Hacker News
Despite predictions, agentic AI has struggled with enterprise-level implementation due to complexities in security, governance, and scalability. While companies show curiosity, most remain in the experimentation phase, suggesting 2026 may still be a year of pursuit rather than widespread success.
AI 生成摘要
儘管預測不斷，代理式AI因安全、治理和擴展性等複雜性，在企業級實施上面臨挑戰。儘管企業對此充滿好奇，但大多數仍處於實驗階段，顯示2026年可能仍是追求而非廣泛成功的時期。
If you keep predicting something every year, sooner or later you may be right. At the end of last year, we kept hearing 2025 was going to be the year when agentic AI really took off. It didn’t really work out as predicted. It’s all well and good to talk about the potential of agents, but the fact is that it’s complex to implement this kind of technology for all the reasons it’s hard to implement any tech in the enterprise.
It may sound like a broken record, but we still need security, governance, observability, identity and authentication. It's easy to talk about this stuff, but it’s much harder to put it together and make it work. As 2025 closes, we are hearing that, yes, 2026 is going to be the year that agents scale and ROI fairy dust gets sprinkled everywhere, but so far that has mostly eluded everyone.
Study after study tells us that companies are still struggling to implement AI at scale. For example, the McKinsey State of AI in 2025 report from November found lots of agentic curiosity from business leaders with “sixty-two percent of survey respondents saying their organizations are at least experimenting with AI agents.” Yet that didn’t translate into agentic success at scale. “Most organizations are still in the experimentation or piloting phase: Nearly two-thirds of respondents say their organizations have not yet begun scaling AI across the enterprise,” the report found.
Vendors see the interest and sense there is something real here, but they could be victims of their own hype. Last year at Dreamforce, Salesforce pushed the idea of how easy it would be to implement agents. As the company’s Dreamforce 2024 recap put it, “Agentforce enables companies to augment their employees and scale their workforces on demand with a few clicks.” This year CEO Marc Benioff acknowledged enterprise complexity is real in an interview with CNBC’s Jim Cramer. While he didn’t come out and say it, he implied that the few clicks statement may have been an exaggeration.
“The speed of innovation, [which is] huge, is outstripping customer adoption. These customers have to go back and modify massive architectures they have and systems they're running—the world's most important companies, mission-critical systems,” Benioff told Cramer.
The dream for agents is to move work across systems from multiple vendors in an automated way, but it’s hard to do that. As Workday CIO Rani Johnson told me in an interview earlier this year, she has been confining agents to a single domain because it’s much more challenging to move between applications and platforms. “We've not been able to start to do what we call that cross-domain yet, because the passing of access control and identity cross-domain gets much more complex,” she said.
And there’s the wrinkle, right? But that didn’t stop Accenture CEO Julie Sweet in an interview with Jon Fortt from predicting that AI agent products will finally flourish this coming year and deliver big outcomes. “So I think the defining trend for next year will be seeing scaled value [when it comes to AI agents],” Sweet said in the interview. She sees companies shifting from looking at AI purely as tech to something that is going to change how work gets done in their organizations.
“And that is an inflection point where organizations are really saying we're going to take on transforming the whole enterprise and getting value from it. And so that's what I think 2026 will be,” she said.
But if we’re being honest, that’s exactly what we heard at the end of last year too — and it didn’t happen. Maybe it was the hype getting ahead of reality, but for whatever reason, it didn’t work out as planned in 2025. The question is whether companies will really start to see value from AI in 2026, or whether we will be singing the same song at this time next year. It’s going to be interesting to watch and find out.
Featured photo by Getty Images for Unsplash+.
We keep hearing how AI agents are going to dominate in the enterprise, but so far the hype has not matched the reality. In part, that’s because implementing AI agents is complex, and many companies are struggling to build them at scale. Enterprise software companies like ServiceNow have been
If there's a light up ahead, well brother I don't know
But I got this fever burning in my soul
So let's take the good times as they go
And I'll meet you further on up the road
~Bruce Springsteen, Further on
A year ago this week, I launched FastForward #1 at AWS re:Invent. It was a heck of a place to officially start my journey, amidst the tumult of a ginormous conference in one of the most distracting places on the planet, but we were ready to go and there
When Amazon announced Bedrock AgentCore in July, it felt significant. Here was a major cloud platform adding a component-based way to build agents with many of the pieces a developer would need to feel comfortable building something in an enterprise setting. This included runtime, browser, observability, identity, memory management components,
The enterprise tech newsletter for forward thinkers