AI automation services, built and shipped by one person
I build AI assistants, agent workflows and automations for small businesses, agencies and studios, and I run them in production. This is the work an AI automation agency sells, done by the person who writes the code: I scope it, build it, ship it and stay responsible for it. Systems I have built are live with paying customers, not sitting in a case study deck.
What are AI automation services?
AI automation services means someone builds the software that does a repetitive part of your business for you, using language models where judgement is needed and plain code everywhere else. Search for the term and you get two very different answers: platforms you sign up for and configure yourself, and people who build the thing for you. This page is the second kind.
The practical difference from classic automation: a fixed rule breaks the moment the input varies. A customer writes "can I move Thursday to next week?" instead of picking a slot, and a rule-based flow gives up. An AI layer reads it, resolves it against your calendar, replies in your tone and escalates the cases it should not decide alone.
Is this an AI automation agency?
An AI automation agency builds and runs AI assistants, agent workflows and automations for other companies. This is that work done by one person. I scope, build and ship the system myself, which means no handoff between the person who sold it and the person who codes it, and no team to bill for coordination. It also means I take on a limited number of projects, and I will tell you when something is not a fit.
The honest version of the tradeoff:
| One engineer (this) | A typical agency | |
|---|---|---|
| Who writes the code | The person you emailed | A team you meet later |
| Scoping | Email, then a written scope and a fixed quote | Sales call, then a proposal |
| Capacity | Limited, a few projects at a time | Scales with headcount |
| Accountability | One name | Shared across roles |
| Cost drivers | Scope, integrations, how much has to run unattended | The same, plus overhead |
What can actually be automated?
The candidates that reliably pay off are repetitive, text heavy and bounded by rules you can write down:
- Customer conversations. Booking, rescheduling, FAQs, intake and qualification, on the channel your customers already use. fahrchat is exactly this: a live WhatsApp assistant that books real appointments for a driving school, built on Twilio for messaging and Stripe for billing.
- Unattended background work. Watching sources, aggregating data, scoring records, drafting and sending messages, tracking replies. JobMachine, the outreach system behind this site, does that daily without me touching it.
- Product features that happen to use AI. An assistant, a scoring model, a generated report inside your own app. baselit turns a selfie into a score and a routine. aimentionsyou.com tracks how often ChatGPT, Claude, Perplexity and Gemini mention a brand and turns it into a report.
- Creative and studio pipelines. Tooling that sits between 3D and generative AI. AIRESHAPE is a Blender plugin I built for a client that turns simple 3D animations into cinematic AI video, orchestrating Gemini, Kling and Meshy.
What should not be automated?
Anything where every case needs a human decision, anything where a wrong answer is expensive and unrecoverable, and anything you cannot describe as a process yet. If you cannot write down what a good outcome looks like, an AI system will not invent one for you. In those cases the useful first step is smaller: automate the retrieval, the drafting or the routing, and leave the decision with a person.
How does an AI automation project run?
- You email me the process that eats your time, in plain language.
- I reply with a few scoping questions: volume, channels, the tools it has to touch, what must never happen.
- You get a written scope and a fixed quote. No sales call, no booking widget, no packages.
- I build it end to end: integrations, the model layer, the fallbacks, the admin view you need to trust it.
- It goes live behind a narrow path first, then widens once the edge cases are known.
- Handover or ongoing operation, whichever we agreed.
What does AI automation cost for a small business?
To judge whether it is worth it, put your own numbers into the website chatbot ROI calculator first. There is no price list here, because a WhatsApp booking assistant and a full agent workflow are not the same object. What moves the number: how many systems it has to integrate with, whether it runs attended or unattended, how many languages and channels, how much failure handling it needs, and whether you want the code handed over or operated for you. Message volume and model usage are an ongoing cost that belongs to you, not to me.
Why hire one engineer instead of a team?
Because most small business automation is not a headcount problem, it is a scoping problem. The systems here were designed, built and run by one person: custom AI agents for the autonomous parts, plain software for everything else. If you would rather buy a fixed share of my week than a single project, that is fractional AI engineering.
See the work and the proof, then write me at hello@marcomori.net. I reply personally, usually within a day.
Questions people ask
What is an AI automation agency?
An AI automation agency builds and runs AI assistants, agent workflows and automations for other companies. This is that work done by one person: I scope, build and ship the system myself, and the proof section shows systems that are live with real customers rather than case studies about them.
Is this an agency or a freelancer?
One person. I do not have a team, account managers or subcontractors, so the person you email is the person writing the code. That is a limit on how much I take on at once, and the reason nothing gets lost between a brief and an implementation.
What kind of processes are worth automating?
Anything repetitive, text heavy and rule bound that a person currently does by hand: answering the same customer questions, booking and rescheduling, moving data between tools, monitoring sources, drafting and sending outreach. If a task needs judgement on every single case, automation is the wrong tool.
What does an AI automation project cost?
Pricing is project-based and quoted after a short scoping conversation by email. No sales calls, no fixed packages. A WhatsApp booking assistant and a full agent workflow are very different scopes.
How does an AI automation project start?
With an email describing the process that eats your time. I reply with a few scoping questions, then a written scope and a fixed quote. No sales call, no booking widget, no packages.
Do you work with small businesses, or only startups?
Both, plus agencies and creative studios. fahrchat is a WhatsApp assistant a driving school uses to book appointments, which is about as small-business as it gets, and baselit is a consumer app on the App Store.
Can you work with the tools we already use?
Usually yes. I build on APIs rather than replacing your stack, and I have shipped against Twilio and the WhatsApp Business API, Stripe, Supabase and Vercel. If a tool has no usable API, I will say so before you pay for anything.
Who maintains the system after it goes live?
That is part of scoping. Some clients want the code handed over and run it themselves, others want me to keep operating it. Systems I run for myself, like JobMachine, run daily and unattended, so I build with that expectation either way.
Start with an email. Describe the process that eats your time. I reply with a few scoping questions, then a written scope and a fixed quote. No sales call, no packages.