AI agents and automations
I build automations that run on your own infrastructure instead of per-task pricing, and AI agents that draft work for a person to approve. Typical jobs are an inbox that drafts replies and quotes, invoice and document processing, and the glue between your tools. Nothing is sent to a customer without a human approving it.
What is included
- A map of the workflow as it runs today, and the steps worth automating
- Automations built in code or on n8n, whichever costs you less over time
- AI drafting for replies, quotes and summaries, with an approval queue
- Document and invoice extraction into your system of record
- Logging, so every run can be inspected and a failure explained
- A kill switch and clear limits on what the agent may do
Who this is for
- Businesses whose Zapier or Make bill grows with every extra task
- Teams answering the same customer emails and quote requests all day
- Finance or ops staff retyping invoices and documents by hand
How pricing works
You get one fixed price for the whole build, agreed before I start. After handover, optional monthly maintenance covers updates, fixes and small changes. If you would rather run it yourself, you get the code and a handover walkthrough instead.
Questions people ask
Will an AI agent send messages to my customers by itself?
Not unless you ask for that and we agree on limits. My default is human approval: the agent drafts, a person reviews and sends. That keeps wrong prices and wrong promises out of customer inboxes.
Why not just use Zapier, Make or n8n?
Often you should. They are fast to start with and fine at low volume. Custom automation starts to win when volume is high, when the logic is awkward for a visual builder, or when you want the data to stay on your own servers.
What does it cost to run after you build it?
Hosting for a small automation is usually small, and AI usage is billed by the model provider based on how much you use it. I show you the expected running cost before you commit.
How do you stop the AI from making things up?
The agent drafts from your own data such as price lists and past replies, and a person approves the result. Where a number matters, such as a price, it is read from your data rather than generated.
Specific jobs I take on
The concrete problems people bring me under this service. All solutions
- AI inbox that drafts the reply and the quote for you to approveA shared inbox where an AI drafts the reply and a quote from your own price list, and a person approves before anything is sent. How it is built and where it fails.
- An AI chatbot that answers from your own content, with human handoffA website or WhatsApp-style chatbot that answers customer questions from your own documents, hands over to a person when it is unsure, and states its limits.
- Custom Chrome extensions and web scrapers that save to Sheets or a databaseManifest V3 Chrome extensions and scheduled scrapers that extract data into Google Sheets or a database, with the legal and technical caveats explained plainly.
- Fix emails landing in spam: SPF, DKIM and DMARC set up properlyInbox placement fixes for business email: SPF, DKIM and DMARC setup, alignment, bulk sender rules from Google and Yahoo, and what I check before changing DNS.
- Gmail sorting, Sheets extraction and auto-replies with n8n, Zapier or MakeAutomations that sort Gmail, pull data into Google Sheets and send auto-replies. How I choose between n8n, Zapier, Make and custom code, and what to expect.
- Private AI assistant over company documents and ERP data, read-onlyA local or on-premise AI assistant that answers questions from your own documents and ERP data, read-only, with permissions mirrored. Retrieval, models, hardware, quality.
Related reading
Tell me what is slowing your business down.
Twenty minutes, no charge. If I can fix it, you get a fixed price. If not, I will say so.