Rogue Startups

In this episode

Craig talks with Aaron, who built DocsBot, an AI customer support agent that's almost four years old, about what AI is really doing to SaaS. Aaron shares DocsBot's numbers: slower growth and fewer new customers, but more revenue per customer and lower churn after narrowing in on the right customers. They also get into what it costs to run an AI product, how to choose and switch models, and the AI tools Aaron is excited about next.

Topics discussed

Growth in the AI era: new customer acquisition is down year over year, but revenue per customer is up 20% from last year

Finding the ideal customer: talking to customers, cancellation surveys, and pointing AI agents at all of the product and Stripe data, after reading Jason Cohen's book Hidden Multipliers

Signal density: why founders with only a handful of customers struggle to draw conclusions

Where customers come from now: organic search, AI search, referrals, conferences like WordCamp, and an agency partner program

Why big companies move slowly: and why that makes larger customers stickier

Churn to DIY: churned customers citing AI or their own custom solution went from 1.6% to 13%, and most weren't a good fit anyway

Pricing around token costs: from bring-your-own-key to included usage to AI credits

Evaluating new models: why letting customers edit their system prompt makes evals hard, and a classifier that behaved differently on GPT-5.4

Choosing models: balancing cost, speed and quality instead of reaching for the newest model

Local and specialised models: embedding models, Cohere for other languages, and Mistral for PDF extraction

Privacy: a tiny model in the chat widget that redacts personal information before it leaves the user's browser

What's next: using a new model to re-rank search results at a fraction of the cost

Key takeaways

The biggest cheat code for finding your ideal customer, in Aaron's words, is to "talk to your customers".

Don't pick your ideal customer by revenue alone: DocsBot's biggest enterprise customer uses the product completely differently from almost everyone else.

Customers who leave to build their own alternative are often not best-fit customers: too price sensitive, a tiny team, or not enough support tickets.

Bring-your-own-key let a fully bootstrapped company launch an AI product without paying for inference for about the first two years.

For an AI support agent, speed matters, so the newest model on its highest reasoning setting isn't always the right choice.

Model upgrades can cause regressions that evals don't necessarily show.

Chapters

(00:00:00) - Intro: Is the SaaS apocalypse here?
(00:00:59) - Growth metrics and MRR trends over 12 months
(00:02:08) - Focusing on ICP to boost revenue per customer
(00:03:11) - Churn rates and pricing restructuring
(00:04:49) - Talking to customers to define the ideal customer profile
(00:07:32) - Signal density and sample size in customer data
(00:09:13) - Customer acquisition channels and what's working now
(00:10:47) - Events, conferences, and agency partnerships
(00:13:05) - Enterprise inertia vs. consumer switching risk
(00:15:19) - Enterprise adoption pace and pilot deployment friction
(00:17:03) - How AI changes customer support buying behavior
(00:19:07) - DIY/AI churn: are customers building their own solutions?
(00:20:45) - Price increases and churn reasons breakdown
(00:21:55) - Token costs, pricing strategy, and bring-your-own-key history
(00:26:06) - Evaluating and switching AI models internally
(00:29:24) - Personal model stack and balancing cost, speed, quality
(00:34:00) - Local models, embeddings, and specialized extraction tools
(00:36:25) - Small/local models and future of specialized AI
(00:38:56) - PII redaction with tiny on-device models
(00:40:16) - What's exciting next: re-ranking, Gemini, and new paradigms
(00:43:19) - Wrap-up and where to find DocSpot

Hiten Shah has built Crazy Egg, KISSmetrics, and a long list of products — and right now he’s shipping a new batch of AI tools while stepping back into Crazy Egg after nearly 20 years. In this conversation we get into what’s actually real in AI right now versus what’s hype.

We start with Noats, Hiten’s local-first meeting recorder, and why “the tech wasn’t there” a year ago. That leads into local models: how much hardware you really need, why open-weight models still have a “flavor” from how they’re trained, and why even the frontier labs don’t fully know what a model will be good at until it’s done.

Then we get tactical. Hiten explains how he evaluates new AI tools within minutes of launch, why he treats his work like an “AI factory,” and the big idea of the episode: in the past, go-to-market was a wrapper around the product — today, the product is a wrapper around the go-to-market. He walks through how that plays out across his own products (paid ads, freemium, and a niche community play), and then turns it on my business, Castos, to show what the “obvious future” of a commoditized category looks like.

