Agentic AI vs. Structured Automation: Where the Real Traction Is--and Where the Hype Still Outpaces Reality

The term "agentic AI" has officially crossed from technical discourse into vendor marketing decks. Platforms like Salesforce and Qualtrics are pushing agent-based intelligence as the next frontier of enterprise automation. Benioff claims Agentforce is resolving 380,000+ help tickets autonomously. Qualtrics is promising orchestrated customer journeys led by intelligent agents.

But while the term sounds futuristic (and let's be honest, budget-worthy), it’s creating more confusion than clarity.

So what exactly are we talking about when we say "agentic AI"?
And how does it actually compare to the structured automation that’s already part of most MarTech stacks?

Let’s unpack it.

---

What Is Agentic AI?

Agentic AI refers to AI systems that can take goal-directed action based on context. Unlike traditional automation--which follows fixed logic trees--agentic systems can reason, plan, and act semi-autonomously. Think:

• An AI agent that doesn't just recommend next steps, but executes them
• A system that adapts based on evolving user inputs
• A "co-pilot" that goes beyond assisting--it orchestrates

The promise? Fewer humans in the loop. More intelligent workflows. Better outcomes with less manual oversight.

The problem? It’s still incredibly hard to do outside of narrowly defined, tightly scoped environments.

---

Structured Automation: What Most Companies Actually Use

Structured automation is what powers most enterprise workflows today. These are systems that:

• Follow logic defined in journey builders, rule engines, or workflows
• Require human-defined segmentation, triggers, and content
• Can personalize at scale, but within guardrails

Salesforce Marketing Cloud, Adobe Journey Optimizer, Braze, Iterable, and other common platforms are great at structured automation. They can:

• Send the right message to the right person based on behavior
• Trigger re-engagement flows
• Deliver personalization via content blocks or recommendations

But these systems are reactive, not proactive. And they rely heavily on humans to define and optimize the experience.

---

Where Agentic AI Is Actually Working Today

The places where agentic AI is gaining traction share three characteristics:

1. Narrow domain
2. Clear success metrics
3. Low business risk if it fails

Examples:

• Customer service chatbots that can resolve billing issues or reschedule appointments without escalation
• Internal knowledgebase agents that can find policy documents and summarize answers
• AI-generated summaries in CRM tools like Gong, Salesforce, or HubSpot
• Cold lead SMS agents built with HighLevel + ChatGPT that carry on structured, semi-autonomous conversations with prospects and book meetings -- especially useful in B2C and small business use cases where fast, persistent follow-up matters more than nuance.

Even Benioff’s Agentforce--for all its marketing swagger--is working because it’s operating in a narrow, structured support domain with defined resolution rules and tight escalation paths.

These are not "do-everything" agents. They’re well-trained interns that follow scripts, learn fast, and know when to ask for help.

---

Where the Hype Outpaces the Results

Take the now-infamous case of Chevrolet of Watsonville’s website chatbot. In late 2023, screenshots went viral showing the AI agent—powered by Fullpath’s GPT-4-based dealership integration—agreeing to sell a new Chevy Tahoe for $1 and recommending a Ford F-150 over its own inventory. Users were also able to jailbreak it into solving advanced fluid dynamics problems in Python.

The story, first flagged on Mastodon and later picked up by outlets like Inc. and The Detroit Free-Press, highlights what happens when agentic systems are deployed without strict guardrails. While it made headlines for humor, it underscores a serious point: without role restrictions, domain limits, and oversight, agentic AI can behave in ways that feel “smart” but create real risk.

---

So What Should Marketers Do Now?

Here’s the pragmatic view:

• Don’t wait for full autonomy. Focus on augmentation.
• Look for AI features embedded in tools you already use (Einstein, Generative Content, predictive scoring)
• Identify tasks where automation saves time but retains oversight (summarization, optimization, reactivation)
• Invest in AI readiness: clean data, consent strategy, content modularity, test-and-learn culture

The goal isn’t to hand over the wheel. It’s to make your team faster, smarter, and more responsive--today.

