Automating an immigration firm:
what actually works.

The specific problems.

Before talking about solutions, it is worth being precise about what the problems are. They are not the same for every firm.

• Documents in multiple languages. Clients submit evidence in their native language. Somebody has to translate it before it can be reviewed. This is slow, expensive, and if done by machine and not checked, unreliable.
• Evidence that is not text. A client photographs their passport and sends 30 images. Each image is one page. A human opens each image and types out the relevant information. There is no shortcut in a standard legal system.
• Decisions from authorities that arrive by post. A determination from a migration authority arrives on paper, in Swedish, on a Tuesday afternoon. Someone reads it, translates the relevant parts, and summarises the outcome for the client. That is an afternoon of work per case.
• Legal guidance that changes. Immigration law is not stable. What was grounds for a positive decision eighteen months ago may no longer be. A firm's internal guidance has to be updated regularly, and every case open at the time of a change has to be reviewed against the new reality.
• No standard tool handles this. General practice management software is built for typical legal workflows. Immigration is not a typical legal workflow. The firms that try to fit it into a generic system spend as much time managing the software's limitations as the cases.

What a custom system actually does.

The system we built for a three-person immigration firm starts the moment a new case folder is created. The case handler drops the client's evidence into a designated folder on the firm's local machine. That triggers the process.

The system reads every document in the folder. For images, it uses optical character recognition to extract the text. For handwritten documents, it applies a more specialised reading pass. For any document that is not in the firm's working language, it translates. This step, which a human would spend two to three hours on, takes about four minutes.

Once the documents are readable, the system does what the senior lawyer would do on a first pass: it extracts every date, every name, every jurisdiction reference. It checks whether the dates in the client's written statement match the dates in the documents. It notes the ones that do not. It produces a structured brief — in the format the firm uses — that the lawyer can open and work from immediately.

Sara · Lead lawyer, three-person firm — her estimate before we built the system was five to seven hours per case for the initial evidence review. After: about forty-five minutes, mostly spent reading the brief and deciding the legal strategy. The cases she handles have not changed. The part that required human attention has been reduced to the part that actually requires a lawyer.

The legal guidance problem.

Immigration law in most countries changes at a pace that is hard to keep up with. Sweden, the UK, the Netherlands — all have had significant policy shifts in the last two years that affected open cases.

A system that was trained on last year's legal framework is a liability, not an asset. The way we handle this is to keep the legal guidance layer separate from the rest of the system. When guidance changes, the firm updates a specific document that the system draws from. It does not require a rebuild. It requires an update to one file, in plain language, that describes what changed and from what date.

The system can then flag any open case where the new guidance is relevant, rather than requiring the firm to review every open case manually.

What this does not replace.

The system handles the first pass. It does not handle the case. The legal strategy, the relationship with the client, the judgment about how to frame an argument, the decision about which evidence to lead with — none of that is automated. All of that still belongs to the lawyer.

The intention is not to replace professional judgment. It is to make sure that professional judgment is not being spent on reading the same document three times in three languages because there is no other way to do it.

Red flags to watch for.

• Any tool that sends client evidence to a cloud service.

  Immigration evidence is protected under GDPR and in many cases under specific asylum data protections. A tool that processes it through an external API is a compliance problem, whether the vendor mentions it or not. Ask specifically where the data is processed.

• Generic legal software claiming to “support immigration.”

  Supporting immigration usually means having a case type called “immigration.” It rarely means handling multilingual photo evidence with automatic translation and date extraction. Ask for a demonstration with a real case file before you commit.

• A system that cannot be updated when guidance changes.

  If updating the legal guidance the system draws from requires calling the builder and paying for a rebuild, the system is not built for a law firm. Guidance changes. The update process should be something the firm can do itself.