Your models are getting smarter. Your AI is getting faster. And somehow the gap between insight and action feels… exactly the same.
This month we're digging into why that is — what's keeping AI stuck in the demo phase, why your models can tell you what will happen but not why, and what teams are doing to close the distance between knowing and doing.
Below you'll find:
📖 Our Recommended Read: Your model knows what happens. It Doesn't Know why.
🔥 Hot Topic: Your AI models: brilliant, expensive, & lost
🛠️ Emerging Tech Insights:ThoughtSpot, Databricks and the agentic shift
📋 From the field: What our clients inherit before a project even starts
😂 LOL Moment:Agentic AI walked so this meme could run
📖 Your Model Knows What Happens. It Doesn't Know Why.
What if the insights your models are delivering are answering to the wrong questions entirely?
This article tackles one of the quieter gaps in enterprise analytics – the difference between prediction and causation – and why it matters more than most teams realize.
The author, JL Verboomen, explores:
A plain-language breakdown of why correlation keeps driving the wrong decisions, and the causal inference techniques that close the gap
How major tech companies have already embedded casual inference into their decision-making infrastructure, and why it hasn't traveled far beyond them
A practical on ramp for teams ready to move from "what happened" to "what will happen if we do X"
Our experts agree: confident predictions without explainability isn't intelligence, it's an oracle. The "why" is the missing link.
If your team has AI that can explain, summarize, and recommend but still hands the actual work back to a human to execute, you're not alone — but you're also not done.
Now it’s time to give your AI solution more context: metadata, semantic layer, lineage visibility…
As Russell Christopher from dbt put it in a recent panel:
"AI without a semantic layer is like you hire a brilliant and very expensive consultant — and then you don't onboard them. They're technically capable, contextually lost, and you wonder why they're not effective."
The fix is context — metadata, semantic layers, lineage visibility — so your AI can stop guessing and start doing.
Emerging Tech Insights
A few of the tech updates on our radar:
ThoughtSpot Opens the Door to Third-Party AI Orchestration
ThoughtSpot just made it possible for AI agents to query your analytics directly — no custom integration, no rebuilding your stack. Ask a question in plain language, get data back, create a dashboard. All from inside whatever AI environment your team is already working in.
This meansthe wall between your AI tools and your analytics layer just got a lot thinner. Instead of AI telling someone where to find the answer, it can now go get it.
The practical implication: If your org is serious about moving AI beyond experimentation, this is the kind of connectivity that makes that possible without a major lift.
Databricks Just Made It Easier to Build Your Own Agent
If your org is already running on Databricks and wants to build a custom AI agent, the path just got shorter. Custom Agents (now generally available) lets teams build, test, and deploy agents without rebuilding their existing workflows or managing infrastructure. Use the frameworks and tools you already have, deploy when ready, and governance travels with it.
The jump from "we built something in a demo environment" to "this is running in production" is where most agent projects stall. Not only does this close the gap without forcing a re-architecture but it also has enhancements built in like memory.
The practical implication: If you've been waiting for a cleaner on-ramp to agentic AI inside Databricks, this is it, especially if getting AI into production without a major infrastructure lift has been the blocker.
From the Field
📋 What our clients inherit before a project even starts
Simon Collis,, our Data Engineering Practice Lead, recently led our latest IlluminA8 session — Analytics8's internal forum where consultants share best practices and lessons from the field. This session focused on Databricks and featured the A8 Intelligence Hub: our internal home for documentation, standards, and delivery playbooks.
The Intelligence Hub ensures we don't start from scratch on every engagement. It centralizes proven architectures, implementation guidance, governance standards, and real lessons learned — so every project builds on collective experience.
Tech certifications matter, but what matters more is turning real-world experience into repeatable delivery.
Every client benefits from what we've already figured out.
LOL Moment
Agentic AI didn’t fix your data problem. It just gave it more friends…
