All Things AI 2026 in Durham marks the shift from experimentation to real AI execution
The All Things AI 2026 conference came in strong in Durham, North Carolina, and delivered a very clear message: artificial intelligence is no longer that shiny new thing everyone keeps testing just to see what happens.
More than 4,000 professionals showed up for the second edition — a massive jump from the 1,600 attendees at the debut — and it wasn’t just the numbers that stood out. The crowd wasn’t limited to devs and researchers. It was a broad cross-section of the market: people from business, operations, management, and all sorts of other fields trying to figure out, in practical terms, how to fit AI into their daily work.
And that’s exactly where the shift is. The conversation changed its tone. Nobody was asking if AI works anymore — the question now is how to use it properly, with responsibility and real results. The experimentation phase is in the rearview mirror, and what dominated the stages and hallways of Durham was one thing: execution. 🚀
The DNA of All Things Open and the origin of the AI event
To understand why All Things AI has such an accessible and diverse profile, it helps to look back. The event was born as an extension of All Things Open, which has been running for over a decade and is considered one of the best tech conferences in the United States. What has always set ATO apart is the intention behind it. Todd Lewis and his team built a stage where global experts and first-time speakers share the spotlight on equal terms, where early-stage startups sit side by side with established companies, and where voices often overlooked in the tech industry are not just welcome but prioritized.
Sure, there’s an expo floor, there are deals happening in the hallways. But the center of gravity has always leaned more toward education, access, and inclusion than spectacle. And when ATO announced a new conference series focused entirely on artificial intelligence, the expectation was that the same spirit would carry over. According to Tom Snyder, who has organized the annual RIoT Demo Night in partnership with All Things Open for more than ten years, the inaugural edition attracted over 1,600 attendees and received very positive reviews. The second edition confirmed the bet paid off: it sold out, grew, and delivered top-tier content.
The human factor at an artificial intelligence conference
It might seem contradictory, but what stood out most right away at All Things AI 2026 wasn’t the technology — it was the people. At an event dedicated to artificial intelligence, every seat was filled by a human being. Yes, many were actively using AI during the sessions. Yes, one of the keynotes incorporated an AI agent delivering part of the presentation. But nobody sent an agent in their place. The value of being present in the conversations, hearing the questions, and participating in collective learning was unmistakably human.
The scale of the event reinforced something important. This wasn’t a niche gathering. It was a broad sample of professionals trying to understand how AI fits into their work. The first day, for example, featured a business workshop led by Mark Hinkle, a longtime open source advocate and co-organizer of All Things AI alongside Todd Lewis. The room reflected this shift in profile: these weren’t people questioning whether AI works. They were professionals wanting to know how to use it responsibly, effectively, and immediately.
The demand is no longer theoretical. It’s practical, and it’s urgent.
The most memorable moment: when an AI agent took the stage
The highlight of the conference happened during the keynote by whurley, a globally recognized expert in quantum computing and CEO of Strangeworks, based in Austin. After finishing his talk, he walked off the stage — and an AI agent took over to deliver a second presentation that wasn’t on the official schedule. It was a planned reveal, but executed in a way that made the audience pause and pay attention.
whurley later explained that he had built the agent during the flight from Austin to Raleigh, instructing it to listen to the keynote live, incorporate previous talks, and generate a new presentation — including slides — in real time. He stated that he hadn’t seen beforehand what the agent would say.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
At one point, the agent demonstrated how it could triage whurley’s inbox across multiple companies — a moment that drew laughs and, at the same time, a certain unease from the audience. It was equally impressive and unsettling. The capability was clear. So was the risk. And it was, without a doubt, a glimpse of where AI is headed. 🤖
The pattern behind the noise: AI as infrastructure
Conferences like this generate a lot of ideas, but not every idea is a signal. The real value comes from stepping back and identifying the pattern that connects everything. And what emerged in Durham wasn’t just that AI is advancing — it’s that it’s embedding itself more deeply into how we work, make decisions, and organize our businesses.
AI is becoming infrastructure.
The major platforms are no longer tools you open when you need them. They’re environments you start to build on. Your prompts, your workflows, your custom instructions, and your projects over time accumulate into something bigger than simple usage. They become a work system.
