What it is
AI automation is the practice of wiring the repetitive parts of marketing and sales to run themselves. Enquiries routed and answered in minutes, CRM records updated without typing, follow-ups sent on schedule and reports assembled from the systems that already hold the numbers. We build the workflows on n8n and connect them to your tools through their APIs, and where a step needs reasoning the workflows call the frontier AI models from OpenAI, Anthropic, Google and the rest. It suits any business where the gap between an enquiry and an answer is costing customers, and any regulated market where every automated step must leave a record behind it. We do not promise a percentage of time saved before we have seen your process, because the honest number depends on how much of your week is currently spent moving information between boxes. We promise machinery you can read, documented and handed over, that does the dull work every time rather than most times.
Most enquiries are not lost to competitors. They are lost to silence. The form nobody saw until Monday morning, the quote that took three days to leave the building, the review request written once and never sent again. Every one of those gaps is a small leak, and together they are the difference between a full diary and a quiet one. Automation closes them quietly and permanently, which is why it belongs beside the marketing that creates the demand in the first place.
We build the workflows on n8n, the open source automation platform, and connect them to the tools you already run through their APIs rather than asking you to migrate anything. Where a task needs judgement instead of rules, drafting a reply, classifying an enquiry, summarising a long thread, the workflow calls one of the frontier AI models from OpenAI, Anthropic, Google and the rest, working through their APIs inside a structure you can read. The models are used where they genuinely earn their place and nowhere else, and every step is documented so that a new person joining your team can open the hood and understand what moves what.
The discipline underneath is the same one we apply to everything else. Measure first, so the saving is a figure rather than a feeling. Design the human review steps before building the clever ones, because a machine that confidently does the wrong thing costs more than a person doing it slowly. And hand the whole thing over properly at the end, with documentation and access, because automation you cannot change yourself is a subscription wearing a disguise.
What we do
Lead capture and response
Every enquiry routed, answered and logged within minutes, whichever channel it arrived through. Forms, calls, chats and inboxes feed one place, the right person is notified with the context attached, and the customer hears back while they are still shopping. For local service businesses this is the single highest-return automation there is, because the first business to respond usually wins the job.
CRM and pipeline automation
Follow-ups, reminders and record updates that happen without anyone remembering to do them. Stages move on real triggers rather than optimism, stale deals surface before they die quietly, and the pipeline in your CRM reflects what is actually happening instead of what someone last had time to type. The books stay clean because nothing depends on Friday afternoon data entry.
WhatsApp and email flows
Messages triggered by real behaviour, written with restraint and stopped when they should stop. Appointment reminders, quote nudges, review requests and win-backs that fire on what a customer actually did rather than on a calendar guess. Consent is recorded first, because in regulated markets every message is a document that may one day be read back to you.
Reporting automation
The numbers gathered, checked and delivered on schedule, so nobody assembles a report by hand again. Dashboards and digests that pull from the systems that already hold the truth, with the anomalies flagged rather than buried. The monthly scramble becomes a Monday morning read, and the figures are the same ones anyone can trace back to source.
How we work
Audit
We follow a real enquiry through your business and mark every place it waits, gets retyped or gets forgotten. We also look at what your tools can already do, because the fastest win is often a feature switched on rather than a workflow built. The output is a map of the leaks with a rough value attached to each.
Plan
A short build order agreed with you. What gets automated first, which steps stay human and why, where the AI models appear and where they are deliberately kept out. Compliance review happens here rather than after launch, so nothing has to be unwound later.
Build
The workflows get built in n8n against your live systems, tested with the awkward cases before anything goes live. Error handling and alerts are part of the build rather than an afterthought, because automation that fails silently is worse than no automation at all. Everything is documented as it is built, not written up from memory afterwards.
Prove
Response times, follow-up completion and hours returned, measured against the audit figures and reported on the schedule you prefer. What works gets extended, what does not gets simplified or switched off. The honest end state is fewer moving parts than people expect, doing more work than anyone has time to do by hand.
What it costs
Automation work is priced by the build and the running of it. The build depends on how many workflows, how many systems they touch and how clean the data is underneath, since connecting two tidy tools costs far less than untangling a decade of manual entry first. The running is a retainer for monitoring, small changes and new workflows as the business grows, because automation that is not maintained drifts quietly out of date. The AI models are billed by the providers on usage, and we keep those costs visible rather than folding them into a fee. We scope and price before anything starts, and we tell you plainly when a problem is a people problem that no workflow will fix.
Who it is for
For local service businesses where a missed callback is a lost customer and the owner is still answering enquiries at nine at night. For regulated markets that need every automated step recorded, reviewable and defensible, where a careless workflow becomes evidence. For marketing and sales teams losing hours to the typing between systems. What they share is a leak between interest and answer, and the sense that the tools they already pay for are not talking to each other.
Two kinds of business come to us: local service businesses and companies in regulated markets. See everything in Be chosen.
Questions we get asked
What do you build on?
n8n, the open source automation platform, connected to the tools you already use through their APIs. Where a step needs reasoning, drafting or classification, workflows call the frontier AI models from OpenAI, Anthropic, Google and others, through their APIs and on your accounts where that matters. Everything is documented and handed over, so nothing is a black box you cannot open. We choose n8n because the workflows are readable, portable and yours, and because it can be self-hosted by businesses that need their data to stay put.
Does this replace our team?
No. It replaces the work nobody should be doing by hand. The typing between systems, the chase-up emails, the monthly scramble for figures, the reminders people forget to send. Your people keep the conversations, the judgement calls and the relationships, which is what they were hired for in the first place. Clients who automate the dull work reliably find their team does better work, not less work.
How do you measure automation?
Response time, follow-up completion and hours returned to the business. We measure before and after, and we show both figures, because an automation project with no baseline is a story rather than a result. If the saving is too small to be worth reporting, we will say so before you spend the money rather than after.
What about AI mistakes?
Anything that touches a customer gets a human review step, and anything in a regulated market gets an audit trail showing what the model did and why it was approved. The models are tools with known failure modes, and we design around those rather than pretending the technology is infallible. Where a mistake would be costly, the machine drafts and the person decides. That line is drawn deliberately, with you, before the build starts.
How is this different from your email automation service?
The email service decides what gets said and when, the strategy and the words. This service builds the machinery underneath, the routing, the triggers and the handoffs across all your systems, of which email is usually one. Many clients run both, and the two fit together without overlap.
What if we already use Zapier or Make?
Fine. We work in what you have when it is doing the job, and we recommend a move only when the tool is genuinely the limit. A migration is a cost paid once, so we treat it like any other investment and show the arithmetic before asking for a decision.
Can you work in a regulated market?
Yes, and it changes the design rather than just the paperwork. Consent records, review steps and audit trails are built in from the start, every automated action can be traced to what triggered it, and the whole system is documented well enough to show an auditor. Where a question is genuinely a legal one, we point you at proper legal advice rather than guessing in code.
More Be chosen services
- PPCPaid media managed against profit, not clicks.
- Paid searchGoogle and Bing, built around intent and margin.
- Paid socialThe feeds your buyers scroll, and WhatsApp where they actually talk.
- Google ShoppingFeed, structure and bids for ecommerce that has to turn a profit.
- ChatGPT advertisingPlacement inside AI conversations, run as a careful test rather than a promise.
- CROTurning the traffic you already have into more customers.
- Landing pagesPages built for one job, the next click that matters.
Tell us what you need from ai automation. We will tell you what we would do about it.