Leading Voices: most brands fail at agentic AI. Our solutions experts share how yours can win.

Every major AI wave arrives with a new vocabulary. Generative AI gave way to copilots, copilots gave way to agents, and the cycle keeps turning. In this episode, Jeff points to a pattern worth taking seriously: 'We had a session just a little while ago talking about the big fails that happened at industry events recently, and I think all of them were agentic-AI-created fails.' That’s no coincidence. When a technology becomes a branding requirement rather than a deployment decision, businesses implement it to signal progress rather than solve problems. The failures that follow are not primarily technical failures. They’re strategic ones, and Cortney and Jeff spend the episode explaining exactly why.

The first problem is definitional. The industry has stretched the term ‘agentic AI’ so far that almost any automated process now carries the agentic label. Cortney draws the line directly, 'If we're making an API call, that's not automatically considered agentic AI. I think a lot of vendors are actually stretching that definition.' In response, Jeff recounts advising a client who believed they needed agentic AI to connect two platforms, when the actual solution required basic automation. 'It's interesting how people are not perceiving the exact use case of when agentic AI makes sense or when it doesn't,' he adds. The clearer definition Cortney puts forward? agentic AI describes a system that can autonomously make a decision or take an action on behalf of the customer and the company, across multiple systems, multiple decisions, multiple steps, and conditions that change. Most customer service interactions don’t come close to clearing that bar.

Once the definition is clear, the strong use cases become obvious, and they’re genuinely compelling. Jeff builds the argument through a specific use case: enterprise technical support triage, where an agentic system reads incoming cases, assesses complexity, and routes contacts without requiring a human to pre-write rules for every conceivable scenario. Cortney sharpens the logic: 'agentic AI could make a decision about the complexity of the situation and decide whether to route a customer to tier one or tier three. Maybe the customer bypasses some of those lower-level tiers because the issue is obviously complex and automatically falls into tier three. There's some automatic savings, some automatic efficiency built into a process like that.'

The customer experience benefit runs deeper than routing speed. 'It can deliver the context it used to make that decision to the agent who receives the contact,' Cortney says. 'By the time a customer reaches that tier-three agent, the agent already knows what the problem is. The customer doesn’t have to repeat themselves.' Jeff captures the implementation upside: 'I don't have to establish all the rules. I can let the agentic AI, based on what I give it access to, decide where to put things without creating all these scripts with rules around it. Theoretically, it’s faster to implement and potentially more accurate.'

The pattern that makes these use cases work is consistent: multiple systems, multiple touch points, and multiple decisions in play simultaneously. Customer service orchestration, claims processing, IT ticket resolution, workforce management, and supply chain workflows all reward the kind of adaptive decision-making that agentic systems handle exceptionally well. That consistency is exactly what leads Cortney to a bigger point about sequencing.

The most valuable insight Cortney shares in this episode is about sequencing, and it applies to the majority of organizations currently being pitched on agentic AI. 'We're sort of leapfrogging all of these capabilities that we need to implement and trying to go straight to agentic,' she says. The capabilities she identifies as higher priorities for most businesses right now include AI search, knowledge management, copilots, agent assist, AI quality assurance for compliance, and workforce optimization and analytics. 'I feel like companies will see more gains with those capabilities than they will if they go straight to agentic and have a really hard time implementing it.' Jeff adds the cost reality: ‘You could probably solve some of those problems without agentic at a lower cost right now too.’

This matters because you can’t layer autonomy onto chaos and expect better results. Agentic AI needs reliable processes, accessible knowledge infrastructure, and quality systems to act effectively. Getting those pieces right first isn’t falling behind. Cortney names the diagnostic that separates strong deployments from struggling ones: 'Your underlying process has to be pretty solid for agentic AI to take an action within it. So you need to ask yourself if your processes and systems are really mature enough for that agentic layer.' Organizations that answer that question honestly give themselves a real advantage.

For leaders navigating vendor conversations, Cortney offers a set of questions that shift the dynamic entirely in the organization's favor, and the most powerful one reframes the whole conversation. Rather than asking what agentic AI use cases the organization can find, leaders should ask what business problems exist and whether agentic AI solves them. 'It's sort of just flipping those,' Cortney says, and the flip matters because it puts the business's real needs at the centre rather than the vendor's narrative. 'Otherwise, you implement AI and AI becomes your new business problem.' Starting from the problem rather than the product is where the strongest AI programs begin.

Her second question is just as direct: Does autonomy actually improve the outcome relative to what already works? If a password reset process runs well today, adding an agentic path for the same task creates complexity rather than value. Jeff makes the point precisely: 'If I have a rules-based system in place that's working well, there's really no reason for me to put agentic AI on top of it.'

Finally, their third question is the one that keeps every investment honest: 'Are we buying value? Are we buying outcomes for our customers? Or are we getting sucked in by this buzzword?' Cortney puts it directly, and it’s well worth keeping this top of mind in every vendor conversation.

The most interesting part of the episode is when Cortney runs Jeff through a live exercise: what would he let an AI agent handle without human approval? Some answers land exactly as expected. Surgery gets an immediate, cheerful no. A $50 customer refund gets an easy yes, with guardrails around the dollar amount. Others are more surprising: approving or denying a mortgage lands as a nuanced yes since the process is largely rules-based already - as long as escalation and bias controls are in place.

The full discussion and the reasoning behind every call are well worth watching in the episode itself. Cheeky as some scenarios may be, the exercise surfaces the examination that should govern every agentic deployment: How high are the stakes, how reversible is the action, and where does model bias introduce unacceptable risk? Those three variables, not the capability of the underlying model, determine where autonomous action earns trust now and where it might earn it next. 'We could have this conversation in six months or a year and have completely different answers to that exact same list,' Cortney says. 

And that’s the point. Technology will keep evolving, and it’s an exciting thing to see. Jeff's one fixed position through all of it concludes the episode perfectly, however: 'I'm always going to be cautious about my surgeon using agentic AI to do surgery.'

At a time where noise and hype cloud AI conversations across the board, Cortney and Jeff’s approach to agentic AI is ultimately about clarity: starting with a real problem, understanding how your own processes actually work, and knowing where the technology genuinely fits. That kind of thinking doesn’t slow down AI adoption. It makes the adoption more intentional, more practical, and far more likely to deliver lasting value.

Businesses that build durable AI capability treat each wave as a question to answer rather than a trend to follow. Agentic AI is a truly powerful capability, and for the leaders who get the foundation right, it’s also an achievable one. Enjoy the full episode and all Cortney and Jeff’s expert insights here: Episode 9: Is Agentic AI Being Oversold?.