Service — Automation
You bring the bottleneck. I build the fix
AI automations, agents, and Claude integrations — built, deployed, and documented for you. Every build is scoped to your workflows and custom-quoted.
scoped & custom-quoted
Sound familiar?
“I want the outcome. I don't want another do-it-yourself project.”
“We tried a no-code bot. It broke, and nobody here could fix it.”
“I'm not handing an AI vendor our customer data.”
What I can build for you
Six of the most common builds — if your bottleneck isn't here, it's probably still buildable.
Inbox triage & drafting
Inbound email classified, routed, and answered with drafts waiting for your approval.
Meeting notes → CRM
Recordings become structured notes, CRM entries, and follow-up drafts — before you are back at your desk.
Document & invoice data entry
PDFs and attachments read and entered into your systems automatically, with a human check on anything unusual.
Proposal & quote drafting
First-draft proposals assembled from your price book and past work — you review, adjust, send.
Scheduled operations briefings
A morning email that pulls live numbers from your business systems and tells you what needs attention. I have shipped exactly this, in production.
Custom MCP connectors
Secure bridges that let Claude read your CRM, calendar, and files — so it answers from your business, not the internet.
Every build ships with
- Human-in-the-loop approvals on anything customer-facing or irreversible.
- Guardrails: the automation can draft and flag — it cannot spend, delete, or promise without a person.
- Documentation and a runbook, written for your team.
- You own everything — the code, the prompts, the workflows. No lock-in.
- Monthly care plans available: monitoring, fixes, and a report on what ran.
Jargon, translated
The words you'll hear — in plain English.
- AI agent
- Software that can use your tools — look up a record, draft an email, update a spreadsheet — and work through a multi-step task, not just answer a question.
- Agentic workflow
- An automation where the AI checks what happened at each step and picks the next action, inside limits you set.
- MCP (Model Context Protocol)
- A standard plug that lets Claude securely connect to your business tools — instead of you copy-pasting into a chat window.
- API integration
- Getting two pieces of software that weren't built to talk — your booking system and your accounting — to share information automatically.
- Human-in-the-loop
- A person stays in the approval chain. The AI drafts and recommends; a human signs off before anything real happens.
- Guardrails
- Hard limits around what an automation may do on its own — the reason it can draft a refund but never issue one.
Questions, answered
Why is there no price on this page?
Because every build is different. A single inbox automation and a multi-system integration are not the same project — you get a fixed quote after a scoping conversation, before any work starts.
Where does our data go?
Builds use business-tier AI access — Anthropic does not train on your data through it — and nothing leaves your accounts without being agreed in the scope. Sensitive fields can be scrubbed or excluded entirely.
Who fixes it when it breaks?
Every build ships with documentation and a runbook. If you want ongoing coverage, a monthly care plan keeps it monitored, maintained, and improving.
What do we own when it is done?
Everything: code, prompts, workflows, documentation. Built on standard platforms, handed over with a runbook — you can run it without me.
Bring your most annoying workflow
No obligation, no pitch deck.