Chapters
0:00 Noats and local-first meeting notes
4:10 What it really takes to run local models
6:28 Why model training is still a black box
10:22 Where are the US open-weight models?
11:55 Local vs. frontier: how Hiten actually works
13:18 “X is where everything breaks” — evaluating new AI tools fast
15:46 Do you really need to switch tools every week?
18:16 Big-company AI vs. startups for personal productivity
20:39 How to think about growth for an AI product
24:24 The product is now a wrapper around go-to-market
26:30 Does every product need a free tier?
29:19 Coming back to Crazy Egg after 20 years
32:16 What’s exciting — and what keeps Hiten up at night
34:47 Finding the “obvious future” of your business (Castos breakdown)
38:22 Why “grow my podcast” is the new job to be done
40:50 What Hiten’s excited about in AI right now
49:17 Wrap-up: pick who you serve

Links
Noats: https://noats.ai
Hiten on X: https://x.com/hnshah
Hiten’s weekly Friday calls: https://www.hiten.com/
Castos: https://castos.com

Chapters

(00:00:00) - Notes app and local model parity with Granola
(00:02:08) - Local models, product motion, and go-to-market teaser
(00:04:08) - RAM requirements and practical limits of local compute
(00:05:50) - Why training is a black box and model 'flavor'
(00:09:25) - Chinese-trained models and the search for US open weights
(00:11:14) - Personal workflow for testing new models and Twitter as signal
(00:15:42) - Setting up systems to experiment like a factory
(00:18:03) - Muse vs startups: trust, security, and Meta's advantage
(00:20:32) - Notes go-to-market and distribution strategy
(00:24:22) - Product as a wrapper around go-to-market
(00:28:29) - Flexibility across enterprise vs consumer go-to-market
(00:29:10) - Returning to Crazy Egg and AI-driven efficiency
(00:32:09) - Obvious futures: websites that change themselves
(00:34:49) - Finding your business's obvious future: Castos case study
(00:37:43) - Podcast platforms needing to drive growth, not just hosting
(00:40:52) - Excitement about agent tools: Muse, Grokbot, Instinct comparisons
(00:47:17) - Differentiation, distribution, and closing thoughts

What does it take to turn an audience into a business—and how do you do it without losing the enjoyment that got you started? This week on Rogue Startups, Craig and Colin compare notes on building creator businesses. Colin shares what he’s exploring after selling his business, including his golf content, sponsorship opportunities, and fractional roles. Craig talks about growing his YouTube channel toward 25,000 subscribers and building a second revenue stream through ads and brand partnerships.

We dig into the gap between getting views and building a business: growing an email list, creating relevant lead magnets, nurturing subscribers, and developing your own products. We also explore how to find a clear coaching niche by combining your skills and experience—and why honest feedback from your team can help you focus on the work where you add the most value.

Then we get practical about YouTube growth and podcast repurposing. We discuss strong premises, titles, thumbnails, and why the first minute deserves so much attention. We wrap up with how AI tools are changing content production and software businesses, including the opportunities—and uncomfortable questions—that come with automating work you once paid people to do. Fittingly, this episode includes an intro reflecting on using Codex and DaVinci Resolve to edit the conversation itself.

Talking points:

Building a creator business without turning a hobby into a chore.

Monetizing through sponsorships, brand deals, and fractional roles.

Why smaller audiences can still be valuable to sponsors.

Turning podcast and YouTube audiences into newsletter subscribers.

Using lead magnets, segmentation, and your own products to build more predictable revenue.

Finding a coaching niche through a distinctive combination of skills.

Asking your team where you can contribute most effectively.

Giving each piece of content a clear job: views, subscribers, or revenue.

Creating stronger YouTube hooks, titles, thumbnails, and opening minutes.

Repurposing podcast conversations into standalone videos with custom intros.

Using AI to streamline editing, copywriting, and publishing.

How changing customer needs could reshape SaaS and service businesses.

Subscribe to Rogue Startups wherever you listen, and subscribe on YouTube for future episodes. Follow Craig @thecraighewitt and Colin @colinmcgray on social media. Let us know what you thought of the AI-edited episode—and which practical AI workflows you’d like us to explore next.

Get goodies via email: https://craighewitt.kit.com/a7a6c4b5ef

AI agents are getting a lot more capable, but are they actually making work easier yet? 

Craig and Colin compare their latest experiments with GrokBot, Hermes, OpenClaw, Claude Code, Codex, and other AI tools, including what works, what breaks, and where agents are genuinely reducing cognitive load. 

