Automation that removes work, not automation that impresses.
We build workflow automation and AI systems into Odoo — targeted at specific repetitive tasks, measured against the hours they actually save.
Most "AI for business" is a chatbot nobody uses
We build automation that removes specific, measurable work — not features that demonstrate well and then sit idle. The starting question is never "where could we use AI?" It is "which task is your team doing repeatedly that a machine could do more reliably?"
Sometimes the answer involves a language model. Often it does not — a well-designed workflow rule solves the problem with less to go wrong. We will tell you which.
Automation that earns its place
Document processing
Supplier invoices, purchase orders and delivery notes read automatically and posted into Odoo, with exceptions routed to a human rather than guessed at.
Lead qualification
Inbound enquiries scored and routed on the signals that actually predict a deal — company size, sector, message content — so your team works the right ones first.
Anomaly detection
Stock counts that drift, margins that slip, invoices that look wrong. Surfaced when they happen rather than at month end.
Workflow automation
Approvals, escalations, reminders and handoffs that run without anyone remembering to chase.
Reporting narratives
The weekly operations summary written from your own data, so the numbers arrive with context rather than as a spreadsheet.
Data enrichment
Records completed and deduplicated automatically, because dirty data is what makes every downstream system worse.
Start with the task, not the technology
Find the repetitive work
We sit with your team and watch what they actually do. The candidates are usually obvious once you look — the task everyone dreads, the one that generates the most errors, the one that only one person knows how to do.
Quantify it
Hours per week, error rate, cost of getting it wrong. If we cannot put a number on it, automating it is hard to justify.
Build the smallest useful version
One workflow, in production, doing real work. Not a pilot that runs in parallel forever.
Measure, then extend
If it saved the time we said it would, we do the next one. If it did not, we find out why before building more.
A note on where AI does not belong. Anything requiring judgement you would defend to an auditor, a regulator or a client. We automate the mechanical work so your people have time for the judgement — not the other way round.
What clients usually ask
Does our data leave the building?
- Not unless you want it to
- We can run models on your own infrastructure
- Where a cloud service makes sense, we tell you exactly what is sent
- Nothing is used to train anyone else's model
What does it cost to run?
- Workflow automation: no ongoing cost beyond hosting
- Document processing: cost per document, usually cents
- Language models: priced per use, modelled before we build
- We show you the running cost before you commit
Where this fits
Automation works best on top of a system that already holds clean, structured data. If your processes are still spread across spreadsheets and email, the higher-return move is usually getting the ERP right first — then automating on that foundation.
Where you already have Odoo running properly, this is often the fastest way to get more out of it.
Common questions about AI and automation
Do we need AI, or would simple automation do?
Usually simple automation. A workflow rule that routes approvals or chases overdue invoices solves the problem with far less to go wrong than a language model. We reach for AI when the task genuinely requires interpreting unstructured input — reading an invoice, classifying a message, summarising a document.
We will tell you which category your problem falls into, and the cheaper answer is often the right one.
Does our data leave our systems?
Not unless you choose that. We can run models on your own infrastructure so nothing leaves. Where a cloud service is the sensible option, we tell you exactly what is sent, and nothing is used to train anyone else's model.
What does it cost to run once it is built?
Workflow automation has no ongoing cost beyond your existing hosting. Document processing is priced per document, typically cents. Language model usage is priced per call and we model the running cost before building, so you see the number before you commit.
How long does an automation project take?
A single well-defined workflow is usually two to four weeks from scoping to production. We deliberately build the smallest useful version first, measure whether it saved the time we predicted, and only then extend.
Can this work if we are not on Odoo?
Yes, though it works considerably better if you are. Automation depends on clean, structured data. If your processes still live in spreadsheets and email, getting the ERP right first usually gives a better return than automating around the mess.
Which task should we look at first?
Tell us the job your team dreads. We will tell you honestly whether it is worth automating.
Book a discovery call