The AI Analyst That Never Sleeps: Burak Karakan of Bruin
with Burak Karakan, Bruin
The AI Analyst That Never Sleeps: Burak Karakan of Bruin
Show Notes
Most founders think they have a data problem. They do not. They have an access problem. The data is already there - sitting in a warehouse, a database, a dozen disconnected tools - and 80 percent of the answers they need are buried inside it. The problem is that getting to those answers requires an analyst, a dashboard, a ticket, and three days of waiting. By the time the answer arrives, the decision has already been made.
Burak Karakan built Bruin to collapse that window. As co-founder and CEO, he has spent two and a half years turning what used to take hours into something that takes 90 seconds - and in the process, has started building something more interesting than a faster BI tool. He is building the data layer that every AI agent in your company will eventually need to function.
This episode is about what happens when your data stops being a report you order and starts being a teammate you can talk to - one that knows your blind spots before you do.
From Data Platform to AI Analyst
Bruin did not start as an AI product. Burak and his co-founder had spent years in the data world - working inside large companies, helping smaller ones build infrastructure - and they kept running into the same problem: data was hard to use and even harder to manage. Their first instinct was to build a better data platform for humans.
Then AI agents became real. And Burak made a straightforward bet: if they had gotten good at making humans productive with data, the same principles would apply to agents. Bruin's AI analyst was born out of that pivot. Connect it to your data warehouse, ask any question in Slack or Teams or the web UI, and get a reliable, traceable answer in under 90 seconds. Every step the agent takes is logged and verifiable. You can see exactly how it got to the answer.
Two and a half years later, Bruin works with companies from small gaming studios to large multinational retailers with thousands of physical stores. None of them have voluntarily stopped using it.
Frameworks from This Episode
These frameworks have been added to the AI for Founders Frameworks Library. Filter by Data or Burak Karakan to find them.
The Centralized Data Layer
One AI data analyst serving the whole company beats fifteen teams building their own agents from different sources.
- •When every team spins up its own AI agent, each pulls from different data sources and generates different answers. Sales doesn't match Marketing. Marketing doesn't match Ops.
- •A centralized analyst gives every person and every agent in the company access to the same reliable source of truth.
- •It also creates visibility: you can see who is asking what, where data governance gaps exist, and where teams are making decisions on bad numbers.
- •Bruin replaces 15 competing internal products with one that onboards in two minutes and returns answers in 90 seconds.
- •The moat is not just speed - it is trust. Traceable, verifiable answers that teams can actually act on.
Build for Agents, Not Just Humans
Agents have fundamentally different interface preferences than humans. Designing for both requires treating them as separate audiences.
- •Humans prefer names. Agents prefer IDs - names create ambiguity when the same pipeline exists across multiple teams.
- •Humans want tables and visuals. Agents want raw, programmatically accessible data they can consume at depth without parsing a display layer.
- •Humans want multi-step flows. Agents want single commands that encapsulate as much as possible - fewer round trips, more efficiency.
- •Bruin tests new CLI tooling by running AI agents against it as if they were human users and benchmarking which interaction model produces faster, more accurate results.
- •Building for agents often produces a product that looks incomprehensible to humans - and that is by design.
The Three Stages of Data Intelligence
Most companies are stuck at Stage 1. The real value - and the real competitive advantage - lives in Stages 2 and 3.
- •Stage 1 - Retrospective: What happened? This is standard reporting. It is slow, painful, and most companies never get past it.
- •Stage 2 - Predictive: What is going to happen? Your data already knows your capacity is about to collapse. It knows where you are bleeding before you feel the cut.
- •Stage 3 - Prescriptive: What should we do about it? Agents that notice an upcoming capacity crunch - vacations + increasing inbound calls - and proactively suggest solutions before the problem becomes a loss.
- •Burak's example: an agent that sees two weeks of upcoming team vacations coinciding with a projected sales call surge and surfaces the conflict before a single deal is lost.
- •The question to ask your data every morning: what do you already know that I have been too afraid to ask?
Agent Access Control as a Safety Foundation
You cannot let agents go wild if they have read-only access to two tables. Granular permissions are the foundation of safe agentic deployment.
- •Every Bruin agent can be scoped to specific data sources, specific Slack channels, specific query types, and specific cost thresholds (e.g., no query that costs more than $2).
- •A marketing agent in a marketing Slack channel only sees marketing data. An executive agent in a private channel sees sensitive financials.
- •You can restrict internet access, external network sources, and query complexity at the agent level.
- •The output of every agent is checked by another agent before it surfaces to the team. Every answer links to a full session log.
- •If an agent can only see two of your marketing tables in Snowflake, there is a hard ceiling on the harm it can cause - regardless of how the model behaves.
Founder Experiment: Ask Your Data What You've Been Afraid to Know
Burak's favorite onboarding question is not a business metric. It is this: "What can I ask you?" The agent surfaces its own suggested questions based on what data it can see. Follow the vague ones. That is where the blind spots live.
- 1Pick your single most important data source - your CRM, your data warehouse, your product database. Connect it to an AI data analyst (Bruin, or the open source Bruin Academy stack).
- 2Do not ask it what you already know. Ask: "What questions should I be asking you based on the data you can see?" Let it lead.
- 3Take the three most uncomfortable-sounding suggestions. Dig into each one. These are likely the areas where your intuition and your numbers have quietly diverged.
- 4Run the same question through two different agents or models. Compare the answers. If they differ, you have found a data quality or governance gap worth fixing before it costs you a decision.
- 5Set up a weekly Slack prompt - automated, every Monday morning - that asks the analyst one forward-looking question: "Where is my business most likely to have a problem this week?" Review the answer before your first meeting.
Stretch goal: Identify one support, sales, or operations workflow where an agent currently has to look up data manually before responding. Connect that agent to Bruin's API or MCP server and measure how the time-to-response changes. Burak's customer went from 2-3 hours per support ticket to 40 seconds. Find your equivalent.
Key Terms
These terms have been added to the AI for Founders Glossary. Search by Burak Karakan to filter them.
Tools from This Episode
Bruin
AI data analyst that connects to your data warehouse and answers business questions in Slack, Teams, or a web UI in under 90 seconds. Every answer is traceable. Access controls are granular. Deployment options include cloud, hybrid, and full on-premise. Core technology is open source.
Bruin Academy
A free module inside Bruin's open source ecosystem that walks you through building your own AI data analyst using Bruin's open source tooling. No payment required. Available on GitHub.