And that leads to a second insight that’s easy to overlook but absolutely critical.
The real vendor lock-in isn’t the model — it’s the memory
Every interaction you have with these systems contributes to a growing body of context that exists within a specific platform. That context makes the system more useful, more personalized, and more embedded in your daily operations. But it also makes it harder to leave.
In the cloud computing era, lock-in was about where your applications ran. In the AI era, lock-in is increasingly about which system understands you best. And that raises an uncomfortable but necessary question about portability. If your workflows, your thinking patterns, and your institutional knowledge are captured inside a single platform, how easily can you move them somewhere else?
The agentic AI opportunity
On the business side, another pattern is forming. The major model providers are approaching the market in a way that feels familiar to anyone who lived through the rise of cloud computing. They want to be your primary platform. They want you to standardize on their tools. They want long-term relationships that anchor your operations within their ecosystem.
But agentic AI introduces a different possibility. Agents can orchestrate across multiple models — large and small, specialized and generalist. Because of that, the value may not reside in any single platform. Instead, it could shift to the layer that understands how to apply these tools to real business problems. Systems integrators, consultants, orchestrators, and solution providers could become the central partners, managing complexity on behalf of clients while leveraging multiple AI systems under the hood.
In that model, you’re not buying access to a platform. You’re investing in a capability.
The moment AI left the lab and entered operations
For years, artificial intelligence lived inside a cycle that’s pretty well known in the tech world: an innovation team tests it, presents promising results in a controlled environment, and the project keeps running as a pilot indefinitely. The problem is that a pilot doesn’t generate revenue, doesn’t solve operational bottlenecks, and over time loses internal sponsorship. What All Things AI 2026 signaled very clearly is that this cycle has come to an end. Companies still stuck in the endless-testing mindset are, in practice, falling behind at an accelerating pace, because the ones that have already moved into real execution are reaping concrete competitive advantages that are becoming increasingly hard to recover from.
The panels and presentations at the conference showcased real cases from organizations that went beyond proof of concept. This wasn’t stage talk about the future — it was people sharing what worked, what stalled, and what had to be rebuilt from scratch for real AI adoption to happen within their operations. And that level of transparency was, interestingly, one of the most talked-about aspects among attendees. Seeing companies speak openly about implementation mistakes carried far more credibility than any flawless technical demo could deliver.
Another aspect that became quite evident is that executing artificial intelligence projects at scale isn’t a purely technical challenge. It involves organizational culture change, process realignment, team training, and above all, leadership willing to make decisions in environments of uncertainty. Many of the hallway conversations at the event revolved around exactly this: how to leave the comfort zone of the pilot and commit to real implementation, even when the risks aren’t all mapped out yet.
Beyond experimentation: five practical lessons from Durham
One of the most striking comments heard during the conference challenged a pattern that has become common in many organizations over the past year. Almost everyone has heard some version of leaders encouraging their teams to experiment with AI, explore, be curious. That was a necessary starting point. But it’s no longer enough. If AI is becoming infrastructure, then how you adopt it matters more than whether you adopt it.
Here are five takeaways worth considering:
1. Treat AI platforms as infrastructure, not tools.
If you’re building workflows, generating content, or making decisions with AI on a regular basis, you’re no longer using a tool casually. You’re building on top of a system. That means thinking about durability and portability. Capture your prompts, document your processes, and store your work in neutral formats like Markdown. Keep your most important agents, GPTs, and automated workflows in neutral repositories like GitHub. Treat your interactions as assets that can be reused and adapted, not as disposable conversations locked inside a single interface.
2. Start using voice as part of your workflow.
Voice fundamentally changes how you interact with AI. Instead of typing queries and waiting for responses, you can enter a continuous conversational exchange. This is especially useful during commutes, turning what used to be passive listening into active exploration. Podcasts are inherently one-directional. Voice-based AI is interactive — it lets you ask follow-up questions, challenge assumptions, and refine ideas in real time. It’s a more efficient and, in many cases, more productive way to think and learn. 🎙️
3. Build with one model and validate with another.
Large language models are powerful, but they’re not infallible. Their outputs are driven by probability, not certainty. For any work that matters, it’s worth introducing a second layer of review. Use a different model to critique, validate, or score the output based on criteria you define. This simple practice can reduce errors significantly and improve the overall quality of your work.
