Odoo Does 90 Percent of What We Need. How Do We Automate the Last 10?

To automate the last 10 percent of Odoo, first find which manual steps are left, then pick the lightest tool for each: Odoo’s built-in automation rules, a small custom module, or an AI agent for work that involves reading messy text. Most gaps close with simple rules. Only the untidy ones need AI.

Where the “last 10 percent” comes from

One Odoo reviewer on Capterra described it as a 90 percent solution for administrative automation, with the remaining 10 percent difficult and burdensome. The percentages are the reviewer’s own impression, not our measurement, but the feeling is familiar to anyone who has run Odoo for a year. Standard flows such as quotations, invoices and stock moves work well. Then a handful of business-specific tasks remain that someone still does by hand every day.

Those tasks are usually unique to your company: a customer who needs a certain document format, an approval that depends on who ordered what, a supplier email that must be turned into a purchase order. This post helps you sort them, choose the right level of automation, and avoid over-building. If you want help afterwards, our Odoo AI and CRM service covers this kind of work.

Step one: list what is left, and count it

Spend a week writing down every task that people still do outside Odoo or by retyping into it. Note what triggers it, how long it takes, how often it happens, and what goes wrong when someone forgets. You do not need precise figures. Even a rough tally of “daily, a few minutes” versus “weekly, an hour” shows where the time is going.

You will usually find the list falls into four kinds of work: moving data between Odoo and another tool, applying a rule that Odoo does not know about, reading something unstructured (an email, a PDF, a WhatsApp message) and turning it into a record, and chasing people or documents on a schedule. Each kind has a different best fix.

Step two: choose the lightest tool that works

Rules inside Odoo come first

Odoo includes automation rules, scheduled actions, email templates and, in some editions, approval rules and a studio-style customisation tool. Many gaps are closed by a rule such as “when a sales order passes this value, ask a manager to approve” or “when an invoice is a week overdue, send this reminder”. These are cheap, auditable and easy to explain to your team. Check what your edition allows before paying anyone for custom code.

A small custom module when the rule is specific

When your rule needs fields, calculations or screens that standard Odoo lacks, a small custom module is often the cleanest answer. Keep it narrow. Every module you add must be maintained through Odoo upgrades, so a short, well-documented one is far better than a sprawling one. Our guide to Odoo customisation with AI covers how to scope this sensibly.

Integrations when the data lives elsewhere

If the manual work is copying between Odoo and a marketplace, a courier, a bank or another system, you need a connection. Marketplace connectors differ in quality, and some do not work well on every Odoo hosting option, so test before relying on one. Our post on Odoo integration services explains the choices.

An AI agent when the input is messy

Rules need clean inputs. If the first step is reading a free-text email, a supplier PDF or a customer message and deciding what it means, a rule cannot do it. An AI agent can read the document, extract the fields and create a draft record in Odoo for a person to confirm. That is where AI earns its place, and where it is overkill for everything else. See our guide to AI agents for how this looks in practice.

Step three: automate one thing and watch it

Pick the single task with the clearest trigger and the most repeated effort, and automate only that. Run it alongside the manual way for a short time so you can compare. Look at errors, not only time saved. A good automation is boring: it runs, it logs what it did, and someone notices quickly when it does not.

Make failures visible

A frequent complaint in Odoo forums is an automation rule that works when run by hand but never fires by itself. On self-hosted installations, one thing worth checking is whether the scheduler is actually running, though causes vary and we cannot diagnose a specific case from outside. The general lesson holds: every automation needs a way to tell you it failed. Send an alert, keep a log and review it on a schedule. Silent failure is how a helpful automation turns into lost orders.

Keep a person where judgement is needed

Use approval steps for anything that moves money or commits you to a customer. This is the same principle we apply to AI drafting: the system prepares, a person signs off, and you widen automation only after trust is earned.

A worked example, kept deliberately simple

Imagine a distributor whose customers email purchase orders as PDFs. Someone reads each PDF, finds the customer in Odoo, retypes the lines into a sales order, and checks prices. The list of steps shows three separate jobs. Matching the customer and checking prices against a price list can be done by rules inside Odoo. Reading the PDF is the messy part, so that is where an AI agent could extract the lines and create a draft sales order. Confirming the order should stay with a person until the drafts have proved reliable.

Splitting the job this way matters. A single big “automate order entry” project is hard to scope, hard to test and hard to fix. Three small pieces can each be tested alone, and if the AI step is ever wrong, the rules and the human check still catch it. This is the pattern we recommend for most leftover Odoo work.

What to leave manual

Not every leftover task should be automated. Work that happens twice a month, depends on a human relationship, or changes every time is cheaper to leave alone. A rule of thumb: if building and maintaining the automation takes longer than doing the task for a year, skip it. Be wary of any supplier who promises to automate everything. A realistic proposal names the tasks it will not touch and why.

Also be careful about upgrades. Heavily customised systems become expensive to move to a new Odoo version. If you are already struggling with upkeep, Odoo support and maintenance may do more for you than another automation, and our post on what a support contract covers explains the detail.

Questions to ask before you spend money

Ask who will maintain the automation after it is built, and what happens at the next Odoo upgrade. Ask how you will know when it fails, and whether you can see a log. Ask whether the same result could be achieved with a built-in rule. And ask what the supplier would not automate. If the data is sensitive and must stay on your own servers, an on-premise AI setup is possible, and for anything unusual, custom AI agent development is usually the route. We will tell you honestly after seeing your processes whether that is needed or a simple rule will do.