Derek Mills

The Agentic AI Hype Cycle Is Here and Most Businesses Aren't Ready for It

February 26, 2026

I want to be clear about something before I say anything else: I think agentic AI is real. I think it matters. I think it is going to change how businesses operate in ways that most people are still dramatically underestimating. I also think the current moment - right now, in early 2026 - is a disaster waiting to happen for a huge percentage of the businesses trying to get involved with it. These two things are not contradictions. They are the same story.

This is exactly how The Last Jedi gets misread, by the way. People call it a betrayal because it delivers something more complicated than they wanted. The hype set up an expectation the reality couldn't match. Agentic AI is doing the same thing to the business world, and the businesses that can't tell the difference between the marketing and the product are going to get burned.

What Is Actually Happening Right Now

Here is the shape of the moment. AI agents and AI-ready data are the two fastest advancing technologies on the 2025 Gartner Hype Cycle for Artificial Intelligence, experiencing heightened interest, accompanied by ambitious projections and speculative promises - placing them squarely at the Peak of Inflated Expectations. That phrase, "Peak of Inflated Expectations," is the most important phrase in this entire conversation, and almost nobody is treating it that way.

The global agentic AI market is expected to be worth around $196.6 billion by 2034, up from $5.2 billion in 2024, growing at a CAGR of 43.8%. Those numbers will make you feel like you're missing something if you don't get in now. That is exactly the feeling you should be suspicious of.

Despite strong intent, only 2% of organizations had deployed agentic AI at scale by 2025, while 61% were still in exploration phases. Let that sit for a second. Sixty-one percent still exploring. Two percent actually doing it. But every vendor in the space is speaking as if deployment is already mainstream. Tory forwarded me one of these vendor emails last week - something about "autonomous digital workers" - and I asked him if he knew what it actually did. He said it sounded like it wrote his emails for him. It did not write his emails for him.

The Agent Washing Problem Is Worse Than People Realize

Here is the specific thing that should make every business owner angry: many vendors are contributing to the hype by engaging in "agent washing" - the rebranding of existing products, such as AI assistants, robotic process automation (RPA) and chatbots, without substantial agentic capabilities. And this is not a small fringe problem. Gartner estimates only about 130 of the thousands of agentic AI vendors are real.

Thousands of vendors. One hundred and thirty legitimate ones. Those are genuinely alarming odds if you are trying to buy something that actually works.

One executive put it plainly: "In 2025, we observed widespread agent-washing: existing RPA scripts and assistants were relabelled as 'agents' without true autonomy, decision boundaries, or accountability. If a system cannot safely act, adapt, and recover without constant human babysitting, it is not agentic. It is simply automation with better marketing."

"Automation with better marketing." Someone put that in a pitch deck and I guarantee it worked.

Common tactics include rebranding existing features as "AI agents" - like calling a call recording feature a "transcription agent" - alongside vague claims of "full autonomy" without specifics on decision logic or operational guardrails. If you've seen a vendor demo in the last six months and walked out thinking "wait, did they just show me a fancier Zapier?", trust that instinct. You were probably right.

The Real Numbers on the "Fastest-Growing Product Ever"

The most useful case study for understanding this hype cycle is Salesforce's Agentforce, because it's the most visible and the most honest - at least in the data, if not always in the press releases.

Marc Benioff dubbed 2025 "the year of Agentforce," with promises to change the agentic enterprise. And by the numbers they're reporting, it does look like progress. Agentforce reached an annual recurring revenue growth of 330% year-over-year, which is the kind of figure that ends up in a lot of breathless LinkedIn posts.

But here's the other number. Agentforce adoption metrics appear strong in absolute terms, but they represent only a small percentage of Salesforce's 150,000-customer base - around 12% have signed agreements. Twelve percent. And that's among companies already inside Salesforce's ecosystem, who have a financial relationship with them, who have been pitched Agentforce at every conference and email and event this year.

The real roadblocks to enterprise adoption include unclear pricing, messy orgs, and weak enablement. Which means even the companies that want to deploy this stuff are running into practical walls. This is not a theoretical problem. The original $2-per-conversation pricing model was prohibitive - it immediately priced out nonprofits, SMBs, and any organization that did not have a large enough AI budget. They've since reworked it to Flex Credits at $0.10 per action, but the instability of the model itself tells you something about how ready for primetime this actually is.

I was playing with one of the newer sales automation tools recently - trying to get it to handle multi-step outreach sequences the way a human would. It was impressive for about fifteen minutes. Then it got confused about context from earlier in a thread and started doing something that would have been a disaster if I hadn't been watching it. It reminded me of R2-D2 trying to navigate the Death Star - useful, essential even, but you better have someone keeping an eye on it.

Why 57% of Organizations Are Already Set Up to Fail

Here's the structural problem that nobody in the hype cycle is talking about loudly enough. 57% of organizations estimate their data is not AI-ready. More than half. And this matters enormously because agentic AI doesn't run on enthusiasm - it runs on data. Without high-quality, well-structured and accessible data, even the best AI model will fail to deliver.

