Artificial intelligence

AI applied to your business, with your data and under your control

Agents, automations and assistants connected to your CRM, your email and your documents, with EU models and human approval for what matters.

Request a proposal

Pix · escritorio

We start with a case that can be measured

We look at which tools your team already uses and with what data, choose a specific process and take it to production before expanding.

Consultancy and usage policy
A pilot with a clear objective
Training for the team

Agents that work with your tools

They sort email, prepare replies, update the CRM and ask your permission before anything that cannot be undone.

Customer service and sales
Integration with your CRM
Human approval of anything sensitive

With your data, without it leaving home

Models in the EU, permissions like those of your team and a log of what each agent does.

Models hosted in the EU
Access only to what you decide
Everything is recorded

Why AI in your business, and where to start

Most small businesses do not need “an AI”: they need hours to stop being lost on tasks that repeat and that call for reading, understanding and writing. Answering similar emails, turning invoices or delivery notes into a table, preparing quotes, summarising a meeting, finding a piece of data in twenty documents. That is where current AI is genuinely useful, because it works with language and documents, not only with numbers.

What usually fails is not the technology but the starting point: a tool is bought and then a use for it is sought. We do it the other way round: we choose a specific process, see what data it works with, take it to production and only then expand.

Repetitive tasks that a person does today by reading and writing: these are the best fit.
Data you already have in your CRM, your email and your documents, without having to copy it anywhere new.
A result you can check: time per task, tasks per day, errors corrected.
A person who reviews what matters: the AI prepares, your team decides.

The usual options, compared

When a business considers AI, it is almost always choosing between three paths (sometimes there is a fourth: the assistant that already comes with your office suite, on that provider's terms). None is bad in the abstract; what changes is the control over the data and what it can do for you.

	Each person with their own assistant (ChatGPT, Claude…)	Automation with fixed rules	AI agent connected to your tools
What it does well	Writing, summarising, explaining; good for starting to experiment	Repeating the same flow without errors, every time	Preparing work that changes each time and spans several tools
Where the data goes	Whatever each person pastes, on the terms of their plan	Wherever the automation tool is	Models on Azure, in the EU (the ForgeNEX Agent)
Knows your business	Only what you paste into each conversation	What was programmed	What you connect to it: CRM, email, calendar, documents
Control and logging	Each on their own account; the business sees nothing	What was defined when it was programmed	Permissions like those of your team, a log of what it does and approval of anything delicate
When it is enough	Occasional use and no customer data	A flow that never changes	When the work changes each time and touches several tools

Comparison based on ForgeNEX's own judgement, with no third-party figures. The data terms of each tool change depending on the plan and the provider: read those of the plan you use before pasting customer information.

Where your data goes

It is the question that weighs most in a small business, and rightly so. In the ForgeNEX AI Agent, the data is in the European Union: the servers are in Germany, and the models run on Azure, in EU data centres. What the agent does outside the application is logged (to whom and where, never the content) so that whoever administers the account can review it.

On what the Azure platform where the models run guarantees, Microsoft documents that your messages and the replies are not available to other customers or to OpenAI, and that they are not used to train its models. It also documents a nuance worth knowing: there is an automatic check against abusive use and, if an alert is triggered, a sample may reach human review; in the European Economic Area, those people are in the European Economic Area.

Source: Microsoft Learn, “Data, privacy, and security for Foundry Models sold by Azure” (updated 18-05-2026; consulted on 2-11-2026). Microsoft also states that, in EU “DataZone” deployments, processing may take place in any EU country. The location of the ForgeNEX servers and of the Agent is the one already given on our [AI agents](/agentes-ia) page.

