10 questions execs are asking about Agentic AI

In magentIQ’s recent Ask-Me-Anything (AMA) livestream, co-founders Ian Barkin and David Brain tackled head-on the most pressing questions from business leaders trying to make sense of the AI boom.

These weren’t scripted talking points or over-polished sales pitches. Instead, what unfolded was a candid and deeply insightful discussion about the real state of AI adoption, and why a Multi-Agent Workforce – AI plus people – is the only viable path forward for organizations that want to see real ROI, fast.

This recap summarizes the top 10 questions asked during the session, and the subsequent discussion.

Click here to access the AMA recap

AMA Recap

Q1. Why aren’t more companies seeing real ROI from AI?

This question came up repeatedly throughout the session, and the stats are sobering: 92 percent of companies are increasing AI investments, but fewer than 4 percent are seeing tangible returns.

David offered a straightforward diagnosis: “Poor use case selection, bad data, and no operational readiness.”

Ian added, “Companies love the hype. They hear ‘AI can do it all,’ and they want that magic. But it doesn’t work like that. The tech is impressive, but without the right scaffolding – people, process, governance – it falls apart.”

Q2. What is a Multi-Agent Workforce, and why does it matter?

At the heart of magentIQ’s model is the Multi-Agent Workforce. This isn’t just a buzzword. It’s a framework for getting work done using the best of AI and people together.

“AI where it wins. People where it matters. That’s it,” said David.

Ian explained, “You can’t run a business with 100 percent AI. And you shouldn’t. Just like you wouldn’t use a hammer to saw a plank. Use the right tool. And in most cases, the right tool is a blend.”

Q3. How should companies be thinking about AI Agents?

The audience asked about the current definition of an AI Agent. The founders admitted: it depends.

“We love giving new names to things,” Ian quipped. “Right now, ‘agent’ just means a piece of AI that does a task. But don’t get hung up on the term. Get clear on the job you need done. Then figure out which type of agent – human or AI – is right for it.”

David built on that: “The agentic trend is psychologically helpful. It helps us break work down into tasks. That modularity is critical. You want to design systems that can swap in the right agent for each job.”

Q4. Should we build our own AI solutions in-house? Or partner?

One of the most common dilemmas: Build or buy?

“Large enterprises can run a thousand pilots and still be fine,” Ian said. “But most businesses can’t afford that. You can’t wait 18 months to see if something works.”

David agreed: “Building a multi-agent solution in-house requires deep process knowledge, technical talent, and change management capability. Most companies just don’t have all three. That’s why we offer AI as a service. We deploy quickly, we take accountability, and we stick around.”

Q5. What tools are you actually using?

Audience members pushed for specifics.

David didn’t dodge. “We use LLMs, RPA, orchestration layers, API connectors, custom-built apps, low-code platforms. But we don’t lead with the tech. The tool follows the need.”

Ian added, “The tool you choose today probably won’t be the tool you use in six months. So don’t marry the tech. Marry the outcome.”

Q6. How do we navigate AI in regulated industries like finance and healthcare?

This came from several compliance-focused leaders.

“Start small and low-risk,” David advised. “You don’t roll out AI across your compliance workflow overnight. Break work down into small tasks. Use AI where you can explain and audit it.”

Ian expanded: “One of our first ML projects was in anti-money laundering. The model worked, but we couldn’t prove why. That’s a non-starter in finance. So we added a layer of oversight, and made it work.”

Q7. Isn’t RPA outdated? Should we just go straight to LLMs?

The founders were clear on this.

“RPA still works,” said David. “It’s well-behaved. If a task is structured and rule-based, why wouldn’t you use it? What’s new is how we combine it with AI agents for more fluid orchestration.”

Ian put it more bluntly: “RPA isn’t dumb. It’s just disciplined. And discipline is good.”

Q8. What are MCP servers and why should we care?

MCP (multi-component platform) servers came up late in the session, prompting curiosity.

Ian clarified: “They let agents share states, coordinate tasks, and plug into systems in a much more modular, scalable way. They’re a big deal. They’ll reshape how agents interact.”

David added, “It’s the glue we need. Especially if you want to run a real business, not just cool demos.”

Q9. How do you pick the right use case for AI?

David gave a checklist:

“Start with repeatable work. High volume, structured data, measurable impact. Make sure it’s a process you understand end to end. Then Agentify it.”

Ian agreed, but warned, “Don’t pour AI onto chaos. If your process is broken, fix that first. Otherwise you’re just making the mess go faster.”

Q10. How do you avoid AI projects becoming science experiments?

“Outcomes first,” said Ian. “We get paid to deliver results. Not to run pilots.”

David elaborated, “We have a phase called the Living Lab. We stand up a real workflow, measure real outcomes, and prove the business case. Then we scale. It’s not theoretical. It’s operational.”

Final Words:

The conversation ended as it began: with clarity, candor, and a bit of humor.

“This isn’t about the number of agents you deploy,” Ian said. “It’s about the outcomes you drive.”

David closed with a reminder: “AI isn’t here to replace people. It’s here to work with them. That’s what magentIQ does. And that’s the future of work.”

Key questions you should be asking now:

• Where in our business can AI deliver immediate, measurable value?
• Do we understand our workflows well enough to automate them?
• Do we have the talent, data, and process maturity to build in-house?
• Where do we need humans in the loop?
• Are we chasing AI for the headline or for the outcome?