A bit of a longer post today given all the platform deals that have been taking place (Waymo partnering with Uber, ChatGPT rolling out services from booking hotels to buying physical items, etc), that I wanted to share my experience having done these types of deals along with key learnings.
I’ve been reading daily, both articles and through posts on LinkedIn, about OpenAI enabling new services via ChatGPT, everything from being able to get an Uber, to book a reservation at a restaurant to ordering groceries via Instacart. None of this is surprising since when OpenAI launched ChatGPT plugins (and later custom GPT apps), they weren’t shy about the ambition. Sam Altman effectively hinted at an “App Store for AI” being a future where you could order a rideshare, book a table, or grocery shop all through ChatGPT’s interface, without ever touching a separate app. The idea, in theory, would be an AI App Store that would unify countless services behind one conversational interface (ChatGPT), sparing us the jumble of apps on our phones. In practice, however, the execution to-date feels like its fallen short ultimately because OpenAI’s strategy emphasized breadth over depth, integrating dozens of partners quickly, at the expense of building truly seamless experiences for the end user. The result is mostly a collection of clunky, half-baked plugin interactions that feel more akin to affiliate marketing than a truly end-to-end integration.
Every new platform faces a classic dilemma whether to go broad fast or go deep on a few core use cases. Google went deep for example with products like Google Maps enabling only affiliate type links for 3rd party partners. OpenAI clearly chose to go broad, opening a wide range of ChatGPT plugins in months. Unfortunately, many of these integrations lack the tight product coupling needed to make them useful. The user experience feels disjointed. This isn’t a new problem. I’ve seen the pitfalls of “breadth over depth” play out before, across various partnership bets in tech. In fact, five firsthand partnership anecdotes from my career illustrate why shallow integrations underwhelm, and why deep, end-to-end integrations are what truly drives the best experience for the customer. Each story, from ride-hailing to meal kits, carries a lesson that OpenAI (or other platforms) would do well to heed.
Uber × Lime – Deep Integration Done Right
One of the most successful integrations I’ve scaled was the partnership between Uber and Lime. With this collaboration, depth was key. Uber didn’t just place a deep link in their app to download the Lime app if a customer wanted to take one of our scooters. Instead, Uber embedded Lime’s experience natively inside the Uber app. If a user wanted a scooter or bike, they could find, unlock, and pay for a Lime all through the Uber interface they already knew and trusted. Furthermore, Uber’s existing 190M customers didn’t have to go through a new sign up flow, put in new payment details, etc. To the rider, it felt like one cohesive product.
That level of integration required significant coordination though between both of our companies. We had to align product roadmaps, pricing plans, synchronize back-end systems, iron out SOP’s around customer support, etc. But the payoff was huge. Riders embraced the feature because it was convenient and intuitive – essentially an Uber experience. We attained access to new riders and Uber was able to create more LTV from existing ride share customers who now could take bikes and scooters. The data spoke volumes as Lime’s ridership and revenue surged thanks in part to the Uber integration.
The Uber x Lime story shows how powerful a partnership can be when you prioritize the UX. The companies invested the time and resources to make the integration feel native. As a result, riders used it. It wasn’t just a press release partnership or an affiliate marketing program, but rather it became a daily-use feature for thousands. This is the gold standard that an “AI App Store” like ChatGPT should strive for, an experience so well integrated that the user barely notices or cares that multiple services are working in concert. The reason why Uber has been successful with many of their integrations is they only do end-to-end deals like the one I discussed with Lime. They also do less deals – instead often focused on finding one leading partner in each category to work with as this allows both product teams to go deeper on the experience. Not all partnerships go this way, however, to see the contrast, consider another one of Lime’s integrations with Google Maps.
Google Maps × Lime — A Shallow Link that Falls Flat
Around the same time Lime was integrating with Uber, we also teamed up with Google to surface Lime scooters in their Maps app. It seemed like a no-brainer since Google Maps is where people plan journeys, so why not show nearby Lime vehicles there offering another mode of transportation? Google, however, sees themselves as more of a utility. They historically have shied away from end-to-end integrations like the kind we had with Uber given the complexity and competing priorities across Google’s breadth of many products. P0 of this integration entailed Google Maps showing a little scooter icon at the bottom of the maps app and the distance to the nearest Lime, but the moment you wanted to actually rent it, you were bounced out to the Lime app (or to the app store if you didn’t have it installed). In other words, it was essentially a deep-link or affiliate type program.
For users, this wasn’t the ideal experience though. Imagine planning a route in Google Maps and seeing a scooter option; you tap it thinking you can grab a quick ride, and suddenly you’re forced to pause, install or open another app, sign up, enter payment details, and then resume your journey in a different interface. Many would-be riders simply didn’t bother. The drop-off from tap to actual ride was significant and conversion rates were lower than we saw with the Uber integration, where everything stayed in one app. We learned that even a tiny bit of friction can kill an otherwise good idea.
