Video summary
Stop building SaaS. Build one of these 8 instead.
Main summary
Key takeaways
Summary
Chris argues that generic SaaS—such as undifferentiated dashboards, CRMs, and habit trackers—is becoming harder to defend as AI makes software cheaper and easier to copy. Software is not disappearing; the advantage is shifting from code and feature lists to expertise, product quality, proprietary data, and distribution. He expects products to increasingly serve both people and AI agents, with customers paying for outcomes rather than access to features.
What’s Changing
- Development is cheaper: AI coding tools make it possible to build complex software quickly, weakening code itself as a competitive moat.
- Agents are software users: Agents can use products through APIs, CLIs, and MCP servers, sometimes without a human-facing interface.
- Pricing is moving toward outcomes: Customers increasingly value a result—such as completed accounting work—rather than logins or feature access.
- New sources of defensibility: Build around domain expertise, distinctive product taste and experience, hard-to-access data, or an existing audience and community.
Eight Software Opportunities
1. Software-Enabled Services
- Use AI systems to deliver a service more efficiently rather than selling software alone.
- Examples include accounting, legal, design, copywriting, and branding firms that automate routine work while retaining human expertise.
- Playbook: Map the service process, identify work suitable for automation, add AI systems, and sell the resulting client outcome.
- Chris estimates that AI could help a service business deliver work 2–5× faster; he also describes small teams potentially serving clients 5–20× faster. These are projections, not reported results.
- Examples mentioned: Crosby and Pilot in accounting.
2. Vertical Agent Harnesses
- Build an AI agent environment tailored to an industry or niche. A “harness” combines the model with relevant rules, tools, knowledge, workflows, and memory.
- Potential sectors include law, accounting, and other specialized fields.
- Examples mentioned: Harvey, Abridge, and EvenUp.
- Playbook: Start with deep niche expertise and package the specialized operating context around a general-purpose model.
- Chris’s example is a planned Dungeons & Dragons agent to help users plan and run tabletop campaigns.
3. Headless AI Software
- Create tools that agents can access through a CLI, API, or MCP server, without requiring a conventional interface.
- Examples: iDropper, a CLI for retrieving a brand’s design system into projects; and Posters, a tool for scheduling social posts through an agent.
- Best fit: A developer or specialist with a valuable data source or narrowly defined capability.
- Monetization: Usage-based pricing is suggested. Chris cautions that model providers may add similar capabilities, so differentiation matters.
4. Agent-Native Systems
- Combine a plugin or agent skill with a dashboard that shows progress, results, or user data.
- Chris’s Product OS is an example: its workflow spans ideation, validation, proposals, design, development, deployment, and customer onboarding.
- Playbook: Turn a proven, repeatable method into an agent workflow, then add a control panel to visualize and manage it.
- Best fit: Coaches, consultants, educators, and other professionals with established processes.
- Possible revenue streams include paid plugin access, memberships, coaching programs, or dashboard subscriptions.
5. Taste-Led Niche Products
- Build a distinctive product for a narrow community rather than trying to serve everyone.
- Playbook: Define a specific ideal customer and create something that a small, trusted audience genuinely wants and will recommend.
- Chris recommends thinking in terms of “1,000 true fans” rather than mass-market scale.
- Example: his concept for a roller-coaster ride-tracking app, inspired by a personal interest in amusement parks.
- Existing creators and communities may have an advantage because they already have an audience and distribution.
6. Ambient Consumer Agents
- Put agents into communication channels people already use—such as group chats, WhatsApp, or iMessage—instead of requiring a new app or interaction point.
- Examples mentioned include Instinct, Muse, ChatGPT, Grokbot, and Meta AI in WhatsApp.
- Product principle: Make the agent’s value understandable through familiar actions such as planning, booking, and remembering, without requiring users to understand AI terminology.
- This is presented as a high-upside but highly competitive area, especially for builders experienced in consumer products.
7. Proprietary Data Products
- Use distinctive, hard-to-access data that general AI models are unlikely to contain.
- Examples: Strava heat maps, Waze traffic data, and Glassdoor salary information.
- Playbook: Identify valuable data held by a company or community, build a product around it, make it queryable by agents, and create ways to keep collecting useful data over time.
- Chris’s core point: Apps and code can be copied more easily than unique data and the processes that generate it.
8. Open-Source Core Products
- Make core functionality open source, then charge for valuable services around it.
- Potential revenue sources include hosting, cloud access, support, paid tiers, sponsorship, and advanced capabilities.
- Examples mentioned: Supabase, PostHog, and cal.com; Postgres is also discussed as an example of open-source software supported by hosted services.
- Playbook: Use the code and community as distribution, but establish a clear path to paid offerings users value beyond the code itself.
Choosing What to Build
- Service experience or existing clients: Automate and improve a service business.
- Deep expertise in a niche: Build a vertical agent harness.
- Specialized capability or data source: Consider headless software with usage-based pricing.
- A proven professional method: Package it as an agent-native system.
- A passion, community, or audience: Create a taste-led niche product.
- Consumer-product experience and appetite for risk: Consider an ambient consumer agent.
- Unique, non-public data: Build a data product.
- Comfort working in public and giving away code: Consider an open-source core with paid infrastructure or support.
Metrics, Targets, and Timelines
- The opportunity is framed around the next 5–10 years.
- AI-enabled service businesses could aim to complete work 2–5× faster, with some small agencies potentially claiming 5–20× faster delivery.
- Chris suggests software systems might help service businesses improve efficiency or profitability by 5–10×; this is an aspirational estimate, not a measured case study.
- The niche-product model uses 1,000 true fans as an illustrative audience benchmark.
- No specific revenue, margin, CAC, LTV, or churn figures are provided.
Further Recommendations
- Treat product quality, brand, usability, and the user journey as strategic advantages—not just feature count.
- Use AI’s lower development cost to test many design and experience variations quickly.
- Build on top of improving models, while focusing differentiation on the expertise, data, workflow, or distribution surrounding them.
- For any open-source or agent product, define the monetization path early; free code or agent access alone does not ensure a viable business.
Presenter and Source
Chris is the host of Build Great Products. He describes himself as a product and design consultant with 15 years of app-development experience and says he has helped hundreds of people build products and find their first paying customers.
Rate this summary
Your feedback will help improve summaries.
Improve this summary
Reprocess with a stronger model when the summary feels incomplete or inaccurate.
Translate summary in another language
Ask questions to this video
Chat for follow-up questions, clarifications, and source-backed answers.