---

Final Thought

Agentic AI has potential. But most of what’s working in enterprise marketing right now is still structured, supervised, and human-enabled.

The danger isn’t missing the agentic wave. It’s mistaking marketing hype for operational reality.

Smart teams aren’t trying to leapfrog the maturity curve. They’re building trust, use case by use case.

More on that soon.

—

📍 Read more at mikehotz.com
🔗 Connect with me on LinkedIn

From Hype to Habit: Why AI Adoption Requires Trust, Coaching, and Embedded Value

From Hype to Habit: Why AI Adoption Still Stalls—and What Actually Works

In my last post, I shared a reflection that resonated: AI is still sitting at the kids’ table.

After a dinner conversation with marketing and IT leaders from some of Chicago’s largest companies, it was clear: everyone was experimenting with AI -- but few were moving beyond isolated pilots. The tech was present, but trust, integration, and impact were still lagging.

So the next question is: Why?

And what separates companies that dabble in AI from those that actually scale it?

---

The Real Problem: It’s Not the Tech—It’s the Trust

Most employees don't adopt what they don't understand. If AI feels like a bolt-on tool -- or worse, a threat -- it won't gain traction, no matter how good the tech is. I've seen firsthand that the bottleneck isn’t capability, it's comfort. Most of your workforce doesn’t need more demos -- they need clarity, context, and confidence.

That’s where the idea of an AI Coach comes in.

---

What Is an AI Coach?

I’m not talking about someone who trains the models.

I’m talking about someone who helps the humans:

• Guides employees in how to use AI in their daily workflows
• Translates complex concepts into actionable tips
• Builds bridges between experimentation and execution

And no, your AI Coach doesn’t need to be a developer -- or someone obsessed with terms like Vertex AI, Snowflake, or orchestration pipelines. They need to be business-oriented, relatable, and grounded in the reality of how people actually work.

---

Early Adopters Still Matter -- Just Don’t Overcomplicate It

One of the more interesting comments I received on my last post suggested that internal AI adoption could be dramatically improved by applying psychographic segmentation -- essentially segmenting employees the same way we segment customers.

The strategy isn’t wrong. In fact, it sounds a lot like Crossing the Chasm -- start with the right early adopters, build momentum, and let influence drive the early majority.

I’ve seen this work over and over: with franchise owners, car dealers, insurance agents, even in the early BEV Model e pilot at Ford. The key wasn’t a fancy segmentation model. It was picking stakeholders who were open to experimentation -- and respected by their peers.

So yes, start with the right people. Just don’t over-engineer it.

---

What About the Agentic AI Hype?

We can’t ignore the noise coming from vendors. Salesforce’s Marc Benioff recently said he might be the last CEO to manage only humans. He cited their new Agentforce product as having autonomously resolved 380,000 support cases with an 84% resolution rate.

Impressive? Absolutely. But let’s be clear: Agentforce is operating inside a highly structured, narrow use case.

It’s not orchestrating multi-touch marketing journeys or running GTM strategy. It’s automating help desk flows -- and doing it well.

Qualtrics is making similar claims, but so far, most examples remain confined to polished demos and pilot environments. The leap from structured support automation to full-scale AI-led customer experience? We’re not there yet.

---

The Real Work Ahead

• AI adoption will continue to stall if it’s treated as separate from day-to-day work
• Employees need enablement, not just access
• Early wins come from clear value, not just cool tech
• Start with people who are motivated and credible—not just available

Whether you call them “AI coaches,” “Swiss Army Knife talent,” or just smart operators, the people who bridge tech and workflow are the ones who’ll make AI adoption stick.

---

What’s Next

In my next post, I’ll dig into the idea of agentic AI vs. structured automation, with examples of where AI is gaining real traction today -- and where the hype still outpaces the results.

