Running Startup Operations with AI

At two people, "operations" sounds grandiose. But even a two-person startup has operational overhead: payroll, contracts, compliance, internal communication, vendor management. This work doesn't need an ops hire in the first two years. It does need a system.

Here's how we run operations at DenchClaw with two technical founders and no dedicated ops person.

What "Operations" Actually Means at 2-4 People#

Let's be specific. Early startup operations includes:

• Financial: tracking expenses, managing payroll, ensuring compliance with basic accounting
• Legal: vendor contracts, IP assignments, NDAs, employment agreements
• HR: onboarding new contractors, managing benefits if any, keeping personnel records
• Internal coordination: tracking decisions made, who owns what, recurring rituals

Each of these has a lightweight AI-assisted version that's appropriate for a team of 2-4.

Financial Operations with AI#

You don't need a CFO. You need a bookkeeper + AI.

What we use: Mercury for business banking (startup-friendly, free), Brex for expenses, Pilot for bookkeeping (YC-negotiated deal, worth it).

Where AI helps: Categorizing expenses, generating financial summaries, answering "how much did we spend on X last quarter?", drafting responses to investor financial questions.

In DenchClaw, we have a simple Expenses object for non-Brex expenses that need manual tracking. The agent can query: "How much have we spent on software subscriptions this year?" and give a real answer from our data.

For fundraising financial models, AI (Claude specifically) is good at building financial model templates and running scenarios. It's not good at deciding what assumptions to use — that's your job.

Legal Operations with AI#

You need a startup lawyer for: incorporation, early financing rounds, IP assignments, and any unusual contracts.

You don't need a lawyer for: standard NDAs, straightforward vendor contracts, basic employment agreements.

For standard documents, AI (specifically Claude with a good legal prompt) can generate acceptable first drafts. The workflow:

"Generate a mutual NDA between DenchClaw Inc. and Vendor Name LLC. Governing law: Delaware. Standard carve-outs for public information, prior knowledge, independently developed information."

Review the output. Send it to your lawyer if anything looks unusual. For truly standard documents between parties who both want the deal to work, this is adequate.

Where AI fails: anything unusual, anything adversarial, anything with significant consequences if it's wrong. Don't use AI to draft your financing documents.

HR Operations at Micro Scale#

At 2-4 people, HR is: onboarding new people correctly (W-9/W-8, IP assignment, equipment), handling benefits if you offer them, and keeping clear records of who has access to what.

What we use: Gusto for US payroll and contractors (handles compliance automatically). Notion for a simple internal wiki of policies.

Where AI helps: Drafting onboarding checklists, generating employment/contractor agreement first drafts for lawyer review, answering common HR questions ("how does equity vesting work?", "what are our PTO policies?").

Track your team in DenchClaw:

• Object: Team Members
• Fields: Name, Role, Start Date, Type (FTE/Contractor), Equity (shares), Equipment Issued

This takes 10 minutes to set up and makes administrative tasks (equipment tracking, offboarding checklists) much faster.

Internal Coordination#

This is where DenchClaw specifically earns its keep for internal ops.

For a two-person startup, "internal coordination" sounds like you're just talking to each other. But as you add contractors and early employees, you need lightweight systems for:

• Tracking ongoing projects and who owns them
• Recording decisions so you don't relitigate them
• Keeping a log of open questions that need answers

We use DenchClaw's documents feature for a "Decision Log" — a markdown document where we record significant decisions, the context, and what we decided. Query it when a similar question comes up: "Did we ever decide on our data retention policy?" gets a real answer.

For projects, we use a simple Tasks object with Owner, Due Date, Status, and Priority fields. The agent surfaces the "what needs attention today" view every morning.

The Time Investment#

Total time on operations for a two-person startup without dedicated ops: about 3-4 hours per week. That includes:

• 30 minutes of financial review (weekly)
• 1 hour of vendor/contract management (as needed, average)
• 30 minutes of internal coordination (daily standup equivalent)
• 1-2 hours of ad hoc legal/HR issues

AI gets this below 3 hours. Without AI, my estimate is 6-8 hours — more than a full day each week spent on overhead.

What You Shouldn't Do Yourself#

Some operations tasks are worth outsourcing even at micro-scale:

• Bookkeeping: Pilot or Bench, even for early-stage. The cost is low; the liability for getting it wrong is high.
• Tax preparation: A CPA who does startups. Not something to DIY.
• Series A+ legal: Not the time for DIY legal templates.
• Benefits administration: Use a PEO (Rippling, Gusto) that handles this automatically once you have employees.

The rule: outsource tasks where mistakes are costly and your marginal competence is low.

Frequently Asked Questions#

When do you need a dedicated ops hire?#

When you have 10+ people and the ops overhead is meaningfully impacting someone's ability to do their primary job. Before that, a part-time ops contractor is sufficient.

What's the most important financial practice for early-stage startups?#

Monthly reconciliation — making sure your books match your bank account. This sounds basic, but the founders who do this consistently avoid nasty surprises when fundraising.

How do you handle compliance without a lawyer?#

Use Clerky for incorporation and standard financing documents (designed specifically for startups and reviewed by lawyers). Use YC's standard safe for pre-seed financing. For anything non-standard, use a lawyer.

What internal documentation is actually worth writing?#

Decisions you'll revisit, policies that affect multiple people, and anything a new team member would need to understand to do their job. Skip documentation for processes that change monthly. Document the stable stuff.

Ready to try DenchClaw? Install in one command: npx denchclaw. Full setup guide →