People are already using it
Somebody pastes the customer list into the first tool they find. That data never comes back.
Artificial intelligence in your business, and the judgement to know when not to
We integrate language models on infrastructure you control and train the people who will use them. When an ordinary rule is enough, we say so.
What you get, by name
Concrete things you can take away.
- The inventory of which AI tools are used and with what data.
- The models integrated and running on infrastructure you control.
- Training built on real cases from the work, not textbook ones.
- A written list of what these tools can and cannot do.
- The judgement of when AI is worth it and when an automation is.
How AI arrives at a small business
It almost always comes by one of three routes, none of them the one the news describes.
Licences were bought and nobody uses them
They were bought because something had to be done about AI. Nobody knows what for.
A task that looks like AI
Somebody proposes a model for what is four rules and a query. It would cost more and fail worse.
All three are settled the same way: where your data is, and what the technology does well and badly.
Is this what you need?
If your case is not quite this one:
- If the process follows clear rules on predictable data, it is a more reliable automation: Automations
- If an application or a feature that does not exist is missing, that is software development: Software Development
How AI work runs here
Four steps, and the second sometimes ends in “do not use AI here”.
See what is used already
Which AI tools are paid for, who uses them and with what data.
Sort the cases
Ambiguous information to interpret, or rules? If rules, it gets automated.
Build the models
Language models on controlled infrastructure, integrated with your systems: the data does not have to leave the company.
Train the people
For teams and staff, in person or remotely, with Copilot, Claude and the company’s own tools. Use cases, good practice and limits.
Who this is for, and who it is not for
A good fit
- Businesses whose staff already use AI with no written rules.
- Anyone with data that cannot leave the company who wants AI.
- Anyone who wants to know, before spending, whether their case is AI.
Not a fit
- Anyone after predictive models trained on their own data: a different discipline.
- Anyone who has decided they want AI and only wants it built.
Frequently asked questions
Are models trained on my data?
No. Predictive and machine-learning models are not trained here. We apply language models that exist to specific problems.
Does my data leave the company?
It does not have to: the models never go out to a third-party service. What stays inside is decided up front.
What if my case does not need AI?
We say so. On structured data, a query or an integration costs less and runs faster than a model, and is easier to debug.
Is this a project or a monthly arrangement?
Normally it is a monthly fee. The initial implementation can be a closed project.
Tell us about your situation
Tell us what you want to do with AI, and with what data. The first call costs nothing.
Tell us what’s broken+34 600 237 257info@isitechconsulting.com