Ask any random developer at your local coffee shop which AI model they’re currently integrating, and chances are you’ll hear the word “OpenAI” spill from their lips before the steam fades off their oat-milk flat white. This turbo-charged household-name status isn’t mere hearsay—it’s written in the survey data inked into tech history by Vercel’s latest research, interviewing 656 app builders in a digital straw poll for the age of LLMs. But if you think that means OpenAI has no competition on the horizon, it’s time to step away from your codebase and have a look at how the back alleys of AI development are beginning to bustle with new vendors and, shockingly, even a whiff of developer disloyalty.
Let’s just lay it on the table: 87% of these surveyed developers—practically everyone with keyboard access—are currently using OpenAI’s models, and 83% count on its inference APIs to get their ideas out the door. That’s the sort of dominance that, in tech history, could previously only be rivaled by Microsoft’s grip on operating systems, or the collective guilt over never quite finishing that Udemy course on Kubernetes.
However, dear reader, there’s a big twist worthy of a Netflix thriller: developers now report using an average of not one, but two AI providers. What once was a monogamous love affair with one LLM API has turned into a polyamorous relationship with the entire vendor landscape. 60%—that’s right, more than half!—admitted changing their vendor in the last six months. Loyalty in AI, it turns out, is as stable as the versioning system on a JavaScript framework.
Developers are looking for speed, flexibility, and solutions to today’s problems, not a wedding ring from OpenAI. If you’re in software leadership, take note: that threat to your AI vendor lock-in just became real.
And then there’s Microsoft. Despite its multi-billion-dollar open marriage with OpenAI, only 10% of respondents dabble with Microsoft’s own LLMs. Even more damning: 9 in 10 developers don’t consider Microsoft’s models a viable choice at all. Clippy must be weeping into his recycled paperclip bin. It’s a quietly brutal reminder that Microsoft’s history as a “fast follower” in the cloud might not be translating so successfully to the era of custom LLMs.
It turns out that talent, speed, and answering actual developer needs beat brand legacy. Sorry, Redmond, the market is as loyal as a cat with a new food bowl.
Creatives now jump UI/UX/code boundaries like parkour athletes, thanks to tools like Vercel’s v0, Uizard, and Cursor. Suddenly, the difference between a “junior” and “senior” is less about years in the trenches and more about how quickly you can mix-and-match tools and ship something that wows stakeholders in a Slack call.
This shift is weaponizing productivity in a way traditional org charts can’t keep up with. Because really, who needs a meeting about the meeting when your prototyping tool can crank out four variations before the marketing drip campaign even launches? This is the era of the polyglot builder, and they’re quietly routing around old hierarchies like water around a stone.
Dr. Jan Ittner of BCG X strikes the nail home: "An AI writer or developer tool can be more valuable than another hire." Forget building empires—start building MVPs. In many shops, the “AI department” is a Slack channel, not a corner office suite.
It’s an incredibly liberating, if faintly terrifying, prospect for IT professionals who remember when every new technology merited funding a new team. Now, the right API key and a workable credit card limit can do more than three new hires and a six-month onboarding plan.
Instead, developers have their sights set on the next wave: website personalization powered by AI. Only 27% currently offer this, suggesting a ripe field. Next time you grumble about yet another “helpful” floating chat icon, remember: the next generation of AI isn’t looking for more ways to answer “Did you try rebooting?”—it’s figuring out how to make sure you never have to ask that question again.
If you’re in IT, consider this crystal ball: Feature checklists at your next product meeting will get longer, but “basic chatbot” starts to look about as sophisticated as a push notification reminding you to hydrate.
If you wonder how they keep costs low, take a peek: only about 14% bother training their own models. The real action? Retrieval-augmented generation, or RAG (no relation to that tattered t-shirt you wear on weekends), and vector databases—used by 60% of devs. The underlying message is clear: when off-the-shelf intelligence is this good, there’s little sense in spending weeks reinventing the neural-networked wheel.
Of course, you could always let your engineers go wild on custom training and pay for it with your lunch budget. But if you enjoy a regular paycheck—and your company enjoys remaining solvent—today’s wisdom is: parse what you can, personalize what you must, but always respect the line item labeled “cloud hosting.”
But then comes the clincher: when asked if AI will reshape their industry in the next twelve months, the answer jumps to 7.7 out of 10. Translation? Most developers know the marketing noise is thick enough to cut with a butter knife—but they’re hedging their bets that smart, practical use of these tools will still deliver real impact, fast.
The pragmatic takeaway for IT pros: ignore the hype, but don’t ignore the tools. Today’s AI might not be a magic wand, but it sure beats cleaning up after the last failed “digital transformation.”
For the modern IT leader, this means flexibility is king. If the past year has taught us anything, it’s that making any technology decision feel “permanent” guarantees you’ll be proven wrong at the most inconvenient time. “Vendor lock-in” now feels like a quaint problem of the 2010s—today, the fear is more about being locked out of breakthroughs by slow, monolithic processes.
So, if your CTO is busy writing a 12-month roadmap with ink instead of pencil, now might be a good time to buy them a whiteboard and a big eraser.
