What an AI agent already does in your sector

In a services company, an AI agent today does mainly three things: it reads documents and pulls out the data, drafts the first version and answers first-level enquiries. A person reviews what matters and signs. What changes from one sector to another are the documents, the rules and the criteria used to review.

The short answer: an AI agent does not decide on its own and does not replace the professional. It prepares the repetitive work on documents (reading, checking, drafting, alerting) and leaves the result ready for a person to review. Which documents, which rules and what kind of review depends on your sector, and this guide goes through fourteen of them.

What we mean here by an AI agent

An application with AI connected to your documents and your software, which carries out a multi-step task and leaves the result ready for a person to review. It is closer to an assistant who prepares the work than to an employee who decides alone. The limits it works within (what it can read, what it can write and when it must stop and ask) are explained in what permissions to give an AI agent in your company.

The three jobs that repeat in every sector

1. Read and extract. Invoices, delivery notes, reports, case files, policies: the agent reads documents that arrive in different formats and puts the data into the system, with a citation of where each piece comes from.
2. Draft the first version. Briefs, reports, offers, summaries, communications: it starts from your data and earlier documents, in your company's tone, and leaves the draft for review.
3. Handle the first level. Routine enquiries from customers, patients, students or prospects, answered from your own knowledge base. It passes to a person whatever it does not know and whatever it should not decide.

What it does in each sector

What follows comes from the use cases we work on with each sector, grouped by activity. The sector pages, with more detail, are linked from each name (they are in Spanish).

Professional services

• Law firms. Summarises case files and reports with a citation of the source, drafts briefs and proposals from the file and filters new enquiries. With medical or court data it always works on pseudonymised data, with the lawyer supervising each document. In one of our projects, analysing legal and medical documentation went from two or three hours to minutes per document (case, in Spanish).
• Tax and accounting advisory firms. Reads invoices and receipts to feed your systems, drafts reports and communications to the client and answers frequent questions from your tax and employment knowledge base.
• Notary offices. Prepares draft deeds from the transaction data and reviews the documents supplied by the parties to flag what is missing or does not fit. The notary reviews the result.
• Consultancies. Drafts first versions of reports and proposals with your methodology and previous cases, and searches accumulated knowledge so work is not redone.

Technical: engineering, architecture and construction

• Engineering firms. Drafts technical reports and specifications, prepares offers and tender documents from earlier projects and finds the applicable regulations.
• Architecture firms. Drafts project descriptions and the justification of planning regulations, and handles documentation with public bodies. Specific examples in the archive of approved projects.
• Construction and development companies. Prepares offers and tenders, drafts reports and documentation and looks up regulations. We have a daily financial control of each site under development, and site visit reports by voice are another example.

Health, insurance and real estate

• Clinics and private healthcare. Drafts report drafts for the professional to review, in a secure environment and in line with the GDPR, and handles administrative enquiries and reminders. In a hospital medical-equipment maintenance project, an application warns of servicing according to the actual use of each machine, with 75% fewer incidents (case, in Spanish).
• Insurance brokerages. Summarises policies and cover to advise faster, organises the documentation of each claim and prepares renewal communications.
• Real estate agencies. Answers prospects instantly and filters the serious ones, drafts property descriptions and prepares draft documents.

Industry, logistics and retail

• Industrial companies. Prepares technical offers from earlier cases and drafts or reviews procedures, data sheets and manuals with your knowledge.
• Transport and logistics. Reads and classifies delivery notes and consignment notes and prepares the documentation of each incident. With the obligation of the digital control document, that reading carries more weight.
• Retail and distribution. Creates product sheets and descriptions for the whole catalogue, handles enquiries and supports order processing. One example is taking orders by voice and chat connected to the ERP.

Training and marketing

• Training centres. Prepares materials and exercises from your programme, answers students and families and supports marking, always with teacher review. In a continuous-training project with AI for companies, the time spent on training fell 45% (case, in Spanish).
• Marketing and communication agencies. Produces content drafts in each client's tone, prepares proposals and presentations and writes campaign reports from the data.

What it does not do in any sector

• It does not decide alone anything with legal or financial consequences. It prepares it and a person approves it. There are levels of human oversight, and the level is best chosen according to what a mistake costs.
• It does not work on data that does not exist. If tariffs or criteria only live in someone's head, putting them in order is the first job and it is yours.
• It does not replace professional judgement. The lawyer, the notary, the doctor or the architect still sign what they sign.
• It does not start without measuring. Before scaling it, check whether AI really saves you money across the whole process and not only on one task.

Four questions to choose where to start

1. Which document does your team handle every day, and how many a month?
2. What does a person do after reading it?
3. What happens if the first draft is wrong, and who would notice?
4. Is the data needed already in a system, or in someone's head?

On each sector page we mark the activity with the most potential as "Empezar aquí" ("start here"). The best candidate is usually the one that repeats most, has a clear start and end and works on documents. If the doubt is whether the AI that comes with your software is enough, the answer is in is the AI built into my software enough, or do I need custom development?.

What it costs to start

The proof of concept starts at €850 and takes two to four weeks with real documents from your company. Full development runs from €5,000 to €20,000 excluding VAT, with a fixed price per phase, and the monthly fee starts at €250 with the infrastructure, the AI account and maintenance included. What each amount includes (in Spanish).

Frequently asked questions

Does an AI agent replace someone on my team?

In the cases in this article, no. The agent prepares the work and a person reviews and signs it. That person's time shifts from typing and searching to reviewing and deciding.

What about my clients' confidential data?

It depends on how it is built. At Tecnea, data is processed in your environment or in the European Union and is not used to train any model. With specially protected data, such as medical or court data, it is pseudonymised first. It is explained in private and secure AI (in Spanish).

My sector is not on the list. Does it still make sense?

Almost any work on documents fits one of the three jobs above: reading and extracting, drafting or handling the first level. Tell us which documents you handle and we will see which of the three pays off.

How long until the result shows?

A proof of concept lasts two to four weeks, and measuring the saving takes another four to six. The method is in how to tell if AI is really saving you money.

If you want to see what it would look like in your sector, tell us which documents you handle.

This article is informative. Tecnea develops custom applications with AI, so we have an interest in you considering this route: that is why we also describe what an agent does not do and when it is worth measuring before investing. The result figures are those measured in our own anonymised projects and are explained in each case.

Sources

1. Tecnea, sector pages: AI for law firms, engineering firms, clinics and the other sectors (in Spanish), with the use cases by activity.
2. Tecnea, case studies (in Spanish): projects in production and in development, with the figures measured in each.