To Google’s credit, over time they did enable slightly tighter links (for example, once Lime accounts were linked, the handoff to the Lime app could be smoother). But it never reached the level of seamlessness that was achieved with our Uber deal. Google Maps’ integration emphasized breadth by aggregating dozens of mobility providers globally in their app, but none of those were deeply woven into the Maps experience.
The lesson here is that a shallow partnership can create the appearance of integration without delivering the substance. Google could tout having lots of partners, but if each one is just an external link, the real utility is limited. Users won’t jump through hoops for a marginal convenience. This is a cautionary tale for ChatGPT’s plugin marketplace: simply listing a service as “integrated” means little if the actual user flow still makes people work for it. Next, let’s consider another platform deal I did from another industry, one that mirrors the ChatGPT scenario of trying to be an all-in-one hub for many services.
Verizon +Play — When More Isn’t More
In 2022, Verizon launched +Play, a digital marketplace where Verizon customers could manage a slew of subscription services in one place. Their vision was to think of +Play as Verizon’s attempt at a mega bundle. They were already offering Netflix and Apple Music for new wireless subscribers and so they wanted to create a larger collection of benefits. After launching, they offered Netflix, Disney+, Peloton, Calm, gaming subscriptions, meal kits, among many other services. On paper, it was the ultimate breadth play: over time Verizon signed up dozens of partners and proudly put their logos on a slide. The strategy was to become a one-stop shop for digital subscriptions, boosting Verizon’s value proposition and subsequently earning referral fees along the way. If that sounds conceptually similar to an app store or ChatGPT’s plugin store, it is.
However, Verizon’s approach lacked depth, and a cohesive UX. +Play was essentially a portal with single sign-on and unified billing. But the user experience remained fragmented. As an example, after you “purchased” a service through +Play, you still often consumed that service in its own app or website. As one sell-side analyst noted, “the streaming partners continued to own the user experience.” Verizon wasn’t integrating the content or functionality into a unified interface. In practice, +Play was basically a large affiliate program with discounts. Users could discover and sign up for services there, but once signed up, they’d be kicked over to each provider’s app to use it. Nothing really tied the experiences together beyond the Verizon billing account.
The result? +Play has struggled based on observations. That said, during my time at Blue Apron, knowing these UX challenges, I built an entirely different type of integration with them. We offered a unique specialized meal kit subscription exclusivity through +Play. This program called Blue Apron +, was designed to attract new users with a set of perks not available through our direct-to-consumer offering and it worked. It became a large acquisition driver for the company. We kept our loyalty program on our native app separate by design, however. This is table stakes if you’re building a platform partnership. Keep your loyalty program (and therefore emerald / high LTV customers) transacting with you directly. When Walmart announced they were going to start enabling some inventory to be shoppable on ChatGPT, they purposefully did not allow any Walmart+ benefits to transfer over. In the Uber deal I did with Lime, I kept our Lime Prime loyalty program unique to the Lime app. The myriad of partnerships that Verizon were doing with +Play looked impressive in a press release, yet a “marketplace” without a superior UX doesn’t usually draw people in for long. From a partnership perspective, Verizon achieved breadth via a lot of deals, but not the kind of integrated product that drives longer term engagement.
This is a fate that OpenAI should strive to avoid. ChatGPT’s plugin store could easily become a similar graveyard of well-intentioned but seldom-used integrations if the focus remains on quantity over quality. Just as Verizon learned, merely aggregating services under one roof doesn’t create a great platform; you have to mirror the experience on the partner’s native app as closely as possible to get traction.
Here are two more examples of partnerships I spearheaded from the world of meal kits and health/voice tech, which further underscore the challenges of a shallow integration.
Blue Apron × Blue Cross Blue Shield — A Missed Opportunity
I bring this up given the news recently about ChatGPT Health. As Blue Apron’s head of BD, I brokered a nationwide healthy food partnership with Blue Cross Blue Shield (BCBS) that, in hindsight, stayed too superficial. The idea was promising, though – healthy meal kits for BCBS’s insurance members. We envisioned an initiative where Blue Apron’s meal plans could tie into health goals (think a recipe suggestion for diabetics or rewards from your insurer for eating healthy). It could have been a powerful trio of brand, product, and healthy lifestyle, but due to various constraints especially on the insurer’s end, what we executed was far more basic. Essentially, BCBS offered its members a promotional discount on Blue Apron subscriptions turning the integration into a marketing partnership, not a product integration. Like the program from Verizon, this experience was another rendition of an affiliate marketing program. Sure, we got Blue Apron in front of a broad new audience (millions of BCBS members), achieving breadth in terms of reach. But there was minimal integration of the user experience or data between the companies. It was basically a referral program with a co-brand wrapper. If a member redeemed the offer, from that point on they were just a regular Blue Apron customer – their health info, dietary needs, or insurer incentives never entered the picture. We didn’t, for example, sync up with any BCBS app or wellness program; we didn’t create tailored meal plans for specific health conditions. In short, we left a lot of value on the table. And this was not by choice, it was because BCBS is first and foremost in the business of insurance. They are an old guard insurer with a ton of technical and regulatory constraints. The theme here, like Verizon, is that truly building a platform and branching out into new areas requires significant resources, talent and c-suite buy-in. It can’t be a “side” project of the company. I’ve watched numerous platform deals fail because one side doesn’t want to invest the effort to make it successful.
