How Block’s custom AI agent supercharges every team, from sales to data to engineering

How Block’s custom AI agent supercharges every team, from sales to data to engineering

How I AIJul 28, 202546m

Brad Axen (guest), Claire Vo (host), Jackie Brosamer (guest)

Bottoms-up + top-down AI adoption at BlockGoose as an open-source, model-agnostic agentMCP as the tool/integration protocol (“arms and legs”)Local-first execution and environment/debug automationVibe data analysis with Pandas + automated recommendationsFrom messy inputs to product catalog + payment links (Square)Vibe-coding new MCPs (Mailgun) + security/permission gates

In this episode of How I AI, featuring Brad Axen and Claire Vo, How Block’s custom AI agent supercharges every team, from sales to data to engineering explores block’s open-source Goose agent connects tools via MCP to automate workflows Block’s Jackie Brosamer and Brad Axen walk through Goose, their open-source, model-agnostic AI agent that gains “arms and legs” by connecting to tools via MCP servers.

Block’s open-source Goose agent connects tools via MCP to automate workflows

Block’s Jackie Brosamer and Brad Axen walk through Goose, their open-source, model-agnostic AI agent that gains “arms and legs” by connecting to tools via MCP servers.

They demo “vibe data analysis” on a messy CSV using Python/Pandas, then automatically generate a shareable HTML dashboard with charts for non-technical stakeholders.

Next, they connect Goose to a Square MCP to convert the same CSV into a real product catalog, generate a live payment link, and then extend the workflow by vibe-coding a new Mailgun MCP to email that link.

Along the way, they emphasize organizational change (not just tech), self-serve data access for every team, local-first control and security guardrails, and pragmatic tactics for debugging and iteration with agents.

Key Takeaways

AI adoption is primarily an organizational transformation problem.

Jackie argues the winners won’t just adopt AI tools; they’ll change operating norms so teams can reliably use them. ...

Get the full analysis with uListen

Non-engineers can be the strongest drivers of AI value.

Block saw sales pushing hardest early, and Jackie notes non-developers are unusually creative at stitching tools together. ...

Get the full analysis with uListen

Goose’s power comes from tool connectivity, not just chat.

Block’s definition of an “agent” is an LLM plus a collection of tools it can call to complete tasks. ...

Get the full analysis with uListen

Local-first agents can reduce friction and increase control.

Goose runs workflows locally, matching developer realities (files, CLIs, environments) and appealing to users who want end-to-end control. ...

Get the full analysis with uListen

Agents turn messy data into usable outputs and shareable artifacts fast.

Jackie drops an “ugly” CSV into Goose, which finds the file, handles Python environment issues, runs Pandas analysis, and generates insights plus recommendations. ...

Get the full analysis with uListen

MCP helps product teams escape rigid import formats.

Brad highlights that instead of forcing customers into a strict CSV schema, an agent can ingest arbitrary formats (CSV, PDF, images) and translate them into the right API calls via MCP tools—reducing import UX complexity.

Get the full analysis with uListen

A repeatable MCP-building workflow is: prove capability → wrap as tool → plug in → test → iterate.

Brad starts by generating a plain Python email-sender, then adds MCP scaffolding using a README snippet, fixes errors in a tight loop, enables the MCP in Goose, and validates by sending real emails—showing a practical “vibe-code to integration” pattern.

Get the full analysis with uListen

Notable Quotes

You tell it what you need it to do by connecting it to different capabilities, and it can just solve any problem.

Brad Axen

The winners… [are] who not just leans into technology, but even more than that, leans into the organizational transformation… humans don't go exponentially so well.

Jackie Brosamer

One of the really underappreciated things about LLMs is how much they function as data duct tape.

Jackie Brosamer

[MCP is] the arms and legs for the model. This is how the model goes and interacts with the real world.

Brad Axen

Find the thing that you don't like doing, and automate that.

Brad Axen

Questions Answered in This Episode

In your internal deployment, what governance decides which teams get which MCPs (e.g., “we have 40–50 internally”), and how do you prevent tool sprawl?