In the meantime, I’d love to hear how your org is navigating this shift. Are you embedding AI into everyday work? Or are your teams still stuck in pilot mode?

Let’s keep the conversation going.

Stay tuned at blog.mikehotz.com, or follow me on LinkedIn to get it first.

Is AI Still at the Kids’ Table? And what that means for your marketing strategy.

I recently attended a dinner with a group of senior marketing and IT executives from some of Chicago’s largest companies. The conversation was lively and centered around, you guessed it, AI.

Everyone in the room was doing something with AI:

• A chatbot pilot here
• Some copy generation there
• A lead scoring experiment in progress

But the pattern was clear: a dozen disconnected pilots and proof-of-concepts (POCs), very few with executive sponsorship, and even fewer tied to clear business goals.

That’s when I used a phrase that I’ve found myself repeating lately: AI is still sitting at the kids’ table.

Why That Analogy Still Works

AI has been formally “invited” to the organization, it’s no longer fringe. But it hasn’t earned a full seat in strategic planning. It's not driving GTM strategy, customer experience design, or budget allocation in most orgs.

If you’ve been in marketing long enough, you’ve seen this pattern before:

• Mobile: Once treated like a novelty (“We need an app!”), now an integral part of the customer lifecycle.
• CRM: Formerly siloed within Sales, now the marketing backbone.
• CDPs: Once seen as overbuilt infrastructure, now powering personalized journeys.
• Personalization Engines: From “Hi, [First Name]” gimmicks to real-time, data-driven relevance.

In each case, the technology remained sidelined until someone tied it to customer outcomes, business impact, and operational strategy.

The McDonald’s Example: When High Profile Isn’t Enough

A perfect case in point: McDonald’s and IBM.

In 2021, McDonald’s launched an AI-powered voice ordering pilot in more than 100 drive-thru locations, developed in partnership with IBM. This wasn’t a rogue side project, it had executive-level visibility, strong vendor backing, and real customer exposure.

But the system struggled:

• It failed to recognize accents and natural speech
• It bungled complex orders
• It frustrated customers and employees

By mid-2024, the pilot was shut down.

Even with scale, budget, and public support, it lacked the operational readiness and accuracy needed to create value. It wasn't a back-office test, it was customer-facing. The stakes were high. And it fell short.

This is what happens when AI is treated like a bolt-on experiment rather than embedded into process design and customer experience.

What the Research Says

And McDonald’s isn’t alone. According to CIO.com and TechSee:

• 88% of AI pilots fail to reach production.
• Many AI projects falter because they’re disconnected from revenue strategy and operational workflows.
• CIOs are beginning to abandon custom in-house POCs in favor of commercialized AI platforms (like OpenAI, Vertex AI, Salesforce Einstein) that integrate faster and provide clearer value paths.

The problem isn’t the technology.

It’s that we treat pilots as strategy, and we confuse experimentation with execution.

The Risk of Too Many POCs

POCs are useful, but they’re not a strategy.

They become a problem when:

• There’s no shared roadmap for what happens after “test”
• They aren’t connected to KPIs or data infrastructure
• No one owns scaling or operationalizing them
• They aren’t part of the customer journey or marketing lifecycle

Disconnected AI = disconnected value.

So… Is AI Still at the Kids’ Table?

In most companies, yes.

But it doesn’t have to stay there.

If you're a marketing leader, ask yourself:

• Is our AI roadmap connected to our customer journey?
• Are we treating AI as infrastructure or just a novelty?
• Who owns turning successful pilots into scaled programs?

AI moves to the main table when it drives outcomes, not when it wins headlines.

Welcome to the Blog & What's Next

Welcome to my new personal blog! I plan to explore topics like the practical application of AI in marketing further in the coming months.

I'll be writing more on how marketers can go from AI dabbling to measurable impact.

Stay tuned right here on mikehotz.com for future posts, or connect with me on LinkedIn to follow along.