Gen H helps first-time buyers get on the UK property ladder. Mortgage lending is complex - each customer takes roughly six months to convert, applications have dozens of possible states, and the data models reflect every bit of that complexity.

Hal Sarjant, Gen H's Chief Customer Officer, knows the business inside out, but he couldn't answer his own questions. The pattern was familiar: business users submit requests to the data team, the data team is slammed, weeks pass, and most questions die in Slack.

When Hal tried self-serve in traditional BI tools, he hit a wall. The data models were built for engineers. Hal could use AI to write SQL, but it didn't understand the business context. He'd get answers that were technically correct but completely wrong.

Critical questions went unanswered: Which mortgage applications will actually convert? What early signals predict success? Where should our sales team spend their time and effort?

The first unlock

Gen H moved from Looker to Lightdash and rebuilt their data models in dbt. That gave them better structure and flexibility, but Hal's workflow was still clunky: ask Claude to write a query, paste it into Lightdash's SQL runner, get results. Download the CSV, upload it back to Claude, repeat.

"I was basically a copy-paste machine," Hal explained. "It worked for simple questions, but I never had time to go deeper."

Every insight required five manual steps and that friction killed exploration. 

Then Hal turned to Agentic BI - using AI agents to build analytics, not just answer questions. With Lightdash's CLI and Claude Code, the friction disappeared. Agents read Gen H's dbt models, wrote queries, and created visualizations - all in one environment.

"With Lightdash, the UI disappears and the thinking goes to the front. I'm not wrestling with tools anymore, I'm exploring ideas. That's when insights happen."

How Agentic BI actually works

Here's Hal's workflow now:

Step 1: Define the business goal

Instead of jumping straight to SQL, Hal has Claude Code interview him about what he's trying to achieve and they document the goal. 

Step 2: Claude reads the data models

Claude Code explores Gen H's dbt repo and Lightdash's semantic layer. It comes back with: "Here's what I found. Here's what these models mean. Am I close?"

Nine times out of ten, Claude nails it. One time out of ten, Hal catches critical business context that's not in the data: "This metric doesn't work the way you think, let me explain why."

Claude writes it down in their shared learnings file. 

Step 3: Iterate toward insights

Claude Code writes SQL, runs it in the Lightdash CLI, analyzes results, and refines the approach, constantly referencing both the goal and the accumulated business context.

In a single session, Hal might run 60+ queries.

“The data team structures our data and keeps our pipelines running - that's their expertise. As a business user, my job is gathering insights and turning them into ideas that move the business forward. Now I can do that without waiting. That's what Agentic BI unlocks.”

The impact: A complexity score that changed how they work

Here’s an example of what Hal was able to achieve with Agentic BI. Gen H needed to predict which mortgage applications would convert. These applications take six months to complete, so early prediction means better resource allocation, smarter sales strategy, and more accurate funding forecasts.

Hal and his Head of Underwriting had intuitions about what made a case "complex." But intuitions aren't metrics, and you can't route work based on hunches.

Using Agentic BI, Hal built a predictive complexity score in two days and shipped it to production.

What happened next:

• Underwriting workflow changed: Cases are now assigned based on complexity, not random distribution. High-complexity cases go to senior underwriters. Simple cases move faster.

• Sales strategy shifted: The team knows which early signals predict conversion, so they prioritize generating business from brokers who have access to these customers. 

• Capacity increased: Because resources are allocated efficiently, Gen H can handle more cases without hiring more underwriters.

• Executive decisions accelerated: Leadership now makes funding and hiring decisions with the support of nuanced and  predictive metrics, not gut feel.

"We started shaping executive-level decisions based on metrics that didn't exist two weeks prior. Having this complexity score changed how we work. And I did it without ever pinging the data team."

Ten days from idea to production. Zero tickets filed. Zero SQL written manually.

Once the metric existed in the semantic layer, other people started using it in their own dashboards as well. The insight soon became infrastructure.

Why this matters

Agentic BI fundamentally changes who can generate insights.

Before: Hal had questions → submitted to data team → waited weeks → maybe got answers. Most questions never got answered because the barrier was too high.

After: Hal has questions → explores them himself in real-time → ships production metrics. New insights drive executive decisions.

"I want to understand our business deeply and help others understand it too. I only have 10 hours a week for this. I don't want to spend them wrestling with tools. I want to spend them thinking."

The barrier between curiosity and action collapsed. Every metric Hal builds goes into the semantic layer, where other business users can discover it, build on it, and ask new questions. The insights multiply without the data team scaling linearly.

"I'm finding insights the data team didn't even know we needed. We're making executive decisions on metrics we didn't have two weeks ago."

Ready to unlock Agentic BI for your team?

See how Lightdash can help your business users find insights without bottlenecking your data team. Start your free trial or see Agentic BI in action.