Howard Street in San Francisco was all abuzz in November, but not with the usual cars and Muni buses flying across downtown. Last week was Microsoft Ignite, one of the mainstay conferences of the tech industry that brings together countless enterprises and their people.
Of course, Stack Overflow was not going to miss it. So our boots and ears were on the ground in rainy San Francisco, seeing and hearing about the newest products, innovations, and ways of thinking in the tech world.
It’s probably no surprise to you that the standout topic at Ignite was AI and the agents that come along with it. But there was an unmistakable difference in the air from the standard agentic AI discourse—and the enterprise marketing materials—at Ignite’s sprawling Hub this year. Enterprises are becoming more focused and steady in their AI strategies, pointing to a deeper need for market fit and proof to customers that this strategy actually works.
AI for the market you know and the customers you already have
In the early days of the AI hype cycle when none of us were really sure where AI would take us, the bullish among us were convinced that LLMs and agents would completely transform every aspect of modern life. The AI gold rush had many companies, from enterprise to startup, racing to go to market with the latest and greatest tool. Because of this, it seemed—allegedly—like many companies were marketing AI features that didn’t yet exist, and were perhaps a little bit outside of their typical market.
This is one of the changes that was evident at Ignite. The race may not be over, but it seems less like the story of the Chinese Zodiac race and more like the tortoise and the hare. Of course, enterprises are still pushing their AI strategies and initiatives forward, but they seem overall slower, and in turn, steadier in their work.
The steadiness comes from a better understanding of market fit and niche. I heard—several times, in fact—that the AI feature I was demoing was actually an AI-powered user interface sitting on top of an existing solution. Many of the agents I saw had quite impressive use cases—such as one VR demo I had for a digital twin within a factory setting—with the potential to really help the work of customers. But ultimately, many of these solutions are just amped up versions of what companies already offered, digital twin included.
One such example was Docusign’s AI-powered forms, which were (as you might guess) very powerful autofillable forms that also addressed many of the drop-off points for users. There were some impressive stats in their presentation about efficiency, cross-collaboration, funnel retention, and reducing friction, which ultimately painted a wonderful picture of Docusign using AI to address real problems that their clients were facing around data, documentation, and security.
The same could be said for other industries, from manufacturing to entertainment, where many of the AI solutions being offered target niche user issues, rather than a blanket-case—and often unneeded—chatbot deployed across the company.
Goodbye hype, hello proof
Another cogent aspect of many of the interactions I had at Ignite was that it’s not enough to have an impressive mission statement and potential solution—now you have to prove it. Many companies are doing just that, showing that AI is well past its early hype stage, where a cool idea was enough to get you into the door.
It was particularly interesting for me to see how many use cases and proof points companies seemed to have around their AI initiatives, showing real-world results for agents and MCP servers that just a year ago were buzzy phrases getting tossed around…if that. The quick innovation and deployment of AI solutions shows how quickly the industry was ready to adopt and utilize these capabilities, and how hungry customers were to receive them.
While the flow charts and KPI talks weren’t the most exciting information I absorbed, they did seem to show that a sort of equilibrium is happening within the AI ecosystem—one that asks for proof from the largest enterprises, and that gets proof in return.
Capabilities, not just automation
One of the best conversations I had at Ignite was with an executive from Microsoft, who spoke to the way that AI is increasing the capabilities of humans, rather than just replacing them. This seems to align with the bell curve of things in a tech-hype world, which is starting to peter off on its excitement that AI will replace every human, everywhere, for all time. Not to say that many companies aren’t still moving forward with their bots-replace-humans agenda (if the latest rounds of mass layoffs are any indication), but it’s a slight change in rhetoric that many are now highlighting the power of human-in-the-loop, rather than the inherent power of the AI solutions alone.
Many of the use cases being presented at Ignite were those that needed the helping hand of subject matter experts. One of these was the case of the solution I discussed with the Microsoft executive. The solution was focused on entertainment and media, taking highlight clips and using AI to personalize the experience for football fans across the world, including language translation for an international audience. Still, the solution would need moderators, editors, and on-the-ground reporters to make it work, with the AI simply delivering this content using its powerful algorithms and predictions to make the experience deeply personalized and curated.
Another example of this was the work by Canary Speech, a company that partners with hospitals to bring voice biomarker technology to doctors to help with early detection of diseases and disorders like Parkinson's and depression. Their technology augmented and supercharged the work already being done by doctors, enhancing their capabilities as practitioners while still deferring to their subject matter expertise.
Now we need infrastructure
For the AI-native companies—and there were quite a few—many of the solutions being presented were in support of the AI initiatives of other enterprises. This regularly included visibility, governance, security, and data management, all key factors to an enterprise’s success with AI. Many of these solutions promised faster testing, automated systems management, and risk detection, allowing companies to work faster and smarter while staying reliable in an AI-world. For me, this revealed a clear push for scale, where AI tools now need to be supported by actual infrastructure for successful implementation.
Much of the marketing I saw highlighted the same roadblocks that came up in conversation at Ignite. Scaling any new technology is not easy, and many of the enterprises present at Ignite were feeling the pain. Some of the phrases I saw include let real intelligence flow and what real agentic power feels like. Perhaps more brashly by one company, make AI real was the slogan they went for. The fact the word “real” came up often in the marketing collateral on the floor of the Hub points to the fact that making agentic AI “real” is not as easy as it seems.
Amongst the noise of the conference, it was clear that security and data are of the utmost importance for companies deploying AI internally and for customers. And the ability to keep data managed and AI applications secure is no small feat for enterprises. Being able to provide governance for these solutions is such a huge weight that they are willing to outsource this work to AI-native companies, whose market niche is providing visibility and efficiency within AI solutions.
It’s no longer about the early adopters
In line with the offerings for security, management, and governance, conversations I had on the floor showed that the AI hype has moved past early adopters…and perhaps even the late majority.
One conversation I had centered around how a company was attempting to bring AI to their most skeptical customers by proving the security of their tools. Reaching this final audience of AI adopters—the ones most suspicious of the technology and least likely to use it—will be a challenge for many enterprises, especially with how young the technology still is. But in combination with the many proof points and case studies that enterprises already have around their AI solutions, it seems to be a challenge that enterprises are actively taking on.
Ultimately, the targeting of the most skeptical possible audience for AI was another indication that companies are settling into their niches, and creating AI specifically for their markets and the challenges these markets face. In order to get these late-stage adopters, enterprises are attempting to meet these customers where they are, providing solutions that easily fit into the work they’re already doing, with proof that the technology actually works.
The buzz at Ignite made it clear that the AI bubble has not yet popped, but that the hype has calmed enough for enterprises to return their focus to what they’re already good at and what they know their customers need. As the technology continues to settle across the industry, it’ll be interesting to see how companies scale, reach their most skeptical customers, and prove the successes of their AI initiatives, especially as the word “AI” loses its heady buzz.