AI won’t run your business. But it’ll do the parts of your business you hate doing.

The 80/20 Rule Holds Everywhere

Across every category in this module, the same pattern shows up: AI handles the first 80% of routine business tasks well — data entry, categorization, initial drafts, pattern matching, scheduling. The last 20% — judgment calls, edge cases, anything with legal or financial consequences — still needs a human.

That’s not a failure. It’s the correct mental model. The ROI is “save 10 hours a week on tedious work,” not “fire your accountant and your lawyer.”

Every recommendation below includes a readiness rating:

• Production-ready — Use this today with normal human oversight
• Useful with caveats — Works for common cases, breaks on edge cases, needs careful review
• Emerging — Promising but not reliable enough for anything consequential

Bookkeeping & Accounting

What AI Handles Well

Transaction categorization is the most mature AI accounting capability. Tools like Booke AI connect to QuickBooks Online and Xero, read your bank transactions, and auto-categorize them based on patterns learned from your historical data. After a few weeks of corrections (you fix its mistakes, it learns), accuracy typically hits 90-95% for routine transactions.

Receipt processing via OCR + AI is solid. Dext (formerly Receipt Bank) photographs or scans receipts, extracts vendor, amount, date, and category, and pushes line items directly into your accounting platform. Vic.ai does similar work for invoice coding, learning from your historical patterns.

Bank reconciliation — Xero’s AI assistant (JAX) and QuickBooks’ Intuit Assist both offer intelligent matching of bank transactions to existing records. They suggest matches with confidence scores. You approve or correct.

Natural language queries — Intuit Assist lets you ask “What were my top 5 expenses last quarter?” in plain English and get answers from your books. Handy for quick lookups without running reports manually.

Where It Falls Down

Judgment calls on categorization. Is that dinner a business meal or personal? Is that home office expense 100% deductible or split? AI categorizes by vendor pattern, not business context. You still make the call.

Tax strategy. AI can categorize transactions and spot potential deductions (Keeper and FlyFin do this for freelancers), but tax planning — timing income, structuring expenses, choosing entity types — requires a human who understands your full financial picture.

Reconciliation edge cases. When a bank transaction doesn’t match anything in your books (or matches multiple things), AI guesses. Sometimes it guesses wrong. The monthly close still needs human eyes.

Payroll compliance. State-specific rules, multi-state withholding, contractor vs. employee classification — too much regulatory nuance for current AI.

Tool Recommendations

The Workflow That Works

1. Bank transactions flow into QBO/Xero automatically
2. Booke AI (or your platform’s native AI) categorizes ~90% correctly
3. You review and correct the remaining ~10% weekly (30-60 min)
4. Dext handles receipt capture — photograph receipts, they appear in your books
5. Monthly: review the AI’s categorization accuracy, correct systematic errors, close the books

Time saved: Roughly 8-12 hours per month for a typical solo operation, compared to manual categorization and data entry.

Legal Review

What AI Actually Does Well Here

Contract review for common agreements is the strongest use case. Tools like Spellbook and goHeather read standard contracts (NDAs, service agreements, leases, employment contracts) and flag: missing clauses, unusual terms, deviations from market standard, one-sided provisions, ambiguous language.

First-pass review is the right mental model. The AI reads the contract faster than you can, highlights the parts that need attention, and translates legal jargon into English. This gets you to “I know what questions to ask my lawyer” much faster.

Template generation — AI drafts first versions of common contracts from your specifications. “Draft an NDA for a consulting engagement with a 2-year non-disclosure period and mutual obligations.” The output is a reasonable starting point that a lawyer can refine.

Compliance checking — Some tools check documents against regulatory frameworks (GDPR, HIPAA, SOC2) and flag potential gaps. Useful as a screening tool. Not a compliance certification.

Where It Can’t Help

Anything with real consequences. AI legal tools are screening tools, not legal counsel. They miss nuance, don’t understand your specific business context, and can’t account for jurisdiction-specific rules that might flip the meaning of a clause.

Negotiation strategy. Knowing a clause is “unusual” doesn’t tell you whether to push back or let it go. That requires understanding what matters in your specific deal and relationship.

Litigation risk assessment. AI flags problematic language but can’t gauge the practical likelihood of a dispute or its likely outcome.

Tool Recommendations

The Workflow That Works

1. Receive a contract
2. Run it through AI review (Spellbook, or paste into Claude with “Review this contract and flag unusual, one-sided, or missing terms”)
3. Read the AI’s flagged items — these become your shortlist of things to understand
4. For anything standard with low stakes: you’re probably fine with AI review + your own judgment
5. For anything material (equity deals, large contracts, IP assignments): take the AI’s flagged items to a real lawyer. You’ve just cut their billable hours because you’re coming in informed.

Do not: Sign anything important based solely on AI review. Full stop.

