2026 State of AI in Enterprise - Fireside Chat

What You Need to Know about AI: Experts Give You What Works, What Doesn’t, and What to Do Now

AI is everywhere, yet it’s still deeply confusing.  

Headlines bounce between hype, warnings, and innovative breakthroughs. As a result, many business leaders lack a clear picture of where AI is headed or how to use it strategically.  

Does that describe you?  

To bring focus, we asked seasoned AI, data, and technology leaders questions that are top of mind for most executives. Here’s some of what they shared.    

Our panel included:  

• Anand Rao, President of BigTapp.ai, who drives enterprise AI modernization and scalable technology solutions. He previously steered cutting-edge global data strategy and digital transformation at Splunk, Oracle, Sony, and First American Corporation.

• Teoman Buyan, Global Chief Information and Transformation Officer at SMJ. Formerly CIO at Coca-Cola and executive leader at Microsoft, Vestel, and Sony, Buyan leads global transformation efforts and is passionate about AI and shaping the future of digital innovation.

• Brian Rowe, SVP of AI and ML at Marlabs, who leads enterprise AI initiatives and pioneers machine-learning and data-strategy innovation across industries. His background spans Perceivant, ExactTarget, iGoDigital, Liberty Mutual, and Cummins.

Read their key insights below or watch the full video above.

‍

Where Does AI Create Real Value Today?

Rao said AI delivers the greatest value when it’s designed to plan, decide, act, and persist, so that whatever you build today should not become a liability tomorrow. The goal is resilience, not novelty, he emphasized.

High-value use cases are already in operation across healthcare, energy, utilities, manufacturing, and law enforcement.

He gave these examples of AI-native drone technology making a dramatic impact today:

• Identifying the location of breathing humans and animals during natural disasters and fires so resources are sent to the right areas

• Detecting potential failures in deep-sea oil pipelines, using geothermal imagery to survey pipelines, track maintenance, check for leaks, and pinpoint issues

• Monitoring and predicting problems with deep-sea fiber optic cables to protect global financial systems  

These cases show how AI moves systems beyond just observing a problem. AI pinpoints issues so they can be solved, generating returns in dollars earned, lives saved, and damaging consequences prevented.  

‍

Will AI Make Data Quality Irrelevant?  

One of the most dangerous misconceptions is that data quality doesn’t matter with AI because AI models can figure it out.  

“I don’t know who said that, but I think they need some education,” Rao said.  

The panel agreed that poor data quality derails AI initiatives.  

Rao stressed that whatever you are building AI for (speed, efficiency, etc.), AI systems are ONLY as good as the data you provide.

Traditional data quality standards (clean, complete, and accurate data) apply. But they aren’t enough with AI because you’re allowing a system to plan, decide, and take action on your behalf, making the stakes higher if something goes wrong.

You also need data assurance, confidence that you have the right data for the right context at the right time to support a specific use case.  

Who is responsible for the AI decisions to stop a train or stop a plane or handle air traffic control? Humans are ultimately accountable when bad things happen.

If anything, AI amplifies and elevates the need for high-quality, battle-tested data.

You should train AI on both good and bad data so it can recognize what looks like failure. But that is different from allowing a live system to act on low-quality data in real time.  

Acting on flawed input can have significant negative repercussions, especially in high-stakes environments.  

‍

Why Do Most AI Pilots Fail, and What Should You Do to Be Successful?

You’ve likely heard that 95% of generative AI pilots fail to produce an ROI, while the 5% who succeed show significant returns (MIT, July 2025).

Failure isn’t inevitable. It’s predictable.  

“It’s not really the AI technology itself [that determines success]. It’s how it’s operationalized, how enterprise execution is done,” Buyan said.  

Many organizations fall into the “pilot trap” from the very start. Critical mistakes are made before the pilot even begins and during early planning stages.  