"Most agentic AI projects right now are early stage experiments or proof of concepts that are mostly driven by hype and are often misapplied," said Anushree Verma, Senior Director Analyst at Gartner. "This can blind organizations to the real cost and complexity of deploying AI agents at scale, stalling projects from moving into production."

The consequence of all this? Over 40% of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value, or inadequate risk controls. Four out of ten. That is an enormous failure rate for a technology that is being sold as inevitable.

And it tracks with what we've actually seen. We stress-tested our own software stack a while back, and the thing we kept running into was that the more ambitious the automation, the more you need clean data and airtight process documentation underneath it. Most businesses don't have that. Not because they're lazy - because nobody told them they needed it before the agents arrived.

Linda mentioned that Gerald has been running the same Excel-based inventory system for eleven years and his company just bought an "AI agent" to modernize their workflow. She seemed proud. I gently asked what their data structure looked like. She said she'd ask him. I am not optimistic about the answer.

The Businesses That Are Actually Winning

Here's the thing - and I want to be precise about this because I don't want to be the person who just rains on everything - the businesses getting real value from agentic AI right now have two things in common. First, they started with a specific, bounded problem. Not "let's be an AI-first company." Second, they already had their data in order.

Consumer-facing applications tend to be complex, messy, and unforgiving of errors - not ideal for agentic AI. Back-end operations, however, are a better fit. That's not a knock on the technology. That's just an honest assessment of where it is right now.

IBM's Danilevsky is quick to ground expectations: "Enterprises need to be careful to not become the hammer in search of a nail." She's right. And the hammer-looking-for-a-nail pattern is exactly what's happening at a lot of companies right now. They've been told agents will transform their business, so they're trying to find something - anything - to transform, whether the problem warrants it or not.

The honest vendor conversations are the ones where they tell you what the product actually can't do yet. Those vendors are rare right now, but they exist.

What the Hype Cycle Means for People Running Actual Businesses

The businesses that treat agentic AI as a competitive emergency right now - rushing implementations, buying whatever a vendor calls an "agent," deploying before their data infrastructure can support it - are going to make a painful and expensive mistake. "Most agentic AI propositions lack significant value or return on investment, as current models don't have the maturity and agency to autonomously achieve complex business goals or follow nuanced instructions over time." That's Gartner, not me. They're not anti-AI. They're telling you to slow down.

The businesses that win this cycle will be the ones that invest right now in the boring prerequisite work. Clean data. Documented processes. Understanding which workflows are actually candidates for automation versus which ones just sound like they should be. "The value is going to be with those organizations that take their private data and organize it in such a way so that the agents are researching against your documents."

This is the Rey's arc problem, honestly. People wanted her to have a prestigious bloodline because the hype set up that expectation. The actual story was about someone building something from scratch with no inherited advantage - which is a better story if you're willing to sit with the discomfort of it. Agentic AI has a better story than the vendors are telling. You don't need to be a Fortune 500 company with a $10 million AI budget. You need clean data and a real problem. That's a story most businesses can access. But not if they're too busy chasing the vendor hype to do the prep work first.

We've written before about the white-collar AI panic that hit our own office, and the honest truth is the panic is less useful than the preparation. The people on our team who are fine are the ones who figured out specifically what AI could take off their plate. The ones who aren't fine are still waiting for someone to tell them what the technology does in general.

Stephanie, for what it's worth, immediately wanted to license every agentic AI platform on the market simultaneously. She said the cost was "basically nothing" for something this transformational. She was looking at the enterprise tier. I sent her the Gartner failure rate data. She said that was "probably for the companies doing it wrong." I respect the confidence. I disagree with the logic.

The Real Timeline You Should Plan Around

Gartner predicts that at least 15% of day-to-day work decisions will be made autonomously through agentic AI by 2028, up from zero percent in 2024. Additionally, 33% of enterprise software applications will include agentic AI by 2028, up from less than 1% in 2024.

Those numbers tell you two things. One: this is moving fast and the window to prepare is shorter than you think. Two: 2028 is not 2025. You have time to do this right. The companies that are going to have a real competitive advantage in 2028 are not necessarily the ones who shipped the first clunky agent deployment this year - they're the ones who spent 2026 getting their data clean and their processes documented so that when the tools mature, they can actually use them.

"The next phase will be less forgiving," one industry exec noted. "Buyers are now asking whether agentic systems reduce cycle time, improve resilience, or lower operational cost in production environments. If the answer is no, the terminology no longer matters."

That's the right question. Not "are we using agentic AI?" but "did it do a measurable thing we can point to?" The hype cycle rewards the first question. Survival rewards the second.

If you're looking at sales engagement platforms or evaluating tools that are suddenly slapping "agentic" on their feature lists, ask what it can do without a human in the loop. Ask how it handles errors. Ask what happens when it's wrong. Those three questions will cut through any demo faster than anything the vendor prepares for you. The hype cycle is real. The technology underneath it is also real. The gap between those two things is where most businesses are going to get hurt. Don't be in that gap.