Block’s Jackie Brosamer and Brad Axen walk through Goose, their open-source, model-agnostic AI agent that gains “arms and legs” by connecting to tools via MCP servers.

Get the full analysis with uListen AI

What specific permissioning model does Goose use when it “denies” reading a .env file—can organizations enforce policy centrally, or is it user-configured?

They demo “vibe data analysis” on a messy CSV using Python/Pandas, then automatically generate a shareable HTML dashboard with charts for non-technical stakeholders.

Get the full analysis with uListen AI

For self-serve data, how do you validate correctness and prevent confident-but-wrong analyses when finance or sales runs Goose without a data expert reviewing?

Next, they connect Goose to a Square MCP to convert the same CSV into a real product catalog, generate a live payment link, and then extend the workflow by vibe-coding a new Mailgun MCP to email that link.

Get the full analysis with uListen AI

Brad mentioned Square’s MCP avoids overwhelming the model by using a “lookup what you can do in catalog” pattern—what are your design heuristics for MCP ergonomics and tool discoverability?

Along the way, they emphasize organizational change (not just tech), self-serve data access for every team, local-first control and security guardrails, and pragmatic tactics for debugging and iteration with agents.

Get the full analysis with uListen AI

What’s the best practice for long-running “walk away and come back” tasks—do you use checkpoints, audit logs, or replayable tool-call traces for debugging and compliance?

Get the full analysis with uListen AI

Transcript Preview

Brad Axen

Gooses are AI agents. We kind of designed it to be really agnostic. You tell it what you need it to do by connecting it to different capabilities, and it can just solve any problem. What I'm gonna do is take that CSV that Jackie was working with, I'm just gonna pop it over, and I'm gonna say, "Can you read through this data? Use it to create items in my Square dashboard."

Claire Vo

So we took a CSV that we didn't even have to look at, that probably had data that was not in this exact format, and you created a product catalog.

Jackie Brosamer

What a lot of my work and what my team is trying to do is not just make it faster for someone like me or on my team who has data expertise to go through this, but to allow our finance team, our sales team, anyone in the company, to be able to dig in and self-serve a lot of this data, rather than having to ask an expert and wait for that to come back.

Claire Vo

You have been the first person that's told me [chuckles] it was the salespeople that were begging for it. We hear so much, you know, engineers are leading in, and they wanna vibe code, and they wanna do all that. But it's nice to hear that the folks close to customers and close to revenue are also seeing the value. [upbeat music] Welcome back to How I AI. I'm Claire Vo, product leader and AI obsessive, here on a mission to help you build better with these new tools. Today, we're speaking with Jackie and Brad at Block, who are gonna show us how their open-source AI agent, Goose, can be used to do everything from vibe data analysis to vibe coding in MCP. Let's get to it. This episode is brought to you by CodeRabbit, the AI code review platform, transforming how engineering teams ship faster with AI without sacrificing code quality. Quality code reviews are critical but time-consuming. CodeRabbit acts as your AI copilot, providing instant code review comments and potential impacts of every pull request, beyond just flagging issues. CodeRabbit provides one-click fix suggestions and lets you define custom code quality rules using AST GREP patterns, catching subtle issues that traditional static analysis tools might miss. CodeRabbit brings AI-powered code reviews directly into VS Code, Cursor, and Windsor. CodeRabbit has so far reviewed more than 10 million PRs, been installed on one million repositories, and has been used by 70,000 open-source projects. Get CodeRabbit free for an entire year at coderabbit.ai, and use the code HOWIAI. Thanks for being here!

Jackie Brosamer

Thanks for having us.

Brad Axen

Yeah, excited to chat.

Claire Vo

I am so excited to have you two on How I AI, because I have been so impressed how Block has, as a full organization, and a large organization at that, really embraced AI, it seems, everywhere and for everything. And, you know, everybody's worried about AI-native startups, but you're one of the companies that I really think is going there very fast as a much larger organization. So how did you lean in as an org so fast and so broad?

Install uListen to search the full transcript and get AI-powered insights

Get Full Transcript

Get more from every podcast

AI summaries, searchable transcripts, and fact-checking. Free forever.

Add to Chrome