They dig into open vs. closed systems, shared company agents, local AI, model costs, and the idea of an AI chief of staff that keeps founders focused. 

Plus, Colin shares how he's using AI to overhaul a 15-year-old WordPress site, and they explore what the next generation of AI-powered websites and marketing tools might look like.

Highlights from Craig and Colin’s conversation:

Open vs. closed AI agents for running a business

Why Craig chose Hermes as Castos' shared company agent

The security challenge of giving AI agents real power

Why AI should reduce cognitive load, not create more work

How Colin uses AI for CRM, task management, and daily planning

Claude Code vs. Codex, Cursor, and other AI development tools

Why expensive AI models aren't always the best choice

The case for running AI locally or privately

How cheap AI access encourages founders to experiment

Can an AI chief of staff actually keep you focused?

Using AI to overhaul a 15-year-old content site

How tiny AI-built tools could replace traditional lead magnets

Resources and Links from This Episode:

Produced by Castos Productions: https://castos.com/productions/ 

The Podcast Host: https://www.thepodcasthost.com/ 

Colin on LinkedIn: linkedin.com/in/colinmcgray?originalSubdomain=uk 

Castos Free Tools: castos.com/tools

Email me: podcast@roguestartups.com 

Find me on Twitter: https://twitter.com/TheCraigHewitt 

LinkedIn: https://www.linkedin.com/in/craig-hewitt-78386a66/ 

Email: podcast@roguestartups.com

Chapters

(00:00:00) - Intro and topic setup: bots and agents
(00:01:23) - Open vs closed agent platforms
(00:03:35) - Why choosing Hermes for a shared team agent
(00:07:07) - Permissioning and security challenges with Hermes
(00:09:08) - Vision for AI teammates across a company
(00:10:44) - Personal use: Grokbot vs Claude code workflows
(00:12:53) - CRM, task management, and cognitive load reduction
(00:16:09) - Mobile workflows, Notion, and GitHub syncing
(00:19:06) - App preferences: Claude vs Codex vs Cursor vs Antigravity
(00:22:00) - Right-sizing models and cost discipline
(00:22:19) - Golf video workflow and downgrading models to save cost
(00:25:33) - Local and private AI compute considerations
(00:29:35) - Enterprise API costs, subsidies, and sovereignty concerns
(00:32:35) - Value of experimentation under usage limits
(00:33:56) - Chief of staff model and multi-agent orchestration
(00:37:59) - Chief of staff for managing priorities and stale projects
(00:41:25) - Old-school business plans teaser
(00:42:01) - Renovating lawns4u.com with AI agents
(00:45:00) - WordPress vs modern stacks (Astro, Next.js) discussion
(00:47:33) - Building tools and lead magnets beyond WordPress
(00:50:15) - Wrap up and future topics

Do you need a team to build a successful company these days? On this episode of Rogue Startups, host Craig Hewitt (Castos) and co-host Colin Gray (Alitu and The Podcast Host) tackle one of the biggest questions facing founders today: how do you hire in the age of AI agents? 

Craig and Colin chat about what an "ideal" team looks like, which roles are still worth hiring for, and which ones AI agents can take over (to save more important tasks for humans). They dig into how their own hiring processes have changed and evolved, from async video interviews to paid work samples and screen-recorded tasks. They also explain why a candidate's CV matters far less than watching them think through a problem.

Highlights from Craig and Colin’s conversation:

Why you might still want a team, even if you don't need one

What a minimal modern software team looks like

What founders should never fully hand off

How AI agents are changing customer research

The best first hires for growth and partnerships

Why every hire should be an AI power user

How to spot great candidates beyond the CV

Using paid work samples and screen recordings in hiring

The hidden costs and risks of hiring overseas

Why EORs and Merchant of Record services can simplify global hiring and sales

When to hire a human vs. build an AI agent

The trade-offs of staying solo with AI vs. building a team

Resources and Links from This Episode:

Produced by Castos Productions: https://castos.com/productions/ 

The Podcast Host: https://www.thepodcasthost.com/ 

Colin on LinkedIn: linkedin.com/in/colinmcgray?originalSubdomain=uk 

Castos Free Tools: castos.com/tools

Email me: podcast@roguestartups.com 

Find me on Twitter: https://twitter.com/TheCraigHewitt 

LinkedIn: https://www.linkedin.com/in/craig-hewitt-78386a66/  

Email: podcast@roguestartups.com