● Service · AI agents for sales ops
Agents that qualify and enrich leads, draft replies and flag hot prospects. Built in n8n with Claude or OpenAI, connected to your CRM, and set up so a human approves what matters.
Where agents fit
Good at
→ Reading every inbound message and sorting it: new lead, existing client or spam.
→ Scoring leads against your ICP and enriching the record.
→ Drafting first replies for a rep to check and send.
→ Summarising long threads and call notes into a next step.
→ Flagging deals that have gone quiet.
Keep human
→ Pricing and negotiation.
→ Final replies on high-value deals.
→ Anything a client would be upset to learn a bot decided.
Agents draft and flag. People decide.
Example builds
01
Every new message is checked against the contact's history, classified, scored 1–10 and given a draft reply. Urgent ones alert the owner right away; the rest go into a morning digest. I run this on my own lead inbox on self-hosted n8n.
02
New form fills are enriched and scored before a rep sees them, with the reasoning written to the CRM.
03
Drafts the next follow-up for deals that have gone quiet, for the rep to review and send.
04
Long email threads or call notes turned into a short summary and a next step on the deal.
How I build them
Certified in OpenAI AI Foundations and Anthropic AI Fluency for Builders (October 2026).
Self-hosting keeps your data on your own server and has no per-run fees. n8n Cloud is fine if nobody wants to manage a server.
Claude or OpenAI, picked per task.
Agents write drafts and flags, not final actions, until you trust them. Every decision is logged, so you can see why a lead was scored the way it was.
Prompts, workflows and SOPs live in your accounts.
How a project runs
01
Audit
We map your pipeline together and mark every place deals slip out. Free, 30 minutes.
02
Blueprint
I design the fix: workflows, CRM structure and AI agents, mapped to the tools you already pay for.
03
Build
Built, tested and documented, with SOPs and video walkthroughs your team can follow.
04
Support
I stay on after launch to fix what breaks and tune what works. The Zapier case study ran with 90 days of support.
Only if you want it to. Most teams start with drafts and alerts that a person approves, then automate the safe cases.
With self-hosted n8n, workflows and data stay on your server. Model providers process what you send them under their API terms, so we keep what's sent to the minimum the task needs.
It's my usual choice for agents because of cost at volume and control over models. Simple AI steps can also run in Zapier or Make.
The free 30-minute audit. We'll look for the repetitive sales work worth handing to an agent first.
Free · 30 minutes
A working session on your real pipeline. You’ll leave with the 3 biggest leaks and the fixes worth doing first. No pitch, no deck, no obligation.
Book the audit →