🎙️ Is AI Just Really Advanced Prediction? What Enterprise Leaders Need to Know
"It's a brilliant piece of marketing to call this technology AI — because it implies a level of reasoning and intelligence behind the tool that's not there." That bold claim from data-scientist-turned-consultant Sahil Panicar sets the tone for a frank, hype-free conversation recorded live at SAFe Summit 2026 in Amsterdam. Joined by Atlassian's Zach Brown and host Raghurani from Accenture, the trio unpacks what today's AI actually is (predictive, not deductive), why companies are bleeding money on "AI sprawl," and what individuals and enterprises should really be doing to get value from these tools. If you've felt the pressure to "go get AI" without knowing why, this episode is your reset button.
👤 Guests
Saahil Panikar — CIO of Atlas Revolutions, SAFe SPCT & Fellow, former data scientist. Specializes in lean portfolio management, value-stream management, and cyber-physical systems.
Zack Brown — Partner Solution Strategist at Atlassian. Helps partners understand Atlassian's product strategy and go to market with joint solutions.
Host: Renaud Granier — Accenture Business Agility
⏱️ Timestamps
Time
Topic
00:09
Introductions & guest backgrounds
02:21
Defining AI: Why LLMs are predictive, not intelligent
04:09
Practical advice: How individuals should approach AI
06:32
Business impact: Atlassian's Rovo and enterprise AI use cases
07:33
AI sprawl: The danger of investing millions without a clear problem
08:22
"Garbage in, garbage out" — Why data quality matters more than the model
10:28
Tactical quick wins vs. enterprise-grade AI success
13:44
Change management: Why training and adoption trump tooling
15:13
Gated gardens & RAG: Securing your AI's reference data
16:31
LLM licensing risks — What's happening with your data behind the scenes
17:32
Building adaptive, AI-native organizations
17:48
Individual takeaway: Get your hands dirty and experiment
20:34
Test the AI on something you're an expert in — the "newspaper test"
22:47
AI hype vs. AI sprawl — Two problems that have meshed together
25:01
Atlassian's approach: Enterprise-tuned, choice-driven, integration-first
28:39
Responsible AI operating models — the building blocks
29:27
Where to find the guests and continue the conversation
🔑 Key Takeaways
AI is prediction, not reasoning. LLMs don't know when they're wrong — they predict what you want to hear based on billions of parameters.
Don't invent the technology, then find a problem. Understand your workflows and data first; AI is a force multiplier that amplifies what already exists — good and bad.
Beware AI sprawl. Dozens of siloed AI tools create more confusion, not less. Start with specific, high-value use cases.
Data quality is non-negotiable. A "gated garden" (curated, secure data source) combined with RAG is essential for trustworthy AI outputs.
Change management > tooling. Training, communication, and human-in-the-loop governance drive real ROI.
Check your LLM licensing. Know what happens to your data when you use frontier models — many license agreements allow the provider to reuse your inputs.
Start personally. Experiment with AI on something you're an expert in to understand its boundaries — then bring that competence to your organization.
📢 Call to Action
Test an AI on something you're an expert in this week. Ask it questions where you already know the answers, see where it shines and where it breaks — and share what you learn with your team. That single step is where responsible AI adoption starts.
Find more episodes and subscribe to stay ahead of the AI-in-enterprise conversation.