Insights

The AI that adds most value to a company does not always write: often it predicts. What it can forecast from the data your business software already holds, how much you need and how to start.

On the personal versions of ChatGPT, Gemini, Copilot and Claude, what you type can be used to train their models. On business plans and the API, not by default. We checked it plan by plan.

Dictate the site visit report as you leave the site and get it organised, ready to review and sign: how it works with AI, what it does not replace and how the audio is protected.

A practice's archive of past projects holds many of the answers each new commission needs. How to turn it into an AI knowledge base, step by step and without mixing up projects.

The AI built into your ERP or CRM is enough when the task starts and ends inside that software. When custom development pays off, and five questions to decide.

Half of employees already use AI tools without approval, and more and more of them give those tools access to their email and documents. How to find out which agents are connected to your company and how to put limits on them.

Before connecting an AI assistant or agent to your business software, ask how much its vendor charges for access. Some already charge differently depending on whether the caller is an integration or an AI.

Most law firms do not know what each case file costs them, so they do not know what AI saves them. How to calculate it, where the cost really falls, and what to demand when medical data is involved.

An AI agent is not dangerous because of what it knows, but because of what you let it do. How to allocate its permissions, which actions must always go through a person, and what to log so you know what it has done.

There is a European registry of AI systems, but almost no company has to enter it. What you do need is something else, and that you will be asked for: knowing what AI your company uses and at what risk level.

Almost nobody trains a model on your documents, and that is good news. What is actually done is something else, works better, and has its own traps: the most expensive one is permissions.

A customer ordering by WhatsApp or dictating an audio note, with the order landing straight in the ERP, is now viable. The hard part is not understanding the audio: it is getting the price, the stock and the product code right.

In 2026 every expense tool claims to have artificial intelligence. The real difference is whether the system judges the expense or merely transcribes it. How to test it in a demo, with your own receipts.

The AI Act is already enacted law, with dates confirmed through 2028, not a proposal waiting for approval. This is the guide without an expiry date: the concrete steps to classify your AI systems, comply with what's already mandatory today, and not fall behind on what comes next.

OCR, approval workflows, ERP integration, tax certification... almost every expense management tool advertises the same things on paper. This is the list of real criteria to tell them apart and the questions to ask before signing, without relying on what each vendor says about itself.

From 5 October 2026, the administrative control document for road transport in Spain must be digital: a PDF with a QR code, not a paper copy or just any PDF. Here's what the rule requires, who's liable if a field is missing, and how to avoid generating it by hand on every trip.

"Human oversight" is one of the most repeated phrases in any AI proposal, and one of the least explained. In practice it is a concrete design, with at least four distinct decisions and a legal obligation behind it.

Most legal AI on the market is built to search public case law or draft a first version of a document, not to touch a client's full file with medical or judicial data. That's where an obligation kicks in that isn't in GDPR: the lawyer's duty of professional secrecy.

“We already built an app so that customers could order directly, and nobody used it. Why would it be different this time?” It is the best question a company can ask, and the answer is not “because it has AI”. Traditional applications ask the customer to adapt to them; a person does not. What really changes, and what it does not fix.

There is more supply than ever and almost everyone claims to be GDPR compliant. Ten criteria to separate those who will solve a process from those who will sell you a demo, the clauses worth having signed before giving access to your data, and twelve questions for the first meeting.

The Bank of Spain puts the productivity gap between companies that have adopted AI and those that have not at €6,300 per worker, twice what it was three years ago. We walk through that figure and the 30% of tasks AI already assists with according to SAP and Oxford Economics, caveats included, and what really separates one group from the other.

A manual expense report takes an average of 20 minutes and fails one in five times, according to GBTA. We explain what Captio is, how it automates expense capture, approval and accounting, and walk through the real calculation of how many hours your company saves each month.

A Deloitte study of more than 660 technology leaders puts between 21% and 40% of IT spend into propping up outdated systems instead of creating new value. We walk through the signs your software has already become the problem, why AI changes the calculus of modernizing it, and what to do about it.

In July we saw that much of the real AI use inside Spanish companies happens without leadership knowing. A new digital hygiene survey puts a number on the problem from a different angle: 64% of employees use unauthorized AI, and fewer than a third of companies know precisely what applications run on their network. It is a control gap.

One survey says AI investment is surging in Spain. Another, based on real activity data from 68,000 companies, finds barely 2.9% adoption. We walk through the figures, their sources, and what neither survey captures: how much AI is already being used without the company knowing.

The EU has published and enacted the «Digital Omnibus» (Regulation EU 2026/1744): it's now law, not a proposal. We clarify what actually starts this August, what gains time, and what a company should do today.

From August 2026, using AI in your company has rules. But leading companies don't read this thinking about the fine: they see it as the signal to bring AI into their processes and stay the benchmark in their sector.

The companies investing most in AI are not always the ones getting the best results. What makes the difference is the context you give the model and the experience of whoever decides how to build it. An honest reflection on what really transforms a company with AI.

On 2 August 2026 the full obligations of the AI Act for high-risk systems come into force. Many mid-sized European companies have affected systems without knowing it. Practical guide to classify, prepare and not arrive late.

Anthropic has decided not to release its most powerful model. The press talks of terror. We see something different: a responsible decision, and an opportunity to understand why many companies are still failing to use the AI they already have available today.

Retrieval Augmented Generation (RAG) lets companies query their internal documents in natural language, reduces hallucinations and improves decision-making.

AI agents are moving from experiments to production tools. What they already do in finance, HR, operations and customer service.

Dashboards that nobody looks at are a waste. Learn how to design Power BI visualizations that actually drive action in your organization.

After more than 1000 integration projects, we've seen the same mistakes over and over. We tell you what they are and how to make sure your project isn't another one.