SoloClientStack · Proof layer

Building a one-person firm.
With AI. In public.

The central thesis: AI lets one operator perform like a small company. This page tests that in practice — transparently, with real numbers, and without pretending it's simpler than it is.

Active build Last updated: May 2026

Current build metrics

What one operator can actually run.

These are real numbers from running SoloClientStack as a one-person operation. Updated as the build evolves.

2
Websites built with AI
~12 hrs
Hours avoided / week
~$2,600
Software cost avoided / mo vs. agency
18
Manual tasks automated
1
Operator running it
4–5
Agency functions replaced

What these numbers mean: 12 hours of admin per week that would otherwise require a VA or part-time hire is now handled by the AI stack. At $50/hr for a VA, that's ~$2,600/month in avoided labor cost against $340 in software. The efficiency gap is real — but so are the limits below.

What's been tested

Six experiments. Real results.

Each experiment tests a specific claim about AI replacing or augmenting solo operator work.

Experiment 01 · Content
Built this site with AI as the primary writing tool
Every page on SoloClientStack was drafted with Claude Pro. Editing, structure decisions, and positioning were human. Writing first drafts were AI. Publication rate: 2–3x faster than unassisted writing.
↑ 3× output · Human work: editing + judgment
Experiment 02 · Automation
Automated the full client onboarding sequence
Built in Make: contract signed → welcome email fires in 60 seconds → intake form sent → Calendly kickoff link → Notion project created. Setup time: 4 hours. Time saved per client: 2–3 hours. Running since March 2026.
↑ 2–3 hrs saved per client · Zero ongoing effort
Experiment 03 · Proposals
AI proposal drafting from discovery call notes
Claude takes raw discovery call notes and a proposal template and produces a first draft. Typical time: 90 seconds. Human editing time: 20–30 minutes. Previous total time: 3–4 hours. Quality difference: undetectable in client feedback.
↑ 80% time reduction · Clients notice no difference
Experiment 04 · Research
AI-assisted tool research and review writing
Testing whether AI can produce publication-quality tool reviews. Process: test the tool live → capture real experience → use Claude to structure and draft. Requires genuine human testing — pure AI reviews are detectable and wrong.
↑ 2× faster · Requires: real human testing first
Experiment 05 · Lead Capture
Email list with zero paid acquisition
Built a lead magnet (the $97/month OS PDF), connected Kit, and drove entirely via organic content. No ads, no paid promotion. Growth rate: modest but consistent. Automations fire on every signup. Zero ongoing manual effort.
Organic only · Automations handle all nurture
Experiment 06 · Stack cost
Running a content + affiliate business under $400/month
Full stack: Claude Pro ($20), Kit ($29), HubSpot Starter ($20), Calendly Standard ($10), Make Core ($10), Notion Plus ($10), Surfer SEO ($89), hosting ($12), domain ($12), miscellaneous ($68). Total: ~$280–$340/month depending on month.
$280–$340/mo total · No full-time labor cost

The honest part

What AI still cannot replace.

This site is not a case for AI replacing operators. It's a case for AI expanding what one operator can do. The distinction matters.

What AI cannot replace

  • Judgment — the call only you can make
  • Trust — earned over time, not automated in
  • Taste — what makes work good, not just done
  • Relationships — every referral, connection, and conversation
  • Decision authority — accountability for outcomes that matter

Where AI creates real leverage here

  • First drafts of anything — content, emails, proposals
  • Automation sequences — built once, run forever
  • Research synthesis — faster than manual reading
  • Formatting and structure — consistent without effort
  • Volume — more content, more pages, more coverage
  • Speed — ideas to published faster than unassisted

Build stage

Where the build stands now.

✓ Done

Infrastructure layer

Site built, email list connected, lead magnet live, affiliate programs applied and approved, onboarding automation running.

→ In progress

Content layer

Tool reviews, comparisons, and OS pages publishing 2–3x per month. Internal link structure being built. SEO signals accumulating.

○ Next

Playbooks and implementation guides

Detailed implementation playbooks for each OS type. These take real testing time — each one gets built and run before publishing.

○ Later

Operator community

A place for solo operators to share builds, ask questions, and compare stacks. Not launching until content layer is solid.

Follow the build

New experiments publish 2–3 times per month. Subscribe to get them.

  • Real experiment results — not theory
  • Honest reporting on what didn't work
  • Stack updates as tools change
  • New OS builds as they launch
  • Playbooks when they're ready

Enter your email to follow the build

No spam. Build updates only. Unsubscribe any time.