Most AI automation agencies stack junior operators on no-code tools and call it "agentic." I am Long Nguyen — one senior operator with years in the Zapier and n8n automation world. I wire agents and automations into the tools you already use, then keep them running as the AI landscape shifts. Part of the fractional AI lead relationship, not a one-time deliverable.
There is a pattern in the AI automation agency space right now: a senior salesperson who understands the pitch, a team of junior operators who run the no-code tools, and a delivery model that looks fine on a Loom walkthrough and breaks six months later when you are not looking. The senior person is not on your account after the sale.
I work differently. I am the person on the sales call and the person who builds the automation and the person you reach when something breaks. Years in the Zapier and n8n world means I know the failure modes — the Zap that silently drops records when the upstream API changes its schema, the n8n workflow that works fine until the rate limit gets hit at 4x the volume you tested at. I build to avoid those failure modes, not just to pass a demo.
AI automation for growing teams is part of the fractional engagement — not a separate product. When I am on retainer, the automations I build stay current as the models, the tools, and your workflows evolve. That continuity is most of what makes them useful long-term.
If you want to see what hands-on AI builds look like, the work page has examples. If you want to understand the broader consulting layer, AI consulting covers that. If you already know what you need automated, talk to the duck.
Agents that reason, classify, and decide — not just trigger-and-action chains. I build agentic workflows that handle the edge cases, escalate when they should, and keep running without a human in the loop. Claude Agent SDK for the orchestration, MCP servers for the tool access, your existing stack for the data.
Years in the Zapier and n8n world before the AI layer existed. I know where no-code tools earn their slot and where they fall over. I use them where they are genuinely the right choice, wire AI into them where that adds real value, and replace them with custom code when the job demands it.
Your internal admin, your CRM, your support queue — most of your tools do not have a native AI integration that does anything useful. I write the MCP servers that give agents first-class access to your systems, so they can look things up, take action, and hand off correctly. See also: AI consulting for the broader strategy layer.
Contracts, invoices, intake forms, support tickets — anything that arrives as unstructured text and needs to become structured data. Extraction pipelines that run on Claude, validate their own output, and route exceptions to a human review queue instead of silently dropping them.
Drafts, summaries, first-pass responses — the things your team writes from scratch a dozen times a week. AI workflows that generate a strong first draft and get it in front of a human for the final call. Not fully automated publishing. Useful automation that respects that judgment still matters.
Automations break. APIs change. Models get updated and the prompt that worked last quarter sometimes does not work this quarter. On retainer I keep the stack running — alerts for failures, quarterly reviews of what has drifted, and updates shipped before your team notices something is wrong.
Most engagements start with two or three of these and expand from there. The audit tells us which ones earn their keep first. See also: AI consulting if the primary need is wiring models into your product stack rather than internal workflows.
Describe what you are trying to automate, what tools your team runs on, and what you have already tried. The duck reads the shape of the engagement — one-off build, retainer, or both — names a price in plain English, and sends a brief to both of us.
I map what your team does that is repetitive, where AI can take over the judgment step, and what you already have running (working or broken). The output is a short written roadmap: the two or three automations worth building first, in the order that compounds.
Agents, MCP servers, orchestration layer — shipped into your stack, your accounts, your infra. You see the work as it lands. I test against real data, not demo inputs. When something surprises us mid-build, I adjust the architecture the same afternoon.
You get the source code, the deploy configs, a walkthrough of how each automation works, and a short reference for what to watch. After that: bring it in-house, keep me on retainer to maintain and expand it, or call me in for the next build.
AI automation sits inside a broader fractional engagement. Depending on what your audit surfaces, the work may also touch AI consulting strategy or deeper model integration.
Five-minute conversation with the duck — no intake form, no qualification grid. Describe what your team does manually that should not be manual, what tools you run on, and what you have already tried. The duck reads the shape of the AI automation engagement, names a price, and sends a brief to both of us. If I am the wrong fit, it says so.