In the context of ChatGPT Health, imagine a scenario: say ChatGPT integrates with a healthcare service like ZocDoc and can offer the ability to book a doctor appointment all in ChatGPT. If it just dumps a link or a phone number at the user, that’s like our BCBS deal, it’s a dead-end. A deeper integration would let the user book the appointment within ChatGPT, verify insurance info, co-pays, get reminders, etc. Without that depth, the feature becomes a gimmick rather than a truly novel platform.
My last anecdote deals with voice assistants which is another area where many partnerships ended up more novelty than utility.
Blue Apron × Alexa — Commerce Challenges with Voice
In 2022, Blue Apron made headlines after I created a partnership integrating with Amazon’s Alexa voice assistant aptly named ‘Alexa in the Kitchen.’ You could re-order your Blue Apron meal kits simply by talking to your Echo device. It was announced with a lot of press: voice ordering was cutting-edge, Alexa was in millions of homes, and we were the first meal kit company to do this. The partnership even gave our stock a short-lived boost (causing a ~10% jump on the news of the Alexa deal). From a PR perspective, this was great. From a user perspective, however, there were many challenges.
The Alexa integration was technically functional. Customers could link their Blue Apron account and say, “Alexa, order my next Blue Apron.” But it offered little functionality beyond that. The reality is, meal kits are a visual and planning-oriented product. Customers like to browse recipes, choose their meals for the week, and adjust things which are all tasks better suited to a screen than voice. Voice ordering was practical for a repeat of a past order (and we did see good numbers here) but it didn’t support selection of new recipes and it didn’t address any major pain point in the customer journey. How many people really need to reorder meals by voice when they’re likely checking the app for recipe details? The integration wasn’t sticky because it wasn’t solving a real problem or making the experience easier than the status quo.
Depth Over Breadth: Lessons for AI Platforms
Whether it’s ChatGPT’s plugin ecosystem or any platform strategy, the takeaway from these stories is consistent, deep integrations surpass wide but shallow ones. Users don’t care how many partners or plugins you’ve onboarded if each experience is clunky or limited. They will however care if one or two integrations absolutely nail their needs with a smooth, delightful experience. In platform building, just as in business development, there’s a temptation to celebrate quantity: “We have 100+ plugins at launch” or “We’ve signed deals with half the Fortune 500” But if those connections prove thin then the platform’s value to the end-user remains weak.
OpenAI’s ChatGPT plugins currently feel more like the Google Maps or Verizon +Play approach, that is a plethora of options, few of which feel truly integrated. For example, using the Uber plugin in ChatGPT can yield a fare estimate, but then it just hands you off to the Uber app to actually book the ride. That’s essentially the Lime in GoogleMaps experience all over again. To make ChatGPT’s app store vision work, OpenAI will need to pick a few flagship use cases and go much deeper: collaborate with partners on sharing data and functionality, streamline account linking and permissions, and refine the conversational UX so that a user can accomplish a task start-to-finish in one place. That likely means less is more which means multi-year category exclusive deals and the development of a ChatGPT loyalty program to create stickiness across the services they are offering.
For product and BD leaders more broadly, the mandate is clear. Before chasing the next big integration or signing the next partner logo, ask: What will the actual UX be? Will this integration make the user’s life easier, how will I protect my most loyal customers from cannibalization if I open a new distribution channel on another platform? If the answer is “well, the user will still have to do X, Y, Z manually” or “the benefit is a small coupon,” I would suggest waiting until the platform can enable the functionality you want.
ChatGPT’s AI app store ambition isn’t doomed by any means, it just needs a course correction. There’s immense potential in AI agents coordinating our digital lives, but the key is execution. The companies that win will be those that prioritize the UX, enable few but deeper integrations and who have a scalable playbook on how to cross promote other services adding overall cohesion and stickiness to the platform. It’s a lesson I’ve learned at the margins – sometimes painfully – and one that I hope the next generation of AI platforms takes to heart as they build the future.