Will O.

Claude Cowork expert and Claude Code developer. I turn the work your team still does by hand into Claude skills and AI agents that run inside the software you already pay for: HubSpot, Notion, Slack, Monday, Shopify and your own internal tools. Most teams buy Claude seats and use them for writing emails. I set Cowork up so your non-technical staff run real work by name, every day, and get the same result every time. Tell me what your team repeats every week and which systems it touches. I will tell you honestly what Claude should own, what belongs in plain automation, and what the build looks like. PACKAGES CLAUDE COWORK TEAM SETUP, 7 DAYS: audit what your team repeats, build the skills for it, connect them to your systems, train the team, leave a written runbook. 3, 6 or 12 skills. MCP INTEGRATION, 10 DAYS: one Claude MCP server against your software. Read access first, write actions behind human approval, full audit log, documented for your developers. CLAUDE AUTOMATION AUDIT AND REPAIR, 3 DAYS: your automation breaks and nobody knows why. I find the root cause, write up what I found, and fix it properly. CLAUDE AI AGENT BUILD, 14 DAYS: one agent that does one job well. Qualifies a lead, drafts the reply, updates the record, escalates to a person at the point a person is needed. CLAUDE TRAINING FOR YOUR TEAM, 1 DAY: live session, recorded, worked through your own workflows, with a cheat sheet your team keeps. WHAT I BUILD WITH CLAUDE COWORK: named skills anyone on the team can run without knowing a prompt from a placeholder. The Friday report, the client update, the intake summary. CLAUDE CODE: the tool I build and repair the systems with. You never touch it. It is why a fix takes an afternoon instead of a fortnight. MCP SERVERS: the bridge between Claude and your real tools. Without one, Claude can only discuss your business. With one, it reads and writes directly, inside guardrails you set. CLAUDE AGENTS AND THE CLAUDE API: agents embedded in your own product or running your internal operations. PROOF - Diagnosed and rebuilt an n8n to HubSpot workflow creating duplicate contacts, tracing it to the root cause rather than patching the symptom. - Built an AI lead qualification agent that routes by type, updates the CRM and triggers personalised follow-up. - Audited and rebuilt a Zoho CRM and Creator environment where conflicting workflows had made reporting unreliable, establishing one source of truth per field. - Built a WhatsApp AI automation answering from live business data, capturing structured leads and escalating on price, refund or complaint questions. Top Rated on Upwork, 178 jobs, working with agencies, SaaS teams and founder-led businesses. HOW I WORK I audit before I build, because most of these problems trace to a broken foundation rather than a missing feature. Scope is written down before money moves, so we agree what done looks like. Something works at every milestone, documented so your team can extend it without me. Write actions into live systems sit behind human approval unless you tell me otherwise. QUESTIONS CLIENTS ASK ME WHAT IS CLAUDE COWORK AND HOW IS IT DIFFERENT FROM CLAUDE CODE? Cowork is the one your non-technical team uses daily. They ask in plain English and it does the work, including jobs that run in the background. Claude Code is the developer tool I build with. Most teams need Cowork set up well and never open Claude Code. WHAT IS A CLAUDE SKILL? A job saved by name so anyone can run it. Instead of one person knowing the right prompt, the whole team runs the same thing and gets the same result. WHAT IS AN MCP SERVER? The connection between Claude and your real tools. It is what lets Claude act on your data rather than talk about it. CAN CLAUDE ACTUALLY UPDATE MY SYSTEMS, NOT JUST ANSWER QUESTIONS? Yes, once it is wired through MCP or n8n. That is the difference between a chat window and a system that does work. WILL IT DO SOMETHING WRONG IN MY LIVE SYSTEM? This is the right question. Write actions sit behind human approval by default, every action is logged, and I would rather build something that refuses than something that guesses. DO I NEED AI, OR JUST AUTOMATION? Often just automation. If a process runs the same way every time, plain n8n or Make is cheaper and more reliable, and I will say so rather than adding AI everywhere. WHAT IF SOMETHING BREAKS AT 2AM? Error handling, retries and logging are built in from the start, not bolted on after something fails silently. Tell me what is eating your team's time and what it needs to touch.