Frequently asked questions

We automate the office work that industrial companies still do by hand: orders and quotes, invoices and expenses, email and document handling, and the rest of the tasks where someone moves data from one place to another (price lists, logistics, reports, hand-offs between departments). We replicate the process exactly as the company runs it today, without touching your software; the system does the mechanical part and someone on the team gives the final OK. And from day one we install a counter that measures, on the client's own dashboard, the hours and euros the system gives back.

Order-entry automation is a system that reads the orders arriving by email, PDF or message and enters them into your software without anyone typing them by hand. The AI extracts which product, what quantity and which customer; the system checks each line against the company's master catalogue. It is the flagship product of the orders family, which we call Pedido en Firme (Firm Order).

Zapier and Make move data between apps with fixed rules and route your information through third-party servers. An RPA robot mimics clicks and breaks whenever a screen changes. Alquimia Digital Agency sets the system up on a server owned by the client, hosted in Europe, and uses AI only to read text: if a reference doesn't match your catalogue, the system stops and asks for human validation instead of making it up.

It works best for B2B distributors and industrial firms with a standardised catalogue and repeat orders: electrical supplies, ceramics and construction, packaging, automotive spare parts, or engineering firms with their own product. If you receive lots of orders in free text (emails, PDFs, photos) and type them by hand into an ERP, you fit. If you sell fully custom product with no catalogue, you fit less well.

We don't know until we see your data. That's why the first phase is a technical validation: we look at the state of your data and your processes and tell you honestly whether you'll recover the investment. If you won't, we don't sell it to you.

Between 4 and 8 weeks from the moment we decide to start, depending on volume and complexity. The first improvements show up in the first few weeks.

Around 10 hours in total, spread over several weeks. Roughly 3 hours from you during the technical validation and about 7 more during implementation (check-in meetings, validating examples, fine-tuning). We do the rest in parallel with your operation. It's not a full-time project for anyone at your company.

No. Implementation runs on a separate environment until we validate the prototype. Your day-to-day operation carries on as before. When we move to production, we do it in milestones: first a small flow, we see it works, and we keep expanding. Never an all-or-nothing switch that could leave you without an operation.

No. We work with what you already have. If your ERP has a standard way to connect (an API), we use it. If it doesn't, there are other ways we go through in the meeting.

Three things: access to someone on your team who knows the order flow well (3 hours during validation + 2-3 during implementation), an initial export of your master product catalogue, and controlled access to the email account where you receive orders. Nothing that requires any programming work on your side.

Yes. A modern API is better. If there isn't one, we connect through read-only queries to its database, or by reading and writing flat files (CSV, XML) in a network folder. The innovation happens in the outer layer. We don't touch the core code of your ERP.

On a server contracted for you, hosted in Europe. You own the server: we set it up and maintain it, but the control and the data are yours. We don't use public clouds like Zapier or Make where your information travels through third-party servers.

No. When we rely on AI, we prioritise providers that don't use your information to train their models. And whenever possible, models that run on European servers under data-protection law.

Yes. By architecture, not by a policy written after the fact. The data never leaves the client's server. Only what is strictly necessary is sent to the AI, with no identifiers. Passwords and access keys are stored encrypted. Every action the system takes is logged for audit.

No password, API key or token is ever written directly into visible code. Everything is stored encrypted in the server's internal vault. On top of that, the system only gets the minimum permissions it needs to do its job: for example, "create a draft order" and "read stock", but not "delete invoices" or "modify master price lists".

No. The AI only reads the customer's emails and pulls out the important data (which product, what quantity, which customer). The comparison against your catalogue is done by the system, one line at a time, against your database. If there's no clear match, it stops and asks someone to review it.

It has a confidence threshold. If it's below 95% certainty, it doesn't act: it pauses the process and asks your team to validate. The AI never operates on your ERP on its own.

It stays in a queue and alerts your team with a clear message about what it couldn't process. Your admin person reviews it, handles it as always, and the system carries on with the rest.

No. The machine does the heavy transcription. Your team validates what doesn't fit and focuses on what really adds value: customers, cross-selling, decisions. The workload changes, the people don't.

Quite the opposite. During the build, part of the work is documenting the know-how your veteran staff hold: how they read the customer's language, which exceptions they handle, which equivalences they use. That knowledge is stored in the system. The day your lifelong admin person retires, that knowledge is still there.

No. The interface your team sees is very close to what they already use: notifications by email or chat like "this order needs your validation", and a couple of buttons to approve or adjust. No heavy training.

No. The server everything runs on is yours. The automations are documented. If one day you decide not to continue with us, everything keeps working and another technical person can understand and maintain what we've set up.

Because a system connected to several places isn't a painting you hang on the wall. Your customers change the format of their PDFs, AI models evolve, providers update their tools. If no one is watching, something breaks in silence. The monthly fee covers the monitoring, the small ongoing adjustments, and the real cost of keeping the system running (server + AI usage).

The system keeps working: the server is yours, the configuration is documented and open. Any technician with experience in this kind of tool can carry on the maintenance. We don't design lock-in. We mean it.