibilities of AI in REI store operations, keeping these Co-Op Way values close will be crucial. The goal isn’t to chase technology for its own sake, but to thoughtfully apply it in ways that responsibly look forward, respect and foster human interactions, and make the work experience better for employees in our stores.
Product management is largely about communicating and selling what you aren’t going to do vs. what you are going to do. This can be effective – if accomplished – but it is often the much harder sell.
Tal Raviv believes the role of product manager is an unfair one, therefore we should work unfairly. I respect Tal greatly and admire his approach, but in my opinion these aren’t unfair tactics. They’re mostly common sense, efficiency-driven and modern PM workflows.
AI startup Anthropic yesterday announced an update to the Claude 3.5 Sonnet large language model that brings a new feature called ‘computer use’ to the forefront of the user experience. Available to developers via the API, users can now direct Claude to use computers like people – surveying open windows and performing operations like moving the mouse cursor, clicking links and buttons, and drafting text.
This is a huge development in the AI space, and one that Anthropic’s rivals in the space are pursuing with great priority. While the tech is nascent, slow and error-prone, the potential is immense. Casey Newton writing for Platformer:
But to use another phrase popular among the AI crowd, the agent that Anthropic released today is as bad as this kind of software will ever be. From this moment on, AI will no longer be limited to what can be typed inside a box. Which means it’s time for the rest of us to start thinking outside that box, too.
I’ve never been a huge proponent or advocate for AI1 , but it’s impossible to deny the impact and influence on our daily lives in the wake of developments like this.
Several months ago, I began using both Claude and ChatGPT to understand how I might use LLMs to improve my professional workflows. Personally, I’ve found Claude to be a better fit for my use cases, which are specific to product management duties such as synthesizing user feedback, analyzing value and impact, and specifying acceptance criteria in technical terms.
With advancements like Anthropic’s ‘computer use’ happening so rapidly in the AI space, it’s daunting to think about what might be coming at us next. One might say the future is already here. I might say we’re perpetually living in it.
I do not use AI to write or develop the content on this website. ↩︎
There can be real art and beauty in the engineering, real art and beauty in the design, and real art and beauty in the solutions we build, and this beauty can contribute to the value and desirability of our products.
I’ve always considered product management to be ~ 70% science and ~ 30% art, and over the years that’s been a somewhat controversial opinion. It’s cool to see Cagan acknowledge this important blend of perspectives.
I’m lucky to work on a product team that builds solutions in close partnership with our users. David Allison (engineering lead on our team) describes our process on the REI Engineering blog. This approach creates such a high signal to noise ratio, and it shows in the work we are able to deliver.
As other retailers divest from curbside order fulfillment, Target continues to go all-in by deepening its ties into mobile ecosystems like Apple CarPlay:
Once connected to CarPlay, the Target app will automatically display the Target store where the purchase was made on the car display. Then, shoppers can view order details, get directions to the store via Apple Maps, and notify the store once they have arrived in their Drive Up parking space.
Target is the only place I do drive-up orders anymore, due in large part to how easy their retail technology makes the experience. Kudos.
It’s release day for my team at REI. We’ve been working tirelessly for the past few months to build a tool that makes pricing product in our stores easier and less painful for employees. Our goal is to replace an archaic, manual & paper-based workflow with a modern, scalable, digitally-supported tool and standard operating process (SOP) that streamlines sales floor operations across the Co-op.
This morning, we deployed the tool to our first group of pilot stores. Exciting! As a data-driven product manager, these days are like Christmas morning. It’s like I woke up to some new datasets under the tree and I can’t wait to unwrap them to see what insights might be inside.
Every now and then I catch myself lamenting that I don’t work on products with millions of users or billions in revenue. But then I catch myself on days like this when I can see the thing we’ve built in the hands of REI’s amazing employees. I can see them using it and the positive impact it makes in their daily lives. I can hear the pain points surfaced in our feedback channels rapidly fade off into the distance. I can watch the thing we’ve built make our stores a better place to work for more than 10,000 people who wear the REI green vest with pride.
I think that kind of direct, measurable impact is something special.
The next few weeks will likely be hectic as the team analyzes and responds to usage data, fields feedback from employees, and optimizes for rollout to all 190 locations. When we get to that point, I’ll circle back with some insights.
Dang, the Bluesky product teams have been crushing it. They just dropped some very good anti-toxicity features like detaching quote posts, hiding replies, improved user controls for notifications & blocking lists.
I’ve been a subscriber to Implications, a monthly Substack newsletter written by Scott Belsky, for some time now. The issues I’ve read so far have been quite enjoyable, as Belsky provides deep analysis that explores what we might expect to come from rapid advancements in technology, shifts in culture, and the evolution of product design & management.
The latest installment includes a section called Insights from the Modern Product Leader and there are some great thought nuggets for product managers to consider as we work on our individual practices.
On the topic of resources versus resourcefulness, and the nuances that exist between them, Belsky writes:
I like to say, if resources are carbs that you can throw at your problems, resourcefulness is muscle that has far longer lasting power and is worth building (despite the pain of doing so).
This is great context for approaching resourcefulness with a growth mindset. It’s something the triad of my product team works on consciously and regularly. Building that muscle is important to ensure our resiliency through ongoing organizational change.
And on product vision, he proposes a thoughtful triad of considerations:
Clarity In Product Strategy: Does every product have a flag planted and a roadmap for how to get there? We should always have a 3-year vision coupled with an annual plan, and your teams should be aligned around what this is throughout your organizations.
Great product teams have a clear strategy and are able to articulate the path to achieve it. Great organizations position product teams to ladder their product strategies up to a broader set of strategic objectives for the company._
Steward The Narrative for Your Segment or Function: The narrative of why your work matters and how your strategy impacts customers is yours to write, share, and iterate.
The internal PR for a product is often overlooked by teams because they are largely, and rightly, focused on executing the work. Telling the story of the work – the impact it will create, the benefits it will provide and the process used to deliver – is an important piece to a product team’s success, and can lend weight to the resiliency efforts outlined above.
Optimistic About Future, Pessimistic About Present: Do you lead with a balance of excitement and vision for the future of a segment/function — and willingness to take big bets — coupled with a pragmatic focus on obstacles and tasks to be done? Are you direct with what is going right and what is going wrong?
This is great. Balance is vital. Great product leaders are simultaneously able to understand long-term goals, but are also realists about current state. They can see either the stepping stones of incrementality and/or the seismic shifts they need to force in order to get to the objectives, and they have the wisdom, empathy, and creativity to understand when to employ each.
One of the greatest rewards we get as product managers is seeing the thing we’ve built create joy in the lives of those who use the product. I don’t consider myself in the operational efficiency business; I’m in the delight creation business.
