AI in Action
How our portfolio companies are using AI within their organizations
Last week I had the privilege of moderating a discussion with a dozen founders from our Garuda Ventures portfolio about how they're using AI within their organizations. What started as a simple question - "What's your most daily used AI tool?" - turned into a fascinating 60-minute deep dive into the practical realities of building AI-native companies in 2025.
The conversation was refreshingly honest. No buzzword bingo or marketing speak. Just founders sharing what's actually working, what's not, and where they're seeing real leverage.
The TL;DR
What's working: Claude and Cursor (but interestingly, no mention of Windsurf) are handling end-to-end coding tasks including bug fixes and deployments. Custom GPTs are solving specific workflow problems in minutes vs. hours. AI-native CRMs (Day.ai, Attio, Clarify) are replacing Salesforce and Hubspot for pipeline management and intelligence with over half the group having migrated.
What's not: Tools that support sales (like AI-based BDRs) are still, as more than one founder put it, "crap." And training AI for specific tasks often takes longer than just doing the work manually.
Interesting nugget: Deep Research is becoming an essential tool not just for market analysis, but also technical planning. You can use it to analyze the build plan, identify comparables, review documentation of similar software, and explain how others approached the problem before writing a single line of code, allowing you to streamline the design of technical system elements.
Read on for a deeper dive and a full list of all the tools mentioned!
ChatGPT is a daily use product
Unsurprisingly, ChatGPT dominated the early discussion. Nearly everyone uses, and uses it daily, contributing to the $10B in annualized revenue OpenAI is doing. Most, however, have moved beyond basic prompting to more sophisticated implementations.
One founder described using ChatGPT to build executive summaries for complex deals by pulling in context from "every single conversation, email, Slack" thread across a 10-person sales team. What used to take days of manual work now happens in minutes with better accuracy than human-generated summaries.
Another founder created a custom GPT for marketing blog images that reduced their image creation process from hours of stock photo hunting to 30 seconds of AI generation. "It took me less time than finding a good stock image," she noted, a relatable pain point for anyone who's scrolled through endless generic photos (here is the post she later published on how to do it!)
Multiple founders also mentioned using ChatGPT's Deep Research functionality for everything from market analysis to technical documentation. One founder uses it before building any major system module, having it research comparable solutions and analyze their documentation and implementation approach, streamlining the design of the system before writing a single line of code.
Agentic coding is very real
The technical founders on the call were the most bullish about AI's current capabilities, with Claude and Cursor dominating the conversation. One founder described having Claude handle end-to-end bug fixes: "A Sentry bug came in. I told it to create an issue in Linear, look at the Sentry bug, and with read-only production access to our database, find the bad code, write a test, fix it, and deploy it. And it just did all those things."
Another mentioned throwing away "$100Ks of internally trained computer vision models in favor of 4o when it came out" – a decision that would have been unthinkable even 18 months ago but is happening more and more as the models improve.
Several founders also talked about moving beyond traditional RAG (retrieval-augmented generation) approaches to full agent-driven workflows. Instead of just retrieving information, these AI agents can autonomously tackle problems by running multiple queries, refining their approach based on results, and iterating until they solve the issue. The upside is obvious – more dynamic problem-solving that adapts in real-time. The downside? Like with anything agentic and autonomous, it will require more time and effort to implement properly.
The choice of coding tool sparked some debate. Cursor, Claude, and Copilot seem to get used interchangeably, but not everyone is drinking the same Kool-Aid. One founder's team actually team actually moved away from Cursor back to direct Claude chat bot because they found the AI tools were "randomly changing stuff around our code base" – understandably unnerving when you're trying to ship product. Instead, they built their own tool for selective context sharing, giving them more control over exactly what the AI touches and when.
GTM is benefiting from a CRM makeover but needs better agents
Sales / GTM seemed more mixed on the value of AI so far. The biggest value is coming from AI-native CRMs, with multiple founders mentioning moving away from traditional tools like Salesforce and Hubspot toward platforms like Day AI, Attio and Clarify. One founder asked their AI CRM to generate ROI analyses for a customer. What would have previously likely taken a week of painful analysis to do, Day.ai did in a matter of minutes, pulling context from all their sales conversations (from notetakers like Granola) and emails with the customer to come up with the framework and number. When your CRM can instantly offer insights "What are all the deals in progress?" or automatically generate business cases for prospects, you're operating at a different level entirely.
However, there was certainly some envy on the sales/GTM side of eng teams that have agentic AI capabilities that largely work out of the box and get the job done. A number of founders talked about having to cobble together their own solutions and spending a lot more time offering context and training the systems rather than rely on off-the-shelf AI sales tools. Multiple founders noted that AI BDR tools are all "kinda crap so far” and are largely underdelivering.
What’s next
We're planning to make these conversations regular, focusing on specific functional areas. Our next session is in July, centered on sales and GTM to dig deeper into ways to drive more tangible value. If you've got tools, hacks, or implementations for go-to-market that actually work, we'd love to hear from you!
The Toolbox
Here's a consolidated list of tools mentioned during the call that founders are actively using:
GTM & Sales:
Day.ai - CRM
Attio - CRM
Clarify.ai - CRM
Granola - Note taker
Naro - B2B content management
Kixie - Automated calling
Tely.ai - Lead generation through blogs
Development:
Claude Code
Gemini
Marketing, Content & Research:
Deep Research (ChatGPT)
Custom GPTs - Specialized tools for image generation, content creation
NotebookLM - Google's research assistant
Superhuman - Email with AI features
What AI tools are your must haves for your teams? Leave a comment below and share your favorites!
Special thanks to Steph, Vasusen, Will, Michael, Derek, Ben, Larry, Evyatar, Satish, Adam and Brad for sharing their use cases and perspectives.
Thank you for sharing
I use ChatGPT and Cursor frequently. I tried Gemini for coding with PyCharm but unfortunately it always misses the mark.