← Solutions

        Sector
        
        Technical services and homologation labs

AI for vehicle homologation labs: complete files before anyone signs

In a vehicle modification lab, the time does not go on engineering. It goes on answering the same questions, chasing missing documents and reviewing files that arrive incomplete. Odysi builds three pieces for that work: WhatsApp intake that collects every request complete, a check of the paperwork before the engineer sees it, and one request to the client for everything that is missing. The engineer still decides and signs.

Try it with your customers' messages →

For Technical services, engineering firms, workshops

Where it runs Your WhatsApp and your email

Who signs Your engineer, always

Proof Fixlab, in production

Where the time really goes

A modification lab runs on files. In Spain, Royal Decree 866/2010 lists the documents a vehicle modification can require: the technical project and the end-of-works certificate, signed by a qualified engineer; the conformity report from the designated technical service; and the certificate from the workshop that did the work. What each modification needs is set by the Ministry of Industry's Vehicle Modifications Manual, whose latest revision was approved on 27 January 2026.

The problem is volume, not engineering. Your engineers keep explaining what a tow bar needs, ask for the same photo again, and review files that are one document short. Every round trip is another day for the client and an hour less of engineering.

Fig. 1 · Triage Incoming questions

front door

Answered from the knowledge base

Change the headlights Add a tow bar Standard documentation, filed as a request

Escalated to a person

Van into a camper Viability from a photo

Fig. 1: the high-volume, pattern-following majority is handled in the conversation. The cases that need judgment reach a person.

WhatsApp intake: every request arrives complete

First contact is usually a WhatsApp message with a photo and a question. A WhatsApp intake agent answers from your lab's own criteria, not generic rules, and gathers what the engineer needs:

The modification the client wants and, if your lab works that way, its code in the Modifications Manual.
The vehicle data as it appears on the ITV card.
The photos and documents you ask for on that modification, one by one, until all of them are in. It reads photos and transcribes voice notes.
Anything that does not follow a pattern, flagged for the engineer with the full conversation.

When the client has everything, the agent opens the request in your platform with the data and documents attached. It runs on your WhatsApp Business number; nothing is installed on your website.

The parts that matter A structured knowledge base as the source of truth A conversational front door on the customer's channel Intake that files a complete reform request Clear rules for when to hand off to a person

Fig. 2: the knowledge base is what makes the answers reliable. The model choice matters less than the source of truth behind it.

A check of each file before the engineer signs

Once the file is in your platform, the AI compares it with your checklist for that type of modification: every document present, the vehicle data matching across the project, the workshop certificate and the ITV card, and the photos showing what they need to show.

The output is a review sheet for the engineer: what is fine, what is missing and what does not add up, with the document where it found it. From that sheet goes one request to the client or the engineering firm, with everything missing at once, instead of three emails over three days. How it works in detail: AI document review.

The AI helps the engineer assess the file. Validating the modification stays their job.

If your clients are engineering firms and workshops, not the public

Many labs hardly speak to private owners: their clients are engineering firms that write modification projects and workshops that carry them out. The questions change, but they repeat just the same: where a file stands, what is missing to close it, when the report will be ready, which criterion the lab applies to a case it has settled before.

The agent answers those from your platform and your criteria, at any hour, and passes the engineer anything that needs a fresh opinion.

Email across several inboxes

Part of the work arrives by email, spread over several inboxes. The AI can read each email, tell a new request from a question about an open file, attach it to the right file and draft the reply. A person approves every reply before it goes out. It is one part of AI process automation.

06 · In production

What we built for Fixlab

Fixlab is a vehicle certification lab in Spain. Its support was split across WhatsApp, email, phone and its platform. We built Fixi, a WhatsApp assistant that answers from Fixlab's knowledge base, opens the modification request in its platform through the platform's API, and hands the case to the team's Slack, with the full conversation, when judgment is needed. Before launch we agreed in writing on a test of 50 real scenarios with a threshold of zero critical failures. It has been live since June 2026.

Field metrics · 26 Jun to 24 Sep 2026

279

conversations from 218 people

16 s

to the first reply (median); every conversation answered within a minute

55%

start outside office hours (Monday to Friday, 9:00 to 15:30)

51%

closed without a person

reform requests opened end to end in its platform

critical failures in the 50-scenario test

Volume more than doubled in September, and reply times did not change. Fixi handles intake; the file check and the email work described above are the next step for a lab.

Read the Fixlab case study

What AI can do here, and what it should not

It can answer the documentation questions that follow a pattern, gather the file, compare it with your checklist and prepare the request. It does not approve modifications, sign, issue conformity reports or make technical judgments. Deciding whether a modification is feasible from a photo, or assessing a camper conversion with a gas installation, is the engineer's work. What matters is that the system tells one case from the other and passes it on instead of guessing.

The AI prepares the file. The engineer decides and signs.

Confidentiality: your rules do not go to another lab

If Fixlab is your competitor, it is a fair question. What we build for one lab does not go to another. The system is built on your criteria and your documents, which are yours. What we reuse across projects is the underlying platform (the messaging layer, handover to a person, transcripts), not anyone's rules, answers or data.

How to start without taking on too much risk

Start with one type of modification: the one that brings the most questions and the most back and forth. We write the checklist with you, test the system on files you have already closed before it touches a new one, and during the pilot a person reviews every reply and every review sheet. If it does not save work on your own files, you will know before you extend it.

FAQ

FAQ: AI in a vehicle homologation lab

Can AI approve a vehicle modification?

No. Approval, signature and the conformity report stay with the engineer. The AI handles intake and preparation: answering documentation questions, gathering the file, comparing it with your checklist and preparing the request.

What can AI take on in a homologation lab?

What repeats: which documents each modification needs, which photos, where a file stands, what is missing to close it, and a first check of the file against your checklist. Anything that needs technical judgment goes to a person.

Our clients are engineering firms, not the public. Does it fit?

Yes. An engineering firm or a workshop asks repeated questions too: file status, missing documents, deadlines. The agent answers those from your platform and your criteria, and passes the engineer whatever needs a fresh opinion.

Are our rules shared with other labs? Is it built on Fixlab's flow?

No. It is built on your criteria and your documents, which are yours. What we reuse across projects is the underlying platform (the messaging layer, handover to a person, transcripts), not another client's rules or data.

Does it connect to our file management platform?

It depends on the platform. The first thing we look at is what it allows: an API, database access or exports. At Fixlab, the assistant opens modification requests directly in its platform through the API. If it cannot connect, the file reaches you in order by email or the channel you already use.

What results has Fixlab had?

From 26 June to 24 September 2026: 279 conversations from 218 people, a first reply in 16 seconds (median), 55% of conversations outside office hours, 51% closed without a person, and 20 reform requests opened end to end in its platform. The details are in the Fixlab case study.

Can we change the rules ourselves after delivery?

Yes. We write the documentation for your system and it is fully yours, so your team can update the answers and the rules without depending on us.

How much does it cost?

It depends on what it does (WhatsApp intake only, or the file check and email too) and on whether it has to connect to your platform. After a short call we send a written proposal with the price and the assumptions. We explain what moves the price in what decides the cost of a WhatsApp AI assistant.

Sources, checked 30 Sep 2026 (in Spanish): Royal Decree 866/2010 on processing vehicle modifications (articles 7 and 8) and Resolution of 27 January 2026 approving the revision of the Vehicle Modifications Manual (Official State Gazette, 5 February 2026).