AI Is Not a Magic Bullet: What Enterprise AI Gets Wrong

The Meridian Point Podcast EP174: AI Is Not a Magic Bullet: What Enterprise AI Gets Wrong

Guest: Ashwini Kumar, Product Manager & AI/ML Practitioner

EPISODE SUMMARY

Everyone thinks they understand AI now. Ashwini Kumar has been doing the actual work for six years, long before ChatGPT turned every executive into a self-proclaimed AI strategist. From manufacturing and energy to healthcare consulting and major telecom, he's been in the room when enterprises placed their biggest AI bets and watched those bets play out in real time.

In this conversation, Ashwini unpacks the patterns he keeps seeing: executives who treat AI as a pedestal project for their resume, organizations that automate broken human workflows instead of rethinking them, and the persistent gap between what companies hope AI will do and what it actually delivers. He also walks through his team's hands-on experience building an agentic RAG system for proposal generation that collapsed a two-week drafting cycle into two days with a four-person team. The conversation covers why machine learning was solving real problems long before LLMs arrived, where change management breaks AI adoption, and what a C-suite leader should actually ask before building an AI strategy.

KEY TOPICS COVERED

Machine Learning Before the Hype Ashwini describes ChatGPT as "the front end of AI," something anyone can talk to. But machine learning was already powering Netflix recommendations, optimizing roof tile manufacturing recipes, and running CFD wing design simulations. The work was happening. It just wasn't visible to the public.

The Executive Pedestal Problem AI becomes a pet project for some executives. They push implementation so hard that they stop asking whether it's solving actual problems. The investment gets so large that the project "cannot fail," and when it does underperform, the numbers get fudged to protect the narrative.

Automating Broken Workflows Companies make the mistake of layering AI on top of existing human processes without examining whether those processes make sense. A human might jump between six systems to gather information. An automation doesn't need those same steps. Rethinking the workflow before applying AI is where most organizations skip the most critical step.

Agentic RAG vs. Standard LLM Prompting Ashwini's team built a proposal-generation tool using agentic RAG that produces a first draft in one to two days. In a standard LLM like Claude, you'd still need to prompt section by section. With properly designed agents handling the work, the system generates a complete proposal in one run with minimal follow-up prompting. For consulting companies where proposals typically take two weeks (or a month for government work), the time savings translate directly to more proposals submitted and more potential revenue.

The 20/60/20 Adoption Rule When you implement AI in an organization, roughly 20 percent of people are enthusiastic and jump in immediately. The middle 60 percent are indifferent, using it occasionally. The bottom 20 percent want nothing to do with it. Without engagement from that broader population, implementation goes flat and ROI disappears.

Start with Vision, Not Strategy When a C-suite executive asks for an AI strategy, Ashwini's first move is product thinking: what is your AI vision? What do you see it doing? Some executives are lofty. Some are spot on. Some don't have an answer at all. You can't build a strategy until you know where they actually stand.

The Hope-Reality Gap Ashwini's closing observation: there's still a gap between the reality of AI and the hope of AI. The reality will keep moving forward. But there will always be something on the horizon that it doesn't do yet.

MEMORABLE QUOTES

"ChatGPT is sort of the front end of AI. Something that anybody can talk to."

"It becomes sort of a pet thing for some executives. Like, oh, it becomes sort of a thing that they can put on their pedestal. They push it sometimes so hard that they don't think about what problems it's solving."

"You shouldn't apply an AI process as a blanket way to solve what you're doing as a human workflow."

"A human might be jumping around in different systems to get information. An automation doesn't need to jump around. Well, it does jump around, but much faster."

"When we put the agents in and we did the agents correctly, it made a huge difference. We're able to generate the proposal in one run."

"There's a top twenty percent of people that are just all about it. And then you have the middle which is indifferent. And the bottom that doesn't want anything to do with it."

"The reality of AI and the hope of AI, there's still a gap there."

"Everybody believes AI is a magic bullet, but it's not."

KEY TAKEAWAYS

1. Machine learning was solving real enterprise problems for years before generative AI arrived. ChatGPT made AI visible, not new.
2. Executive ego is a failure mode. When AI becomes a resume item instead of a problem-solving tool, the organization stops asking the right questions.
3. Don't automate the human. Examine the workflow first. Remove unnecessary steps before applying AI to what remains.
4. Agentic RAG outperforms standard prompting for complex knowledge work like proposal generation.
5. Change management determines AI ROI. Without broad engagement, implementation dies regardless of how good the technology is.
6. Ask for the vision before building the strategy. If leadership can't articulate what they want AI to do, no strategy will save them.
7. The gap between AI hope and AI reality is persistent. The frontier moves, but so do expectations.

RESOURCES & LINKS

Connect with Ashwini Kumar:

LinkedIn: linkedin.com/in/shawnakumar

Technologies & Concepts Referenced:

• Agentic RAG (Retrieval-Augmented Generation)
• Semantic Kernel (Microsoft)
• Claude Code
• CFD (Computational Fluid Dynamics) simulation optimization
• Machine learning for chemical manufacturing process control

ABOUT THE GUEST

Ashwini Kumar is a product manager and AI/ML practitioner with six years of experience implementing machine learning and AI solutions across manufacturing, energy, healthcare consulting, and telecommunications. He specializes in identifying where AI can genuinely improve enterprise workflows versus where it's being applied as a superficial fix. His current work focuses on agentic RAG systems for knowledge-intensive business processes like proposal generation.

ABOUT THE HOST

Kumar Dattatreyan is co-founder of Agile Meridian and co-creator of The Disruptor Method. The Meridian Point explores disruption and transformation through conversations with leaders, practitioners, and entrepreneurs who are shaping how organizations think and operate.

Connect: LinkedIn | agilemeridian.com | thedisruptormethod.com

Book a conversation

SUBSCRIBE

The Meridian Point broadcasts live on LinkedIn, YouTube, and Facebook every other Tuesday at 12:30 PM Eastern. Subscribe so you don't miss the next conversation.

#EnterpriseAI #AIStrategy #MachineLearning #AgenticRAG #ChangeManagement