Episode 809 | What I Learned Diving into A.I. for 100 Days (with Craig Hewitt)
Timestamps are as accurate as they can be but may be slightly off. We encourage you to listen to the full context.
Podcast Summary
In this episode, Rob Walling sits down with Craig Hewitt, founder of Castos, to explore Craig's ambitious "100 Days of AI" YouTube series and the insights he gained from testing hundreds of AI tools. (03:28) Craig shares his surprising discovery that ChatGPT may not be the best choice for SaaS founders, revealing superior alternatives like Claude Code and Manus that offer more powerful capabilities for business applications. (08:20) The conversation delves into practical AI implementations, including how Craig's team cut customer support workload in half using AI agents, and explores the realities versus hype of AI automation in business contexts. (18:35) Beyond AI, Craig opens up about his strategic pivot at Castos, discussing his evaluation of three new product ideas as he seeks to break through growth plateaus and build a more scalable business empire.
- Core themes include practical AI tool evaluation, the gap between AI hype and reality, strategic product expansion for mature SaaS companies, and the founder's journey of building beyond the first successful product
Speakers
Rob Walling
Rob Walling is the host of Startups for the Rest of Us and founder of MicroConf, the largest community of self-funded software companies. He's also the founder and General Partner at TinySeed, the first startup accelerator designed specifically for bootstrappers, and has been mentoring SaaS founders for over a decade through his podcasts, events, and investments.
Craig Hewitt
Craig Hewitt is the founder of Castos, a podcast hosting and production platform doing low seven figures ARR after eight years in business. He was part of season one of TinySeed Tales and has appeared on Startups for the Rest of Us multiple times. Craig recently completed an ambitious "100 Days of AI" YouTube series, growing his channel to over 11,000 subscribers while testing and evaluating AI tools for SaaS founders.
Key Takeaways
AI Tools Aren't Taking Jobs Entirely - They're Creating Human-Augmented Workflows
Craig discovered through his 100-day journey that AI consistently delivers about 80% accuracy, requiring human oversight for the remaining 20%. (05:26) This means AI isn't replacing founders entirely but rather amplifying their capabilities. The real value lies in automating repetitive tasks while keeping humans in the loop for critical decision-making and quality control. For SaaS founders, this translates to significant time savings on content creation, customer support, and data analysis, but not complete automation of core business functions.
Claude Code and Manus Outperform ChatGPT for Business Applications
Rather than using ChatGPT, Craig strongly recommends Claude Code for coding tasks and Manus for comprehensive business workflows. (08:20) Manus, which uses Claude underneath with a web interface, can perform complex multi-step tasks like researching competitors, downloading transcripts, and creating frameworks - tasks that would be impossible in ChatGPT's consumer-focused interface. Craig demonstrated this by having Manus analyze 20 YouTube videos from Jay Klaus to create intro frameworks for a client. (13:00) The key insight is that consumer AI tools like ChatGPT aren't optimized for the complex, multi-step business workflows that founders need.
Customer Support AI Agents Provide Immediate ROI
Craig implemented DocsBot, an AI customer support agent, which cut Castos's support workload in half for their 4,000-customer base. (18:44) The tool costs $80 monthly but provides value equivalent to multiple salaries by handling routine inquiries automatically. The key to success was providing an easy escape hatch for customers who want human support and continuously monitoring conversations to improve documentation. This approach ensures customers don't feel trapped in frustrating chatbot loops while still capturing significant efficiency gains.
Build One AI Automation Per Month for Compound Growth
Instead of trying to automate everything at once, Craig advocates for implementing one AI automation or agent per month. (15:21) Whether using tools like N8N, Lindy, or Relevance, the compound effect of twelve automations per year can dramatically transform business operations. The focus should be on identifying repetitive, time-consuming tasks that follow predictable patterns. Examples include content distribution, lead qualification, or data aggregation tasks that currently require manual intervention.
Product Validation Must Include Minimum $100/Month Pricing for SaaS Sustainability
When evaluating his three potential new products, Craig established clear criteria: minimum $100/month entry point, built-in expansion revenue, and alignment with existing customer base. (28:32) His rationale is that selling $20/month products is "super hard" and doesn't provide sufficient unit economics for sustainable growth. This pricing threshold forces founders to create genuinely valuable solutions rather than feature-light tools. The expansion revenue component ensures that successful customers contribute more over time, creating a natural growth mechanism within the existing customer base.
Statistics & Facts
- Craig's AI customer support agent (DocsBot) cut Castos's customer support burden in half for their 4,000-customer base, costing only $80 per month but providing value equivalent to multiple salaries. (18:44)
- Craig grew his YouTube channel from zero to 11,000 subscribers during his 100 Days of AI series, with two videos reaching over 50,000 views each. (24:59)
- Castos has built an email list of almost 40,000 people and serves 4,000 customers after eight years in business, positioning them well for cross-selling new products. (27:25)