For IT management, the mandate is clearer than ever: foster a culture where switching gears, switching vendors, and switching up workflows is not a sign of indecision, but a proof of agility. This might mean fewer grand unveilings of “AI Centers of Excellence” and more regular, incremental product pushes.
Will OpenAI stay on top? Maybe. But one thing is certain: the days of picking a single vendor for everything are gone. Now, it’s about having the right mix of APIs, the right blend of off-the-shelf cleverness and custom logic, and the guts to pivot on a dime.
The joke, of course, is that developers are as loyal as alley cats and just as likely to switch when a new can opens. Today’s “default” LLM could be tomorrow’s deprecated repo. Meanwhile, cost-conscious leaders are learning that the real power move isn’t in custom AI departments or moonshot training runs, but in the agile, cross-functional teams who know how to stitch together disparate tools, prioritize features, and keep one hand firmly on the company card.
In other words: the future belongs to those with a short attention span, a healthy skepticism of hype, an allergy to vendor lock-in, and a knack for turning yesterday’s technical curiosity into tomorrow’s revenue stream.
And if you’re still trying to pick a side, remember: in this new wild world of AI, the only real loser is the one who stops building.
Source: inkl OpenAI continues to dominate AI landscape among developers - but things are changing fast
OpenAI: King of the LLM Hill… For Now
Let’s just lay it on the table: 87% of these surveyed developers—practically everyone with keyboard access—are currently using OpenAI’s models, and 83% count on its inference APIs to get their ideas out the door. That’s the sort of dominance that, in tech history, could previously only be rivaled by Microsoft’s grip on operating systems, or the collective guilt over never quite finishing that Udemy course on Kubernetes.However, dear reader, there’s a big twist worthy of a Netflix thriller: developers now report using an average of not one, but two AI providers. What once was a monogamous love affair with one LLM API has turned into a polyamorous relationship with the entire vendor landscape. 60%—that’s right, more than half!—admitted changing their vendor in the last six months. Loyalty in AI, it turns out, is as stable as the versioning system on a JavaScript framework.
Developers are looking for speed, flexibility, and solutions to today’s problems, not a wedding ring from OpenAI. If you’re in software leadership, take note: that threat to your AI vendor lock-in just became real.
DeepSeek, Microsoft, and the Growing Vendor Menagerie
Just a few months ago, if you’d asked an AI enthusiast about popular models, “OpenAI” and “Google” would’ve been safe bets, while “DeepSeek” might’ve sounded like a deep-sea fishing startup. Not anymore. Nearly a third (29%) of surveyed businesses are using DeepSeek—though whether that means a full commitment or a weekend’s side project fling is open to interpretation.And then there’s Microsoft. Despite its multi-billion-dollar open marriage with OpenAI, only 10% of respondents dabble with Microsoft’s own LLMs. Even more damning: 9 in 10 developers don’t consider Microsoft’s models a viable choice at all. Clippy must be weeping into his recycled paperclip bin. It’s a quietly brutal reminder that Microsoft’s history as a “fast follower” in the cloud might not be translating so successfully to the era of custom LLMs.
It turns out that talent, speed, and answering actual developer needs beat brand legacy. Sorry, Redmond, the market is as loyal as a cat with a new food bowl.
The Techno-Democratization of the Workflow
Perhaps the most profound shift isn’t who’s building the models, but who’s using them and how. The phrase “AI is dissolving the boundaries between roles” pops up, which at first sounds like another self-congratulatory line—right until you see what it means in practice.Creatives now jump UI/UX/code boundaries like parkour athletes, thanks to tools like Vercel’s v0, Uizard, and Cursor. Suddenly, the difference between a “junior” and “senior” is less about years in the trenches and more about how quickly you can mix-and-match tools and ship something that wows stakeholders in a Slack call.
This shift is weaponizing productivity in a way traditional org charts can’t keep up with. Because really, who needs a meeting about the meeting when your prototyping tool can crank out four variations before the marketing drip campaign even launches? This is the era of the polyglot builder, and they’re quietly routing around old hierarchies like water around a stone.
The Disappearing AI Team (and Why You Might Not Miss It)
Maybe the most telling number of all: 45% of these shops reported no dedicated AI team at all. Another 57% have no specific AI leadership structure. Instead, the focus turns decisively toward tooling and clear priorities. In other words, instead of writing org charts, tech leaders are writing requirements and letting their best generalists self-organize like an open-source project on a sugar rush.Dr. Jan Ittner of BCG X strikes the nail home: "An AI writer or developer tool can be more valuable than another hire." Forget building empires—start building MVPs. In many shops, the “AI department” is a Slack channel, not a corner office suite.
It’s an incredibly liberating, if faintly terrifying, prospect for IT professionals who remember when every new technology merited funding a new team. Now, the right API key and a workable credit card limit can do more than three new hires and a six-month onboarding plan.