Website Launch & Management

What Works Today

Site generation from prompts is genuinely useful for getting started. Shopify’s AI, Squarespace’s AI builder, WordPress.com’s AI builder, and Elementor all generate full site layouts with copy from natural language descriptions. “Build a consulting website with a services page, about page, testimonials section, and contact form” produces a working site in minutes.

SEO mechanics — AI generates meta descriptions, suggests keywords, optimizes headings, creates alt text for images. It won’t replace SEO strategy, but it handles the grunt work.

Content drafting — Product descriptions, about pages, FAQ sections, blog post drafts. AI writes reasonable first drafts that you edit for voice and accuracy.

What Doesn’t Work

Design taste. AI-generated layouts are functional but generic. If your brand depends on distinctive visual identity, you need a designer (or significant manual refinement).

Complex custom functionality. Booking systems with custom logic, multi-step checkout flows, integrations with niche business tools — still real development work.

Content strategy. AI writes content. Deciding what to create, for whom, in what order — that’s strategy, and it’s still yours.

The Realistic Path for Solo Operators

Standard brochure site: Squarespace AI or WordPress.com AI builder gets you 80% there in an afternoon. Manual polish for brand voice and visual refinement.

E-commerce: Shopify is the clear winner for AI-integrated e-commerce. Product descriptions, inventory management, customer segmentation all have AI assistance baked in.

Custom tool / web app: That’s Module 02 territory — use vibe coding tools to build it.

E-Commerce Operations

What AI Handles

Product descriptions at scale. Shopify’s AI generates product descriptions from basic attributes (name, category, features). Quality is “good enough for most products” — edit your hero products and bestsellers by hand.

Inventory forecasting. AI analyzes sales patterns and predicts stock needs. Useful for seasonal products and dodging stockouts. Most mature in Shopify’s native tools and dedicated tools like Inventory Planner.

Customer service automation. Chatbots handle FAQs, order status queries, and basic return processing. Intercom, Tidio, and Zendesk all have AI agents. For a solo operation, these deflect maybe 60% of inbound queries without human intervention.

Pricing optimization. AI tools analyze competitor pricing, demand patterns, and margin targets to suggest pricing. Treat suggestions as data points, not directives.

Email marketing. Klaviyo, Mailchimp, and others use AI for subject line optimization, send time optimization, and basic personalization. The AI picks when and what subject line — you still write the email (or have AI draft it for your review).

Should I Automate This? A Decision Framework

Before throwing AI at a business task, run it through this filter:

1. Frequency: Do I do this daily/weekly? Good candidate.
2. Stakes: What’s the worst case if the AI gets it wrong? Low stakes = automate freely. High stakes = human review required.
3. Pattern: Is this task mostly pattern-matching (categorization, formatting, routing)? AI excels here.
4. Judgment: Does this task require understanding context that isn’t in the data? AI struggles here.
5. Volume: Is there enough volume that automation saves real time? Automating a task you do once a month probably isn’t worth the setup.

High frequency + Low stakes + Pattern-based = Automate now
High frequency + High stakes + Pattern-based = Automate with review
Low frequency + Any stakes = Probably not worth automating
Any frequency + Judgment-heavy = Keep human, use AI as assistant

Content & Social Media

What Works

Drafting. AI generates reasonable first drafts of blog posts, social media posts, email newsletters, and marketing copy. Key phrase: “first draft.” You edit for voice, accuracy, and brand.

Repurposing. “Turn this blog post into 5 social media posts” or “Summarize this report into an email newsletter” — AI is excellent at reformatting content across channels. This is genuinely one of its best tricks.

Scheduling and optimization. Buffer, Hootsuite, and others use AI to suggest optimal posting times based on your audience engagement patterns.

Image generation. For social media graphics and blog headers, AI image generation (Midjourney, DALL-E) produces usable assets. Not for brand-critical design work, but fine for the Tuesday LinkedIn post.

What Doesn’t

Fully automated posting is risky for brand voice. Even good AI output has a recognizable “AI voice” that attentive audiences pick up on. AI drafts, you review and edit, then schedule.

Strategy and audience development remain human work. Knowing what to say, to whom, and why requires understanding your business and audience in ways AI can’t replicate.

The DIY Content Pipeline

The most effective approach for a solo operator isn’t a $200/month content platform — it’s a simple pipeline:

1. Ideation: Keep a running list. When you have a thought, capture it.
2. Drafting: Give the idea + context to Claude. “Write a first draft of a blog post about [topic]. Target audience: [who]. Tone: [how]. Key points: [what].”
3. Editing: You edit for voice, accuracy, and anything AI got wrong.
4. Repurposing: “Turn this blog post into: 3 LinkedIn posts, 5 tweets, 1 email newsletter intro.”
5. Scheduling: Post via Buffer/Hootsuite or manually.

This costs $0 beyond your existing Claude subscription and produces content that sounds like you, not like a robot.

Further Reading

Previous: Module 05 — OpenClaw Agent Runtime | Next: Module 07 — Data Engineering Mindset