Companies that fail typically:

• Jump into AI because of buzz, choosing a tool and then looking for a problem to solve — a backwards approach that Rao calls “a disaster”

• Treat AI as merely an IT project

• Set vague success metrics, don’t set metrics at all, or allow IT, not the business, to define success metrics

Successful organizations:  

• Treat AI as a mandatory business transformation, planning and setting expectations accordingly before the pilot begins

• Define a clear, narrow, high-impact business problem that needs to be solved  

• Set measurable success metrics tied to business outcomes

• Require the initiative to prove it will not disrupt what’s already working in the enterprise

• Require the pilot to build buy-in by demonstrating that it reduces costs or saves time  

• Choose a vendor partner to provide speed without rushing for scalability

Buyan underscores that executives must accept early that AI is not an optional experiment but a necessary transformation. How quickly executives, leaders, and teams embrace that mindset will set the pace for everything that follows.

Read more about what determines AI success versus failure in our whitepaper, The AI Divide: Why Most AI Initiatives Fail and How the Top 5% Succeed.

What’s the Most Common Pitfall in Scaling AI from Pilot to Production?  

Even successful pilots struggle when it’s time to scale.  

Rowe said many organizations aren’t ready to operationalize and scale AI the same way they are for normal applications because AI pilots are inherently more difficult to scale.

That’s because AI introduces layers of complexity not present with traditional applications: GPU availability, data requirements, real-time concurrent use, security, accuracy thresholds, and figuring out how much it will cost to scale.  

There’s also a multitude of infrastructure options that didn’t exist a few years ago. Rowe said providers like Nvidia, Google, Microsoft, and gategroup are rapidly rolling out new capabilities, but many still exist only in preview.  

AI scaling is not just a technical problem. There’s a human factor of getting employees to adopt tools they might fear or resist. This is where vendor partnerships help you navigate both challenges.

‍

Do You Need Vendor Partnerships to Integrate AI into Existing Tech Stacks?

“Vendor partnerships have never been more important,” Buyan said.  

Unless you’re a software company, you need an AI vendor partner because you’re not in the business of developing software. You can’t do it alone.  

There are also limits to what you can develop in-house. From a time-to-market and  product-development perspective, AI requires speed and execution expertise.  

Buyan said AI is no longer something you can put on the side and ignore. You need to get AI embedded into your core systems, like your ERP, CRM, and CMS, if you want to stay relevant.

When AI integrates into your workflows and operations, you’re actually building an enterprise capability that wasn’t there before.  

He suggests going slow and deep and highlighting early wins to build trust. When integration fails, that’s when things fall apart.  

‍

AI Vendor Advice: What to Ask, What to Look for, What to Avoid  

Look for a partner who really understands the reality of AI in the enterprise.  

Rao advises that if a vendor leads by talking about features and functionality, that’s not a good sign. Ask what they’ve done for a similar company in a similar industry or use case.  

They should clearly explain where the client was before using their services and where they are now.  

Also look for an exit strategy. Ask how long the vendor will need to support your organization. Can the vendor show you what you don’t know, help you hire qualified people, build your internal capabilities, and transition ownership to your team over time?

If the vendor needs to be embedded in your enterprise forever for the system to work, that’s a red flag, Rao warns.

Look for a partner who:

• Understands your business problem and speaks the language of your business, not just of technology

• Recently solved the pain point you’re experiencing

• Can prove they’ve done a similar implementation for another client recently

• Proves the impact of their work and successes with before-and-after metrics

• Provides a roadmap to build your internal AI capabilities

Avoid a partner who:

• Leads with discussions about AI features, not business outcomes

• Focuses on the software or tools without a plan of how to operationalize them

• Fails to give you a clear exit and transition plan

• Offers solutions where you’re permanently dependent on the vendor

Buyan adds that you should maintain strategic partnership with vendors. Continue using legacy vendors to manage mission-critical operations.  

Don’t make the mistake of juggling multiple vendors handling different small-scale projects. Your business leaders won’t have the bandwidth to manage multiple partners effectively.

‍

How to Prepare Your Organization for AI Now

If budget wasn’t a constraint, Rowe said the most impactful AI investment would be sending executives and leaders who know the operations of the business offsite to plan how to use AI in their jobs incrementally for the next 90 days.

The goal is for someone in each department to do something with AI. This doesn’t have to mean pilots or major AI delivery initiatives.  

