AI Enterprise - Databricks & Glean | BG2 Guest Interview
Focus on what makes your company special. LLMs are commodities—you can switch between them overnight. The real value comes from your unique data and business processes. Instead of building generic AI solutions every competitor can replicate, leverage your proprietary data to create AI that truly und
45mKey Takeaway
Focus on what makes your company special. LLMs are commodities—you can switch between them overnight. The real value comes from your unique data and business processes. Instead of building generic AI solutions every competitor can replicate, leverage your proprietary data to create AI that truly understands your company's secret sauce. This is your competitive moat in the AI era.
Episode Overview
Databricks CEO Ali Ghodsi and Glean CEO Arvind Jain discuss the current state of enterprise AI, debunking myths about deployment failures, explaining why LLMs are commodities, and arguing we already have AGI. They share successful AI use cases across finance, healthcare, and retail while addressing the 'AI bubble' question and predicting where value will accrue in the AI stack.
Key Insights
The 95% AI Failure Rate Is Actually Healthy
The MIT report showing 95% of AI projects failing isn't alarming—it's expected when experimenting with new technology. If all your AI projects succeed, you're not experimenting enough. The key is rapidly testing hypotheses and learning what works, similar to how startups operate. The 5% that work are driving real transformation.
LLMs Have Become Commodities
Unlike previous technology platforms where users develop strong preferences (iPhone vs Android, Excel vs Google Sheets), LLMs are interchangeable. Companies switch between models overnight based on performance and price. This commodity status means differentiation must come from your data and how you apply the technology, not from the model itself.
We Already Have AGI
By the standards discussed in AI labs 15 years ago, we've achieved artificial general intelligence. The goalposts keep moving, but current LLMs can reason and perform tasks at or above human level across many domains. The challenge isn't waiting for AGI—it's making existing AI useful inside enterprises and expanding successful deployments from 5% to 50%.
Three Camps Define the AI Landscape
Camp 1 pursues superintelligence through massive compute and scaling laws. Camp 2 (AI researchers) says that approach won't work and true AGI is 20 years away. Camp 3 (builders like Databricks and Glean) believes we have sufficient AI capability now and should focus on creating economic value rather than chasing theoretical superintelligence.
Data and Apps Will Capture Most Value
In the AI stack, the intelligence layer (models) is becoming commoditized. Most value will accrue to the data layer (proprietary company data) and the application layer (products that solve real problems). Companies should focus on securing unique data assets and building governance/security systems rather than chasing the latest model.
Notable Quotes
"I think the LLM is a commodity. People are not saying that, but it is a commodity. Like you can get gas from this gas station. You can get gas from that gas station. Doesn't matter. Just compare price."
"You you hear these 95% of projects fail, but like you know like that's that's that's actually what you want. I"
"I think we have AGI. I think we have artificial general intelligence. We really have it. We absolutely have it."
"It really comes down to your company what data does your company have that's special that your competitors don't have"
"We don't actually need super intelligence for that. That's good idea. If the super intelligence guys nail it, amazing then we've cured cancer. Um if they don't hopefully the second camp comes up with a new thing in the next 20 years that's also awesome. We already have whatever we need."
Action Items
-
1
Experiment Rapidly with Multiple AI Vendors
Don't lock into long-term contracts with single AI providers. The winners haven't been identified yet. Use shorter-term contracts and test multiple solutions quickly. Focus on products that work immediately rather than requiring 6-month implementations.
-
2
Identify Your Unique Data Assets
Audit what data your company has that competitors don't. Your AI strategy starts as your data strategy. Build AI systems that understand your proprietary data, business processes, and company-specific workflows rather than generic solutions anyone can replicate.
-
3
Focus on Real Economic Value, Not Demos
It's easy to create impressive AI demos, but they need to deliver actual ROI. Target use cases where AI can transform specific business processes—like reducing equity research reports from 2 hours to 15 minutes—rather than pursuing flashy but shallow implementations.
-
4
Invest in Governance and Security Infrastructure
As AI agents access company data, build robust governance systems to control what information is shared and with whom. Ensure agents can't accidentally expose sensitive data like HR information or use untrusted models that might leak proprietary information.