AI Product Features: Support Chatbots Out, Customization In
When developers were asked what they prioritize in AI-powered apps, “customer-facing features” led the way—three quarters, or 75%, rated this as crucial. Contrast that with the currently shrinking hype bubble around traditional support chatbots, included in less than 40% of new projects. It’s as if the world collectively realized one more time that conversation trees alone don’t equal innovation.Instead, developers have their sights set on the next wave: website personalization powered by AI. Only 27% currently offer this, suggesting a ripe field. Next time you grumble about yet another “helpful” floating chat icon, remember: the next generation of AI isn’t looking for more ways to answer “Did you try rebooting?”—it’s figuring out how to make sure you never have to ask that question again.
If you’re in IT, consider this crystal ball: Feature checklists at your next product meeting will get longer, but “basic chatbot” starts to look about as sophisticated as a push notification reminding you to hydrate.
The Pragmatic Developer: Cost Control and DIY Solutions
It turns out “move fast and break things” is evolving into “move fast and stay inside budget.” More than 70% of developers are putting their models through manual testing, which, considering the amount of caffeine and monitor space devoted to AI these days, keeps monthly costs under $1,000. It’s a feat worthy of a standing ovation from any finance department that survived the 2000s SaaS boom.If you wonder how they keep costs low, take a peek: only about 14% bother training their own models. The real action? Retrieval-augmented generation, or RAG (no relation to that tattered t-shirt you wear on weekends), and vector databases—used by 60% of devs. The underlying message is clear: when off-the-shelf intelligence is this good, there’s little sense in spending weeks reinventing the neural-networked wheel.
Of course, you could always let your engineers go wild on custom training and pay for it with your lunch budget. But if you enjoy a regular paycheck—and your company enjoys remaining solvent—today’s wisdom is: parse what you can, personalize what you must, but always respect the line item labeled “cloud hosting.”
Is AI “Overhyped”? Ask a Developer
If you think developer enthusiasm for AI is a never-ending parade of LinkedIn humblebrags and glassy-eyed pronouncements about “changing the world,” let’s temper expectations: when asked how “overhyped” AI currently feels, developers offered a solidly mediocre 6.4 out of 10. Not quite vaporware, but not exactly “moon landing” levels of excitement either.But then comes the clincher: when asked if AI will reshape their industry in the next twelve months, the answer jumps to 7.7 out of 10. Translation? Most developers know the marketing noise is thick enough to cut with a butter knife—but they’re hedging their bets that smart, practical use of these tools will still deliver real impact, fast.
The pragmatic takeaway for IT pros: ignore the hype, but don’t ignore the tools. Today’s AI might not be a magic wand, but it sure beats cleaning up after the last failed “digital transformation.”
The Real Metric: Adaptability Beats Size
Underneath the latest round of vendor bingo and shifting allegiances runs a subtler theme: the best LLM for coding isn’t about the most parameters or even the most integrations—it’s about solving today’s real problems while keeping a wary eye out for tomorrow’s pivot.For the modern IT leader, this means flexibility is king. If the past year has taught us anything, it’s that making any technology decision feel “permanent” guarantees you’ll be proven wrong at the most inconvenient time. “Vendor lock-in” now feels like a quaint problem of the 2010s—today, the fear is more about being locked out of breakthroughs by slow, monolithic processes.
So, if your CTO is busy writing a 12-month roadmap with ink instead of pencil, now might be a good time to buy them a whiteboard and a big eraser.
What’s Next for Developers and IT Teams?
The big story here isn’t a single technology or a single winner. The old battleground of “my LLM’s bigger than yours” is rapidly transforming into “who can ship, iterate, and adapt fastest?” For developers, this means resume skills are migrating from “Support Chatbot Expert” to “RAG-and-Vector-DB Polyglot.”For IT management, the mandate is clearer than ever: foster a culture where switching gears, switching vendors, and switching up workflows is not a sign of indecision, but a proof of agility. This might mean fewer grand unveilings of “AI Centers of Excellence” and more regular, incremental product pushes.
Will OpenAI stay on top? Maybe. But one thing is certain: the days of picking a single vendor for everything are gone. Now, it’s about having the right mix of APIs, the right blend of off-the-shelf cleverness and custom logic, and the guts to pivot on a dime.
The Bottom Line: AI Is Growing Up (Just Not Always How You Expected)
OpenAI is still the alpha in the LLM enclosure, but there’s a growing stampede of alternatives nipping at its heels—and anyone in the sector who isn’t changing things up at least every six months risks missing out on the next software gold rush.The joke, of course, is that developers are as loyal as alley cats and just as likely to switch when a new can opens. Today’s “default” LLM could be tomorrow’s deprecated repo. Meanwhile, cost-conscious leaders are learning that the real power move isn’t in custom AI departments or moonshot training runs, but in the agile, cross-functional teams who know how to stitch together disparate tools, prioritize features, and keep one hand firmly on the company card.
In other words: the future belongs to those with a short attention span, a healthy skepticism of hype, an allergy to vendor lock-in, and a knack for turning yesterday’s technical curiosity into tomorrow’s revenue stream.
And if you’re still trying to pick a side, remember: in this new wild world of AI, the only real loser is the one who stops building.
Source: inkl OpenAI continues to dominate AI landscape among developers - but things are changing fast
