Work with me
Get your week back. I find the workflow eating the most hours, build the system that runs it, and train your team to own it. Leads answered in minutes. Follow-up that never slips. Paperwork that files itself. Live in 14 days, in accounts you own, and if it is not working, you do not pay.
A lead comes in, everyone is busy, and by the time someone replies they have already called the next place on Google.
Intake gets retyped into the CRM, the spreadsheet, and the email thread. Your best people spend hours being a copy machine.
Quotes, renewals, callbacks. All memory, no system. When that person is out, it simply does not happen.
Nothing moves unless the owner pushes it. Your business runs on your attention instead of a process.
You already pay for tools with the exact feature you need, turned off, because nobody ever had two clear days to wire it up.
The only report is the bad news: the quote that died silent, the client who left, the lead nobody touched.
Every inquiry gets a fast, personal-sounding response, even after hours, even when everyone is on the phone.
Quotes get chased, renewals get worked, callbacks happen on schedule. Nobody has to remember anything.
What hits the inbox gets read, matched, and turned into a ready entry. Your team approves instead of retypes.
What ran, what closed, what slipped. On one screen, without asking anyone for a status update.
Built in your accounts, documented in plain English, and your people are trained before I leave.
Fire me the day after it goes live and you keep everything: the accounts, the workflows, the documentation.
What would your team do with the hours back? The first system is live in 14 days.
I have worked in and built automations across nonprofits, legal, marketing, sales, insurance, real estate, and e-commerce. The work changes by industry, but the useful systems usually sit around follow-up, intake, handoffs, reporting, documents, and revenue operations.
Loop design, tool use, planning, self-correction, evaluation. From a first agent to one that can run overnight without blowing up.
Model Context Protocol servers that connect Claude and other models to your real systems. Built, reviewed, or rescued.
Skills, hooks, slash commands, custom harnesses. The canonical surface for autonomous coding agents, done right the first time.
Chunking, embeddings, retrieval eval, production hygiene. Why your RAG hallucinates and how to make it stop.
Offline eval, regression testing, trajectory grading, gold-set discipline. So you know when a prompt change is an upgrade, not a coin flip.
Headless agents, scheduled loops, budget controls, failure routing. Moving from copilot to autopilot without losing the brakes.
I am one person. That's the pitch.
The person you meet on the call is the person who thinks about your problem, writes the code, and replies to your late-night "wait, what about this edge case" Slack message.
If the right answer is "don't build this," I will tell you and hand you back the money for the rest of the engagement.
I like Claude. I use GPT and Gemini. I recommend open source when it's right. I have no partnership checks to chase.
Every opinion I have is one I formed by doing it in production: agents on a schedule, MCP servers on real systems, retrieval pipelines serving real users.
My framework and expertise library are public and free. Consulting is what happens when a team needs your specific problem solved, not a general read.
"I was as skeptical as could be. I didn't trust AI and didn't think it would work, but Sam convinced me to give it a try and I'm so glad I did. He's built workflows that save me time every day." Ashel Troncoso, Owner, ATI Insurance Group
Depending on scope, a typical engagement delivers some combination of:
Different tools for different jobs:
| Situation | Right move |
|---|---|
| You have one hard problem, no AI team, want it solved in weeks. | AI consultant |
| AI is core to the product, you will ship features monthly for years. | Full-time hire |
| You have an AI team but nobody senior; direction is drifting. | Advisory consultant |
| You want a discovery pass before committing to a hire or vendor. | Consultant, short engagement |
| You need 40 hours a week of AI engineering for 12+ months. | Hire. Do not retain a consultant. |
Flat pricing, no hourly billing, ever. You know the number before we start, and it does not move.
One working system, built in your accounts, live in 14 days. $1,000 flat, my founding rate for the first five clients, then $2,500.
If it is not working in 14 days, you do not pay.
Three or four systems built alongside your team, $6,000 to $10,000 flat. Quoted only after a first build, from inside your setup, never from guesses on a call.
Optional, never forced. Everything kept running, tuned, and reported on, flat monthly from $297. You own the system either way.
AI consulting is advisory and implementation work that helps a team figure out where AI fits in their business, what to build, how to build it, and how to keep it running. Good AI consulting produces working systems, not slide decks.
Flat pricing, no hourly billing, ever. Most clients start with one working system built in 14 days for $1,000 flat (my founding-client rate; it rises to $2,500 after the first five). Larger engagements of three to four systems run $6,000 to $10,000, and I only quote those after a first build, from inside your setup. Ongoing help is flat monthly, starting at $297.
Both. Advisory work is short and outputs direction. Build engagements produce working agents, MCP servers, Claude Code harnesses, or retrieval pipelines running in your environment.
Claude and the Anthropic API, Claude Code, MCP servers, agent frameworks (LangGraph, Claude Agent SDK, CrewAI), vector stores, evaluation tooling, and whatever glue the specific problem needs.
Founders, operators, and engineering leaders at seed-to-series-B companies shipping AI features, plus established businesses exploring where generative AI fits without burning a year of budget learning in public.
Yes, happy to. Most engagements start with one.
Remote by default. On-site for a kickoff or a stuck-point intensive, case-by-case.
That's literally question one of the intro call. If the honest answer is "automation with no AI would be faster and cheaper," I will tell you that. It's saved clients more than AI ever has.
Send the process, what breaks today, and what a win would look like. If AI is not the right answer, I will say that too.
Start with one workflowThe definition that shapes every decision downstream.
The Model Context Protocol, the USB-C of AI.
ReAct, planning, reflection, tool use. The shapes that ship.
How to tell whether your prompt change helped.
The definition, minus the hype.
The most valuable page in the agents section.