"Writing SQL is enjoyable. Writing less SQL is profitable."

Imagine building and operating your data team at 1/10th the usual cost, while delivering 10x more value.

Sounds impossible?

Well, let's find out!

Where am I today?

I've been building and optimizing data foundations for high-growth companies across the globe for more than 15 years. Four years ago, I turned my freelance consulting into a productized service.

I build data foundations in three steps:

→ (E)xplore: Understand the people, problems, and priorities. Create a clear blueprint.

→ (T)ailor: Set up custom data infrastructure tailored to key use cases.

→ (L)aunch: Run the data infrastructure efficiently, initially as a one-man-show, before hiring a successor to take over.

Thanks to standardized blueprints, reusable templates, and trusted outsourcing partners, I can build up a solid data foundation in less than 4 weeks and operate it solo for 6-12 months.

But I want more.

The Quest for the 10x Data Team

My goal is bold:

I want to build AI-driven systems that:

→ Reduce operational and buildup costs to 1/10th

→ Increase delivered value by 10x

To achieve this, I'm currently (mainly) exploring two avenues:

1. Dramatically Reducing Custom Coding Efforts

Writing custom code for every data infrastructure is costly and time-consuming. While I currently use reusable frameworks and templates, 60-80% of the work remains manual.

I have started to dig into vibe-coding to outsource large chunks of the code creation to AI assistants.

Here's why I believe it will work:

• I've standardized my data foundation with a re-usable five-layer data warehouse model .
• Metrics and dimensions are defined in a consistent format in my KPI-Dimension Maps during the Explore phase.
• I have repeatable blueprints for everything, including definition of roles & responsibilities , that can be translated into an agentic model.

2. Accelerating Ad-hoc Analysis

Ad-hoc analyses are critical - but often slow. Even though I enjoy writing SQL, manual queries take precious time away from strategic tasks.

To tackle this, I'm building AI agents to minimize or completely replace manual query writing and manual ad-hoc analyses.

My first experiments:

• Traditionally, my data foundations contain 95% structured data. I'm now starting to load contextual data from Slack, Email, Notion, Trello, Confluence etc.
• Connected Wobby to my BigQuery DWH and built my first 3 agentic analysts

I'm not happy with the set-up yet, though.

I believe that without a proper semantic layer in place, it will be impossible to make the agentic analysts work.

That's because natural nanguage input to SQL doesn't work. It has to be natural language to yml to SQL.

The problem with semantic layers:

• Semantic layers are resource-intensive to build and maintain.
• Existing solutions are expensive and cumbersome.

I'm now exploring AI-driven, autonomously created semantic layer approaches such as Connecty.ai to drastically reduce the cost of building and maintaining semantic layers.

What's next

Over the coming weeks, I'll update you on:

• Results of my vibe coding experiments.
• Progress on AI-powered ad-hoc analysis agents.
• Lessons learned from the semantic layer side.

And here's a little teaser: I'm considering to launch a community dedicated to building 10x data teams. Not sure yet how this would look like but I'll keep you posted.

Stay tuned!

Cheers,

Sebastian

P.S.: Disclaimer: I don't have any affiliate deals with the tools I mentioned here. While the Founder of Connecty is a friend of mine, I have no financial incentives and I am completely neutral and unbiased.

P.P.S. Make sure you are subscribed if you don't want to miss my updates on building the 10x data team. Subscribe below, if you aren't 👇