Model Choice, Confidence, and False Positives

Aydin sits down with Mike Potter, CEO and co-founder of Rewind, to talk about how AI is changing both the risk and opportunity landscape for SaaS companies. They cover how AI agents are now deleting real customer data, why backup is more critical than ever, and how Rewind became an AI-native org with dedicated AI ownership, monthly Lunch & Learns, and real internal workflows.

Mike walks through the exact N8N workflows he uses to:

Auto-triage his Gmail into multiple inboxes using AI

Generate a daily AI brief based on tasks, calendar events, and past email context

Analyze churn, win/loss, and internal product data using Claude and MCP

They close with Mike’s “dream automation”: a full AI-generated business review that looks across financials, CRM data, and benchmarks.

Timestamps:

0:00 — Welcome to the show

0:31 — Mike’s intro & what Rewind backs up across SaaS ecosystems

1:40 — AI agents as a new failure mode and how Rewind “saves you from your AI”

4:05 — Turning Rewind into an AI-native company early on

4:53 — First attempt at AI-built integrations (why it failed then, why it might work now)

7:23 — Developers trading tedious integration maintenance for more interesting AI work

9:45 — Code vs architecture: the Shopify webhooks story and handling 1.1B+ events

14:03 — Hiring an AI Engineer: scope, responsibilities, and why background mattered

15:33 — How Rewind drove AI adoption: Lunch & Learns, “use it in your personal life,” experimentation

20:53 — How AI Lunch & Learns actually run across multiple offices and remote folks

23:10 — Examples: CS tools, Alloy prototypes, AI video voiceovers, end-to-end workflows

25:13 — Churn workflows: combining uninstall reasons from multiple marketplaces into Claude

27:06 — Win/loss and internal analytics using Claude Projects + MCP server into an internal DB

29:14 — Choosing between Claude, ChatGPT, and Gemini depending on the task (and re-testing every few months)

31:23 — Mike’s Gmail system: multiple inboxes + N8N + AI classification

36:07 — Inside the email-classifier prompt and AI-powered spam that beats Gmail filters

41:34 — The “Daily AI Brief”: pulling tasks, meetings, and prior email threads into a single morning email

45:02 — Letting AI write and debug N8N workflows (and how assistants in tools are getting better)

48:58 — Wishlist: automated AI business review across finance, Salesforce, and SaaS benchmarks

51:23 — Closing thoughts: so many useful tools are possible, but GTM is the hard part

Tools & Technologies Mentioned

Rewind – Backup and restore for mission-critical SaaS applications.

Claude – LLM used for analysis, projects, agents, and internal tools.

ChatGPT / OpenAI (GPT-4.1, GPT-4.1 mini) – LLMs used for code, prompts, and workflow JSON.

N8N – Automation platform used to build email and daily-brief workflows.

Gmail – Email client where AI-powered labels drive multiple inboxes.

Google Calendar – Calendar data powering the daily AI agenda.

Google Tasks – Task list feeding into the morning brief email.

MCP (Model Context Protocol) – Connects Claude to Rewind’s internal databases.

Alloy – Tool for building interactive product UI prototypes.

Salesforce – CRM used for pipeline, churn, and win/loss analysis.

Gumloop – Workflow tool with an embedded AI assistant.

Zapier – Automation platform referenced for plain-English workflow creation.

Fellow – AI meeting assistant for summaries, action items, and insights.

Subscribe at⁠ thisnewway.com⁠ to get the step-by-step playbooks, tools, and workflows.