Use of artificial intelligence

How falcani uses AI in "Let’s solve it together": what it receives, what it can and cannot say, how long it is kept and what happens if it fails.

In 30 seconds

  • What it is for

    Only in "Let’s solve it together": you write a phrase about your business and the AI turns it into a visual proposal with sample screens. It is not used to answer your messages or to decide anything about you.

  • What it receives

    The phrase you write and your language. It receives no name or email, because that step does not ask for them.

  • What it cannot say

    No prices, no deadlines, no percentages, no result figures, no third-party brands, no promises about who owns the code. A proposal is an idea, not an offer.

  • If it fails

    If the AI does not answer in time or the monthly cap is reached, the site shows you a proposal from its own library and says so.

What the AI cannot write

Pick a rule and see a line that would be blocked and the one shown in its place.

A rule

Blocked

Everything included for a fixed price every month.

Shown

A proposal carries no price: we talk it over with you.

This is a hand-written example to explain the rules. It is not an answer from the AI.

Full text

1. What "Let’s solve it together" is

"Let’s solve it together" is a feature of the site where you describe your business in one phrase and get a visual proposal of how custom software made for you would look. A proposal is an illustration of an idea. It is not a quote, an offer or a commitment by falcani.

2. Who processes your phrase

The proposal is generated by an Anthropic artificial intelligence model, to which the site sends the phrase you wrote and your language. The site does not send it your name, your email or any contact data. Before you write, do not include personal or confidential data in the phrase.

3. What it can and cannot say

The model does not answer with free text: it returns a proposal with a fixed structure (the type of business, the "before", a device, the problem, the solution and some sample screens) that the site draws with its own components. The rules it is given, and that the site checks again, are:

  • No prices or deadlines.
  • No percentages or result figures.
  • No third-party brand names.
  • No promises about who owns the code.
  • If the phrase does not describe a business problem, it asks you to clarify instead of inventing a proposal.

Every output also goes through a list of forbidden words and patterns before it is shown. Even so, a model can be wrong: if you see something that does not belong, tell us.

4. What is kept and for how long

The proposal, with the phrase you wrote inside it, is kept for 30 days so you can see it again and share its link. It carries no name or email. After that it deletes itself. If you decide to write to us from the proposal, the identifier of the proposal travels with your contact details, and the server reads from the saved proposal a summary that includes your phrase and adds it to the note of your contact in Clientary (or, if Clientary fails, to the internal email notice). That note does not expire after 30 days: it stays in Clientary for as long as Clientary keeps it, like the rest of your contact; the site keeps no copy.

The artificial intelligence provider processes the phrase only to generate the answer, under the commercial terms of its API service. A monthly spending cap limits how many proposals are generated.

5. If something fails

If the model takes more than 12 seconds, returns something that does not meet the rules or the monthly cap was reached, the site shows you a proposal from its own library, which does not use AI, and marks the origin of the proposal. It never shows you unchecked text.

6. Automated decisions

The AI on this site makes no decisions about you and does not rate you. All it produces is an illustrative proposal that you can ignore.

7. Your rights

You can ask for a proposal to be deleted before the 30 days by writing to us from Your data, with the link of the proposal.

This text is a draft and describes how the feature is designed. Before it is published as final we still need to confirm: that the contract with Anthropic excludes using the phrases to train models and sets a data processing agreement, the model and the monthly cap that will be used, and that the feature in production meets every point of this text.