Besides my full time job, I run a small coaching business. Like many solo entrepreneurs, I wear multiple hats. Recently, I started using Claude to help with business decisions, and it revealed a problem I didn’t expect.
The Problem: AI That Hedges Instead of Decides
I’d been asking Claude for help with business decisions like “Should I adjust my service offerings?” or “How should I handle client retention after they achieve their goals?” But Claude kept giving me the classic consultant answer: “Well, you could do X… or Y… it depends on your goals…”
Frustrating, right? I didn’t need options. I needed decisions based on my specific constraints.
Then during one conversation, Claude pointed out something brilliant: I was re-explaining my business constraints every time. Build a dedicated skill that encodes your business rules, capacity limits, and decision frameworks. Then I can give you real answers instead of hedges.
That made perfect sense. I needed to give Claude my business context once, as structured knowledge.
Enter Claude Skills
Claude Skills are basically “onboarding documents” for AI. Think of it like hiring a consultant and giving them your company handbook, except the consultant is Claude and the handbook is a skill.
I decided to build my first skill focused on product and operations decisions for my business.
What I Wanted
I needed Claude to answer operational questions with specific recommendations based on my actual business constraints. Things like capacity limits, service boundaries, and decision frameworks that I’d been re-explaining in every conversation.
The difference I was looking for:
Before: Vague options that require me to do the analysis
After: Specific recommendations with calculations based on my constraints
The Architecture
I already had data systems in place using MCP servers to connect Claude to my databases. The skill I was building would add the decision layer on top. Teaching Claude my business logic without giving it direct access to implement changes.
The skill doesn’t do anything. It just teaches Claude how to think about my business and which data sources to query for different decisions.
Building the Skill
A Claude skill has a main file (SKILL.md) with core business context and decision frameworks, plus detailed reference guides loaded as needed. This keeps the main context lean while having depth available.
For my operations skill, I created six core operational capabilities:
- Service design and capacity planning
- Client retention strategies
- Capacity and growth management
- Performance review and adjustment protocols
- Workflow optimization
- Seasonal strategy
Each capability has decision triggers, which data sources to consult, and links to detailed reference guides.
The skill explicitly tells Claude when to query real data versus when to give conceptual advice:
**When to Use Data Tools**:
- Capacity decisions → Query current client count
- Performance review → Pull activity data
- Retention strategies → Identify clients finishing goals
- Workflow optimization → Search existing templates
This teaches Claude when to actually look at my data versus when to just apply principles.
Results: From Hedges to Decisions
The transformation was immediate. Instead of philosophical discussions, I now get specific, actionable recommendations.
Before (without skill):
Me: “Should I adjust my service offerings?”
Claude: “You could offer a lower-priced tier with reduced touchpoints, or maintain premium positioning with full service, depending on your target market and capacity constraints… What are your goals for growth vs. revenue per client?”
After (with skill):
Me: “Should I adjust my service offerings?”
Claude: “Yes. Based on your current capacity utilization, add a mid-tier service level. This fits within your operational constraints and lets you serve more clients without compromising quality. Here’s the math on how this impacts your time allocation: [specific calculations based on my encoded constraints]. Start by testing with 1-2 clients at the new tier before fully marketing it.”
The difference? Specific recommendation with reasoning, based on my actual constraints, not a generic analysis that forces me to do the work.
Key Learnings
Skills Are Not Code—They’re Knowledge
I initially thought of skills as “plugins” that do things. Wrong. Skills are teaching materials that help Claude understand your domain. Your data systems do things. The skill teaches Claude when and how to use them.
Be Specific About Constraints
The power of the skill comes from encoding your actual business constraints with precision. Not “you can handle a range of clients” but “X clients maximum, Y hours/week available”. Specificity enables decisions instead of hedges.
Version Control Your Skills
Skills evolve with your business. When I update capacity, change my approach, or add new capabilities, I update the skill. Having the skill in version control means I can track how my business logic evolves over time. It’s like versioning your company handbook.
What’s Next
I’ve validated that the operations skill works well. My roadmap:
- Marketing Skill: Positioning, messaging, acquisition strategies
- Technical Skill: Service delivery methodologies and best practices
- Cross-Skill Integration: Teaching skills to reference each other’s decisions
The vision: Three specialized “consultants” (skills) that work together. Each focuses on one domain with clear boundaries and the ability to reference each other.
Try It Yourself
If you’re running a business and using Claude, consider building a skill for your domain. It’s especially powerful if you:
- Have consistent decision frameworks you apply repeatedly
- Keep re-explaining your business constraints to Claude
- Have data systems Claude needs to query
- Want AI to give you decisions, not philosophy
My operations skill took about 4 hours to build (with Claude’s help, ironically), and now saves me 10-15 minutes every time I need to make an operational decision.
Plus, it’s fascinating to externalize your business logic into a document. It forced me to articulate principles I’d been applying intuitively but never wrote down. The skill became my external brain for operations. Now I can version control it, iterate on it, and watch it evolve with my business.
The Meta-Lesson
The most surprising insight? Building the skill clarified my own thinking about my business. Writing down “here’s how I make retention decisions” forced me to realize I had a framework I’d never articulated. The skill isn’t just for Claude. It’s documentation of how I actually think.
If you build nothing else, build an operations skill for your business. Future you will thank present you.
Resources
- Claude Skills Documentation
- Run with Mort (My coaching business)
Questions? Thoughts? I’m on Twitter and always happy to chat about AI-powered business operations, making AI give you actual decisions instead of hedges, or the intersection of software engineering and coaching.
Morten Lileng leads product engineering and architecture for dentsu.Connect Audiences, dentsu’s enterprise identity and data platform. Outside of work, he’s an avid marathon runner and coach.