More time for what matters
Process automation with artificial intelligence
We automate the repetitive tasks that until now needed a person, because information had to be read, understood, or classified before deciding.
Reading and understanding: the step that used to require a person
Automating a process has always worked well when it can be expressed as rules: "if the order status changes to shipped, send the email." This is what we have been doing for years with automations and APIs.
The limitation was in processes that start with information that has to be read and interpreted. That step had to be taken by a person, which is why the process could only be partly automated.
AI solves that step, and the rest of the workflow is still traditional automation: reliable, repeatable, and inexpensive.
Traditional or AI process automation: when to use each one
Not every process needs AI. The key question is whether the same input always has to give the same output.
Traditional automation: same input, same output
It is essential when the result always has to be the same for the same data. The PDF invoices generated by the ERP itself always have the same format: a program extracts their data without sending it to an AI model, which would add cost and response time. And to export that data to another platform, an integration with defined communication protocols is used, without AI. It is the basis of custom automation and API development.
- Speed: no need to wait for a model to respond.
- Reliability: the same data always gives the same result.
- Security: the information is not sent to any AI service.
- Control: every step can be reviewed and audited.
AI automation: when interpretation is needed
It is the option when the information has no fixed format or its content has to be understood. For example, knowing whether a customer writing to technical support is happy or angry: with traditional programming it would mean searching for keywords and synonyms, and the result would be unreliable. With AI, it is feasible.
- Data protection: the model has to be chosen carefully, because not all models are suitable for every kind of information. When data cannot leave the company, the option is private AI.
- Cost and time: each query to the model has a small cost and takes a few seconds.
- Measured results: it is tested with real documents before going into production.
An in-between case: supplier invoices
Each supplier sends invoices in its own format, so a program with fixed rules is not enough. But they do not need to be sent to a general-purpose AI model either: there are services specialized in reading invoices, such as the invoice processor in Document AI, from Google Cloud, subject to its data processing terms for businesses. It offers more confidentiality guarantees than freely using a commercial model, and it is integrated with traditional programming through its API: the rest of the workflow does not need AI.
Examples: what each one solves
Extracting the data from the PDF invoices generated by the ERP itself
How it is solved: Traditional programming
Why: They always have the same format: a program reads them at no cost per document.
Exporting that data to the gestoría (the accounting firm that handles taxes), the bank, or another platform
How it is solved: Traditional programming
Why: An API integration or a defined exchange format gives an exact result.
Checking that a Spanish tax ID (NIF), an IBAN, or a date is valid
How it is solved: Traditional programming
Why: There is a validation algorithm: AI adds nothing and could make mistakes.
Sending an alert when a product's stock falls below the minimum
How it is solved: Traditional programming
Why: It is a fixed rule: if the stock is below the minimum, the alert is sent.
Reading supplier invoices and delivery notes, each in its own format
How it is solved: Specialized service and traditional programming
Why: The format changes, but there are services designed for that document, such as Google Cloud Document AI.
Knowing whether a customer writing to technical support is happy or angry
How it is solved: AI
Why: It requires understanding the tone of the message, not searching for words.
Detecting whether a message contains the company's financial information before it goes out
How it is solved: AI
Why: The same information can be expressed in many ways.
Sorting incoming emails by topic and priority
How it is solved: AI
Why: Each sender writes in their own way.
Finding a supplier's product in the company's own catalog even if it has a different name
How it is solved: AI
Why: It requires comparing by meaning, not by the exact text.
Month-end invoices: generating them, sending them, and writing each customer's email
How it is solved: Both
Why: The workflow is traditional and the AI only writes the email. This is the example explained below.
| Task | How it is solved | Why |
|---|---|---|
| Extracting the data from the PDF invoices generated by the ERP itself | Traditional programming | They always have the same format: a program reads them at no cost per document. |
| Exporting that data to the gestoría (the accounting firm that handles taxes), the bank, or another platform | Traditional programming | An API integration or a defined exchange format gives an exact result. |
| Checking that a Spanish tax ID (NIF), an IBAN, or a date is valid | Traditional programming | There is a validation algorithm: AI adds nothing and could make mistakes. |
| Sending an alert when a product's stock falls below the minimum | Traditional programming | It is a fixed rule: if the stock is below the minimum, the alert is sent. |
| Reading supplier invoices and delivery notes, each in its own format | Specialized service and traditional programming | The format changes, but there are services designed for that document, such as Google Cloud Document AI. |
| Knowing whether a customer writing to technical support is happy or angry | AI | It requires understanding the tone of the message, not searching for words. |
| Detecting whether a message contains the company's financial information before it goes out | AI | The same information can be expressed in many ways. |
| Sorting incoming emails by topic and priority | AI | Each sender writes in their own way. |
| Finding a supplier's product in the company's own catalog even if it has a different name | AI | It requires comparing by meaning, not by the exact text. |
| Month-end invoices: generating them, sending them, and writing each customer's email | Both | The workflow is traditional and the AI only writes the email. This is the example explained below. |
Processes that can be automated with AI
They all start the same way: with information that arrives unstructured. AI turns it into data and the workflow continues with its usual rules.