What the AI does and what it does not

Setting clear limits is what makes an AI usable in a business. This is what we say today about our Agent, and we keep it up to date with the product:

It does	It asks you first	It does not do
Read your documents, emails and data with the permissions you decide	Send a message, publish something or any action that cannot be undone	Decide for you what is your responsibility: prices, contracts, dismissals, payments
Prepare quotes, email drafts, invoice tables and summaries	Delete, pay and any other action that cannot be undone: they wait for a person to approve them	Be infallible: it can make mistakes, which is why a person reviews anything important
Create tasks and reminders so that nothing slips	Anything that cannot be undone, even if asked in a hurry	Work outside the permissions you give it or skip the approval

How much it costs: work it out

We do not list prices for our services here: the AI Agent has its plans on the product page, and consultancy or integration work is quoted according to your case (you can ask us for a quote). What you can do now is estimate how much the task you want to automate weighs today:

Annual cost of the task = minutes each time × times per day × people who do it × working days per year ÷ 60 × cost per hour.

Assumption in the example	Value
Task	Preparing and logging a customer enquiry
Minutes each time	15
Times per day, per person	4
People who do it	2
Working days per year (assumption)	220
Cost per hour of one person (assumption)	18 €
Hours per year	15 × 4 × 2 × 220 ÷ 60 = 440 h
Annual cost of that task	440 h × 18 € = 7.920 €
If the AI saved half (assumption, not a promise)	3.960 € a year, plus the time spent reviewing

Example calculation by ForgeNEX with explicit assumptions; it is not data from any customer or a guaranteed result. Replace the numbers with your own: if the task does not turn out to be significant, it probably does not deserve an agent.

Case study 1: a firm that answers the same thing every day

Illustrative case, not a specific customer. A professional firm of 6 people receives similar enquiries by email every day, with attached documents, and notes them by hand in its CRM. Whoever handles them loses part of the morning reading, sorting and writing a first draft.

Case study 2: the team already uses AI on its own

Illustrative case, not a specific customer. In a business of 20 people, several already use a free AI tool to write and summarise. Nobody has decided what data can be pasted there, and management does not know what is being shared.

Source: European Commission, “AI literacy – Questions & Answers” (updated 20-10-2026; consulted on 2-11-2026). The Commission states that article 4 applies from 2-02-2025 and that its supervision regime applies from 2-08-2026; the text has been the subject of recent changes (“digital omnibus” package), so confirm the wording in force before taking decisions.

How we do it

We start with a conversation and a measurable case, not a generic demo: we see which tools your team already uses and with what data, choose the process, take it to production with a clear objective and train the people who will use it. The AI Agent is offered with Pix, and if you prefer to stay with Claude or ChatGPT, we can connect them to your CRM and your invoices with your permissions.

We do not offer a free trial of the Agent: we show it to you with your own cases and reply within 24 hours to arrange the demo. To learn more about the product and its scope, there is the AI agents page; to start with a diagnosis of what you have, the consultancy one. If you want to talk about yours, ask us for a quote.

Frequently asked questions

Is my data used to train models?+

In the ForgeNEX AI Agent, the models run on Azure and Microsoft documents that messages and replies are not used to train its models or made available to other customers or to OpenAI. If you use other AI tools, check the terms of their plan.

Where are the models hosted?+

On Azure, in data centres in the European Union. The product's servers are in Germany.

Can the AI send an email or change something without my knowing?+

For anything that cannot be undone, no: sending a message, publishing something, deleting or paying waits for a person to approve it. And what it does outside the application is logged (to whom and where, never the content) so that whoever administers the account can review it.

Which process is best to automate first?+

One that repeats daily, that calls for reading and writing and that has a result you can measure (time per task, tasks per day). If you cannot manage to measure it, it is better to choose another.

Can I use my Claude or my ChatGPT with my ForgeNEX data?+

Yes, by connecting them from your account: they read, create and change with your permissions (you choose whether they only read), but they do not delete or export. What you ask and what it reads goes through that provider, on its terms.

How much does it cost?+

The Agent has public plans on the product page and consultancy or integration work is quoted according to your case. On this page you also have a formula to estimate how much the task you want to automate costs today.

Is there a free trial of the Agent?+

We do not offer a free trial: we show it to you in a demo with your own cases and reply within 24 hours.