Examples might be:

• Using ChatGPT or another generative AI with a custom prompt to accelerates putting together proposals or RFP responses

• Improving the performance of individual work

• Automating reporting

Rowe said AI is severely underutilized. Give employees dedicated time to learn new capabilities of the AI tools they use and put them into practice.

Make AI part of daily work to expand competence. Revisit how you’ll expand use of AI every 90 days, and treat AI adoption as continuous learning, not a one-time event. Recurring interactive AI workshops help move everyone forward.  

‍

Predictions: What’s the Greatest Impact AI Will Have in the Next 3 to 5 Years?

Buyan says the challenges five years from now will be different from today.  

He sees AI becoming a digital workforce. Autonomous or semi-autonomous AI systems will handle repetitive, data-intense computations, make decisions themselves, and escalate to humans when necessary.  

“Humans are always needed, so the AI hype, ‘Everybody’s going to lose their jobs,’ is definitely not the case,” Buyan said.

Humans are not good at computation analysis and looking at details at a mass, CPU level, so AI can handle those tasks.  

But humans are great at judgment, interpreting the outcome of the AI, and creativity (because we created the AI), and that will be our role.  

It’s important that we cannot blame AI for failures. We are accountable for AI’s semi-autonomous and autonomous decisions.  

Buyan also predicts a shift toward manufacturing and operational autonomy.  

We’ll allow AI to adjust purchase orders while we sleep. If a supplier is short on something, it will impact the manufacturing schedule, clients must be notified, and logistics will need to be adjusted.  

The question is whether humans will allow this bidirectional flow of rescheduling, adjusting, and editing and operations to be done by AI.  

Some human intervention is needed to manage exceptions, but Buyan believes AI will handle the rest automatically.

In the next three to five years, you’ll have a workforce that we did not have before, according to Buyan. We’ll use the AI workforce the way we use Excel and PowerPoint today.  

“AI is going to be your next spreadsheet,” Buyan explained.

Rao believes paperwork-heavy roles, such as in healthcare and insurance, will be increasingly automated to eliminate daily drudgery. For example, claims adjusters facing  unsustainable workloads can’t catch up when they get over a thousand claims cases a day, leading to high turnover. This is the perfect use case for AI.  

“All the things that humans don’t want to do will probably go away,” Rao said.

He envisions more natural interactions with data, literally talking to your data rather than navigating screens. If your data is already organized and intelligent enough, with safeguards in place, AI can execute routine actions overnight for you to confirm when available.

So if a printer, cartridge, or an insurance sort “breaks” at night, the whole chain won’t get backed up. AI will automatically create a support ticket, trigger workflows, and resolve operational bottlenecks, and a person can confirm and accept in the morning.  

Rowe said, “Five years is a lifetime in AI.” While it’s unrealistic to expect robots to replace all warehouse jobs, he foresees we’ll go down the path of greater automation, similar to a lights-out warehouse, where robots will be building a lot more things.  

Rowe predicts a shift in interfaces towards voice chat, so people will be “chatting” at the computer rather than typing, particularly for remote or isolated work settings. It will be a seamless interface, where you literally ask for the data you want, and AI responds. The capability is nearly here, and Rowe hopes it will be mastered in five years.

He foresees a small-studio entertainment economy where a few people produce full-length, full-production films at microstudios at a low cost using AI tools, then sell them to streaming services like Netflix.  

Finally, Rao anticipates agent-to-agent ecommerce, where AI agents have the autonomy to handle tasks and even vendor management instead of humans. Human sales won’t go away, but instead of humans, AI agents might be ranking and rating whether a vendor is good enough to do business with them. This will reshape the dynamics of enterprise sales.  

‍

Takeaways

AI continues to change how we work. Your success will require discipline, planning, and accepting that AI is required to stay relevant.  

The road to creating ROI means you must define real problems; build AI on reliable data; scale deliberately; choose partners wisely; and treat AI as a core capability, not an experiment.  

The future of AI is a digital workforce that doesn’t replace humans but equips them with capabilities that didn’t exist before.  

People worried a long time ago that computers would replace jobs, but they created far more new jobs and opportunities. AI will do the same. We just have to develop it responsibly.  

To hear the entire discussion, watch the full fireside chat video.

‍