How I AI – “Vibe analysis”: How Faire’s Data Team Uses AI for Fast, Deep Product Analytics

Host: Claire Vo
Date: November 3, 2025
Guests: Tim & Alexa, Faire Data Team

Overview

In this engaging episode, Claire Vo dives into how Faire’s data team uses AI-powered tools—like Notion AI, ChatGPT, Cursor, and MCPS—to uncover and communicate product insights. The team's approach has transformed everything from identifying “what went wrong” with conversion rates, to running detailed funnel analyses, writing SQL, and even conducting and summarizing user surveys, all while democratizing analytics throughout the org.

Key Discussion Points & Insights

1. The Shift from Context Gathering to Executive Insights

• The Foundation:

  • Product development accelerated by AI, but "are these products any good?" is now a perennial challenge ([03:02-03:40]).
  • Tim explains, “The most important, often most difficult thing is actually just getting the right context in the first place, because that's what separates good analysis from bad.” ([04:55])

• New AI Workflow:

  • AI tools dramatically reduce the time to gather relevant context—searching across all platforms (Slack, Notion, Jira, codebase) enables faster and better analysis ([00:07], [05:50]).

• Demo Walkthrough:

  • Tim demonstrates using Notion AI to surface hypotheses for why conversion dropped, summarizing the friction caused by new features ([05:54-07:30]).
  • “Instead, look, I've got straight away a really interesting list of hypotheses to dig into…” ([07:14])

2. Using AI to Query the Codebase for Deeper Investigation

• Forensics via AI:

  • With deep research modes in ChatGPT and Cursor, even non-engineers can “talk to the codebase” to find when and what changes shipped ([10:07]).
  • Tim's prompt: Forensic investigation of the "EORI collection process at checkout"—outputting a timeline of all changes, linked to impact ([10:22-11:27]).

• Accelerating Troubleshooting:

  • Tim: “Knowing nothing about this feature, I can already start to get really smart on what happened. And I can see... an experiment launched in mid-September, right in the sweet spot of when this drop first happened…” ([16:01])
  • This capacity lets product, data, and design quickly access the technical source of truth, not just product docs ([13:38-15:08]).

• Practical Takeaways:

  • Product managers and designers should get “at least read access to GitHub” as part of onboarding ([14:09]), because code is now an analytical data source.

Notable Quote

• “One of the best things about these AI tools is just the ability of someone who's non technical to access things that they couldn't previously access.” – Tim ([10:38])

3. The Power of Cursor, MCPS, and the Semantic Layer for Analysis

• Funnel Analysis Example:

  • Alexa demonstrates an end-to-end AI workflow: Context gathering → writing/running SQL → dashboarding → exec-ready documentation ([19:19-37:55])
  • Cursor’s workflow eliminates context switching and enables self-QA: "It's not the AI's name on this analysis, it's mine." ([30:39])

• Semantic Layer:

  • Faire’s team built a semantic layer—a JSON “translation” of business terms, canonical queries, etc.—making it far easier for AI agents to generate accurate queries ([25:13-29:55]).
  • This democratizes data: “It's just democratizing data and saving us a lot of time so that we can focus on more deep analysis.” – Alexa ([25:13])

• Iterative Documentation:

  • While AI drafts docs, humans are still needed for the final executive summary. “We still need to do 3 to 4 revs of editing… Humans are still valuable.” – Alexa ([41:23])

Memorable Exchange

• Claire: “Over-comment your code with AI! Engineers hate it, but it’s so helpful for human review and future AI context.” ([32:04])

4. Automation of Routine Experiment Analysis

• AI-Driven Experiment Writeups:
  • Tim walks through how simple it’s become to auto-generate standardized experiment summaries.
    • “Are we going to do this for every complicated experiment? Probably not… But for the simple ones, straight one shot—even the complicated ones, this accelerates you.” ([51:34-52:24])
  • AI generates summaries, takeaways, emojis, and even Slack-ready snippets ([51:15]).

Quote

• “AI as a translation layer between a SaaS interface or a SQL query into natural language in the format… that your boss likes. That's just a time saver in and of it of itself.” – Claire ([51:15])

5. Bonus: Supercharging Survey Design and Analysis with AI

• Survey Workflow:
  • Tim shows how to ideate survey hypotheses, turn them into questionnaires, auto-code them for Qualtrics, and even get a first-pass analysis plan—all in minutes ([53:33-55:46]).
  • Raw survey data is then analyzed by AI to output hypothesis tables, confidence scores, and insights—saving hours of spreadsheet work ([57:06-59:18]).

Memorable Moment

• “I just did this really simple prompt…”
  • Alexa quips about Tim’s “simple” but actually highly structured and verbose prompts ([55:46-56:07]).

6. Reflections on Democratization & Enjoyment

• AI for All:

  • These workflows enable sales, designers, and other business folks to write and execute SQL and analysis, lowering the technical barrier ([43:57]).

• Personal Touch:

  • The team agrees the new tooling makes analytics more fun and creative ([43:05-43:45]).

Quote

• “Lowering the barrier to entry on data analysis is just going to create a whole bunch of really high, high impact folks.” – Claire ([44:18])

Notable Quotes

• “The most important, often most difficult thing is actually just getting the right context in the first place, because that's what separates good analysis from bad.” – Tim ([04:55])
• “I'm not an engineer, I can't write Kotlin or Swift. I used to be a lawyer, for God's sake. Instead, I can run a deep research against our code base…” – Tim ([10:38])
• “You have not used a WYSIWYG analytics tool. You have written straight up good SQL, traceable SQL to do a funnel analysis of that on a daily basis. Very interesting.” – Claire ([37:55])
• “We still need to do 3 to 4 revs of editing... Humans are still valuable. And so this is like a pretty good start.” – Alexa ([41:23])
• “It's not just making the good analyst just incredible, it's also democratizing data...” – Tim ([43:57])

Timestamps for Key Segments

• [03:00] – Claire sets the stage: acceleration of product, the need to know “are products actually good”
• [04:14-08:00] – Tim: Broad context gathering and the importance of AI in analysis
• [10:07-17:03] – Demo: ChatGPT & Cursor for codebase forensics; onboarding implications
• [19:19-37:55] – Alexa: End-to-end funnel analysis workflow, semantic layer, doc generation
• [44:54-52:24] – Tim: Automated experiment result writing with AI agents
• [53:33-59:18] – Tim: Bonus—survey design, coding, and analysis in AI
• [60:57-62:09] – “Lightning Round”: Prompting personalities and troubleshooting trick
• Selected throughout: – Reflections on democratization and the importance of making analytics fun

Resources & Contacts

• Alexa: Alexandra on LinkedIn; Faire’s Strategy and Analytics team hiring ([62:17])
• Tim: LinkedIn; “Come join us. If you love AI, come join us and show us how we can do it more here.” ([62:44])

Final Reflections

• AI isn’t just for building—it’s for closing the loop: analyzing, sharing, and communicating results.
• The democratization of analytics via AI is making teams faster and more impactful.
• It’s not merely about automating SQL or dashboards, but about unlocking more strategic thinking and bigger business impact at every level of the organization.

For more episodes, visit howiaipod.com.