4. Be cautious with long-term commitments to a single AI platform.
The instinct to standardize on a single provider is understandable, especially given the cloud computing precedent. However, the emergence of agentic systems suggests a more flexible future where value is created through orchestration across multiple models rather than dependence on one. Before entering long-term agreements, consider whether you’re limiting your ability to adapt as the ecosystem evolves.
5. Move from experimentation to execution within your organization.
If you’re in a leadership position, it’s time to go beyond encouraging curiosity. Effective adoption requires structure. Identify a champion within the organization who is responsible for testing tools, developing workflows, and documenting best practices. Make it an official part of that person’s job — including training others. If that expertise doesn’t exist internally, invest in external support. The organizations that will succeed in this next phase won’t be the ones that experimented the most, but the ones that learned fastest and scaled that knowledge effectively.
Adoption at scale: the real challenge for organizations
Talking about artificial intelligence adoption at scale means talking about something that goes way beyond installing a tool or activating an API. It’s about getting people from different departments, with different levels of tech familiarity and different natural resistance to change, to incorporate AI into their daily workflow in a consistent and productive way. And that, as became quite clear at the conference, is probably the most complex challenge of the entire journey. Not because of a lack of technology — the technology is available, accessible, and increasingly sophisticated — but because of a combination of human, organizational, and cultural factors that no algorithm can solve on its own.
One of the most debated points at the event was the difference between superficial adoption and real adoption. Superficial adoption happens when teams use the tool because they were told to, but work around it whenever they can, or when usage stays limited to low-impact tasks that don’t actually transform operations. Real adoption happens when people understand the practical value of artificial intelligence in the specific context of their work, when they have the autonomy to adapt usage to their needs, and when leadership consistently reinforces the importance of this transformation.
Another relevant angle that came up frequently at All Things AI 2026 was the role of training and continuous upskilling as a driver of adoption. We’re not just talking about teaching people to use a specific tool, but about developing a work mindset that includes artificial intelligence as a natural part of decision-making and problem-solving. Companies that invest in this kind of structured upskilling, with ongoing programs rather than one-off training sessions, reported significantly higher adoption rates and much more consistent results over time. 💡
Infrastructure: the foundation nobody wants to talk about but everyone needs
If there was one topic that came through loud and clear at All Things AI 2026 and that tends to get sidelined in the more glamorous conversations about artificial intelligence, that topic is infrastructure. When a company decides to bring AI into its actual operations, it quickly discovers that the technology foundation supporting the business needs to be up to the task. It doesn’t matter if you have the best models, the best tools, and the most skilled professionals if your data is scattered across silos, your legacy systems don’t talk to each other, and your processing capacity can’t handle the load AI demands.
The experts on hand were pretty direct about this. Investing in infrastructure before scaling any AI solution isn’t a cost — it’s a prerequisite. That includes well-defined data architecture, robust ingestion and processing pipelines, properly sized cloud computing environments, and clear governance and security policies. Without these elements, execution runs into technical roadblocks that can derail the entire initiative.
There’s also an important strategic dimension. AI infrastructure isn’t static — it needs to be designed to grow alongside business demands. The companies that get ahead are the ones building their foundations in a modular and scalable way, so that each new use case can be incorporated without rebuilding everything from scratch. The most mature organizations on this journey no longer think in terms of isolated AI projects — they think in terms of AI platforms that support multiple initiatives simultaneously, with centralized governance and the ability to expand continuously. 🏗️
The final message from Durham
What All Things AI 2026 left as its central message is that we’re living through a real inflection point in the artificial intelligence market. The conference wasn’t about promises or what the technology might do in the future — it was about what’s happening right now, in the trenches, at companies that decided to stop experimenting and start executing for real.
Durham wasn’t just a gathering of people interested in AI. It was a reflection of a broader shift already underway. The technology will keep advancing. New capabilities will emerge. Entire categories of work will be reshaped. But the most important decisions won’t be made by the models. They’ll be made by the people who decide how to use them.
And as that responsibility becomes clearer, so does the opportunity. Because while evolution may be inevitable, leadership is not.