Incoming emails, messages, and files
An email, a Telegram message, or a file. The AI interprets what is being requested and extracts the relevant data, and the workflow turns it into a task in the ERP, for example in Odoo.
Forms and free-text messages
What a customer writes in their own words in a web form or a message becomes a record with its fields, ready for the CRM.
Automatic summaries of emails, meeting minutes, and incidents
Email threads, meeting minutes, or the week's incidents, summarized in a few lines for whoever has to decide.
Repetitive texts and translations
Draft replies, product descriptions, or translations of product sheets, which a person reviews before they are sent or published.
Delivery notes, orders, and certificates
Each supplier sends them in its own format. A specialized service or an AI model extracts the fields, and the workflow checks them against the ERP data and records them where they belong.
Spreadsheets in different formats
Excel sheets that each department or customer fills in with its own format, with columns that change name or order. The AI interprets them and the workflow loads the data into the right program.
AI automation example: month-end invoices, from the ERP to the customer
The ERP prepares the drafts
At the end of the month, the workflow generates draft invoices from the hours the team has logged in the ERP, for example in Odoo.
Accounting reviews them
The accounting manager receives a notification with the drafts ready for review. No invoice goes out without their approval.
The AI writes the cover email
Each approved invoice goes out with an email written for its customer, not with a template that is the same for everyone, and with the invoice attached.
The rest of the workflow, with fixed rules
The AI only writes the email. Generating the drafts, notifying, and sending are done by the workflow with fixed rules, and the approval is up to a person.
Scope of automations: channels, programs, sensors, and APIs
An automation can receive information and act on almost any system that has a way to connect.
Telegram, WhatsApp, and SMS
A request that arrives via Telegram or WhatsApp becomes a task, and notifications go out through the same channel or by SMS. For example, alerting the person in charge when the deadline for a task is approaching.
Email and files
Incoming emails and files that arrive in a folder are classified and turned into the corresponding task or ticket. In the other direction, each client receives a periodic email summary of the progress of their project.
ERP, CRM, and internal programs
The automation reads from and writes to the company's programs: it creates a task in Odoo, updates a CRM record, or generates draft invoices at the end of the month.
Sensors and machines
Data from the machines in a plant can also trigger an automation: if one has been stopped for more than ten minutes, the maintenance manager receives an alert. It is the basis of AI for industry.
Webhooks from third-party tools
Many tools send a notification when something happens: a payment received, a form submitted, or a change in the code repository. The automation acts at that moment: for example, when the payment gateway confirms a charge, the order is marked as paid in the ERP.
Remote APIs
External services that are queried or that data is sent to: banks, carriers, public platforms, or AI models. For example, reviewing public tenders and using Jev to filter the ones that match the company's profile.
At Gilsys we have also streamlined our internal management by integrating AI automations, which is why we can recommend them from first-hand experience.
AI automation or an agent?
They are similar and easy to confuse. The practical difference is who decides the path.
AI automation
The path is decided in advance and is always the same. The AI handles a specific step within it: reading, classifying, or summarizing. Cheaper, more predictable, and easier to check: if something fails, the step where it failed is identified right away.
AI agents
The path is not predetermined: the agent analyzes the situation and chooses what to do, in what order, and how many times, within the limits set for it. Two similar cases can follow different steps. For processes where not every case can be foreseen: AI agents.
Automation tools: n8n, the model, and the company's systems
n8n and orchestration tools
We program the workflows and integrate workflow management tools such as n8n, which make it possible to view and modify them and to see where each run stands.
The right model for each case
Commercial models such as ChatGPT, Gemini, or Claude when the information is allowed to leave and, when it is not, open-source models in a private AI setup within the company.
Connection to the company's systems
The workflow records the information in the relevant system: ERP, CRM, document management system, online store, or custom applications, through system integration.
A log of every run
Every run is logged: what came in, what the AI decided, and what was done. This is what makes it possible to check that it works and to correct it when something fails.
Frequently asked questions about AI automation
How accurate is AI at reading documents?
Is it necessary to switch software?
Which process should be automated first with AI?
How long does it take to get a first automation?
How much does it cost to automate a process with AI?
Does the automation have to be rebuilt if a better AI model comes out?
Which manual task can be automated in your company?
Tell us about it with an example: an email, a document, or a spreadsheet that is processed manually today. With that example, we will see which part AI can do.