AI architect

Ask an AI about Staffbite's services and EU-first. It can draft a solution sketch and an indicative EU readiness assessment, and the processing is in the EU.

The AI architect is an AI assistant that answers questions about Staffbite's services based on this site's own content. Tell it what you are building or looking into, and it tells you which of our services fits and what a sensible next step would be.

The architect is an AI, not a person, and it can be wrong. Conversations are stored, and Staffbite can read them (Privacy), so do not write sensitive information in them.

Updated

Ask the AI architect

  1. Notice: I am an AI assistant, not a person. My answers come from Google's Gemini model on Vertex AI in the EU multi-region, so the model runs in EU member states. Google is a US company. I can make mistakes. Conversations are stored, and Staffbite can read them. Do not share sensitive information here. How conversations are handled

  2. AI architect: AI

    Ask about the AI Mindset workshops: what a leadership team gets from one and how the 90-day plan is built.

Example questions

At most 2,000 characters. Do not write sensitive information.

Leave your contact details

Nothing is sent until you send the card. Only the email address is required. The conversation is attached so that Staffbite can see what it is about.

How would you like to continue? (optional)

Recipient Karo Tammela, CEO

The card and this conversation are stored in Staffbite's database in UpCloud's Helsinki data centre and kept for 24 months. Staffbite reads them and gets an email notification of the card. After you send the card, Staffbite may make an internal summary of the card and the conversation with the Gemini model in Vertex AI's EU multi-region. The summary contains no price, and only Staffbite reads it. You will not get a confirmation email. Read the privacy policy

What the architect does and does not do

The architect:

  • answers questions about AI-native development, AI visibility, AI Mindset workshops, products and EU-first, and links to the page where the topic is covered
  • suggests one concrete next step
  • suggests 1–3 pages or posts on this site when they answer your next question better than the conversation
  • drafts a solution sketch once you have described your situation and goal
  • makes an indicative EU readiness assessment of your AI use case
  • offers the contact card when you ask about price, schedule or a person, or when the matter needs a person

The architect does not:

  • give prices, because they depend on the scope, and the solution sketch has no price either
  • give legal advice, for example on interpreting the AI Act or the GDPR, and the EU readiness assessment is not legal advice either
  • make decisions, offers or commitments on Staffbite's behalf
  • search the internet
  • see your contact details or send anything itself, including the solution sketch

The architect answers in the language of your latest message. On the English pages it starts in English, and if you write in Finnish, it answers in Finnish.

The architect knows only the site's public content. In an English conversation, its knowledge base is the English part of the llms-full.txt file (Machine-readable interface) meant for machines (open the file): the texts of the English pages and published posts. It contains no internal material and no unpublished content drafts.

Content suggestions appear as cards below the answer. Each one has the title of the page or post, a link and a short reason. The suggestions link only to pages and published posts on this site: the server takes the title and the address from the site's content, and the model cannot give an address of its own.

The aim is that you get a useful answer and a next step in a few messages. A longer conversation also works. However, the conversation closes once you have sent 50 messages, when its usage limit is reached or when it is 30 days old. The architect may sometimes take a break if the site's daily usage limit is reached or the service is busy. The contact card and the contact page still work then.

Solution sketch

Once you have described your situation and what you are aiming for, the architect can draft a solution sketch. It may first ask at most two clarifying questions. You can also ask for a sketch yourself.

The sketch covers the situation, goal, approach, steps, an estimate of the scope and what drives it, data and the EU perspective, risks, stop criteria, assumptions, the first step and the best-fitting service. The scope is one of the scopes on the AI-native development page, and its duration is shown as a range: small 1–2 weeks, medium 2–4 weeks, or a larger whole built batch by batch. The sketch has no price, and the architect does not promise a duration for your specific case.

The sketch is produced by AI from the conversation. It is not an offer or a commitment, and it may contain errors. One conversation can have at most three sketches.

Share link and printing

Each sketch has its own page, which you can open with the card's link “Open the shareable page”. The “Copy link” button copies the page's address to the clipboard, so that you can send it to a colleague, for example. The button only appears when JavaScript is enabled. You can print the page or save it as a PDF with your browser's print function.

Anyone who has the link can open the sketch. The link contains a random identifier that cannot be inferred from other sketches. The page is marked noindex for search engines (Privacy), and it is not in the sitemap or the llms.txt files. The sketch is based on your conversation, so share the link only with people you want to see it.

The sketch is not sent to you by email. Save the link or print the page if you want to come back to the sketch later.

The card's “Ask Staffbite to comment” opens a contact card with the sketch's name prefilled in the message. When you send the card, Staffbite's notification includes links to the conversation's sketches. The card is sent only when you send it yourself.

EU readiness assessment

When you tell the architect what you plan to use AI for, it can make an EU readiness assessment of your use case. The assessment appears as a card below the answer, and it is based on what you have said and on the site's public content.

The card shows:

  • the use case in brief
  • what kind of data it involves, for example personal data, special categories of personal data or confidential business information
  • where the data is processed: in the EU, outside the EU or both
  • the providers you mentioned and their home countries
  • a preliminary assessment of what the AI Act might require of the use case, with the reasoning
  • 1–4 things worth checking next

If you have not said something, the card says that the information is not known or that the assessment is unclear. The architect has been instructed not to invent providers.

The card always carries the fixed text “Indicative assessment, not legal advice.” The assessment is produced by AI, and it can be wrong. The card's links lead to the EU-first page sections What the law says now, Dependency mapping and Location does not decide jurisdiction.

If, according to the assessment, the use case may be high-risk or prohibited under the AI Act, the card recommends that an expert at Staffbite reviews it and links to the contact card below the conversation. For example, systems used in recruitment may fall under the high-risk systems of Annex III of the AI Act, whose obligations start on 2 December 2027 (What the law says now).

The assessment is stored with the conversation, like the other cards.

Model and data location

The answers are produced by Google's Gemini model in the Vertex AI service. We use Vertex AI's EU multi-region, so the model's processing takes place in EU member states (AI processing). The EU multi-region does not cover the United Kingdom or Switzerland (Google Cloud: Vertex AI data residency).

Google is a US company, so the US CLOUD Act applies to it regardless of where the data is. The EU multi-region therefore settles where the processing happens, not the jurisdiction. We explain this in more detail on the EU-first page.

Nor does the EU multi-region guarantee that Google retains nothing. Google does not use customer data to train models without permission. However, Vertex AI's caching is on by default: the in-memory cache is cleared within 24 hours at the latest, and the persistent cache is also on by default. Google may also log prompts to detect abuse (Google Cloud: Vertex AI data governance). We do not use grounding with Google Search, because it stores search logs for 3 days and cannot be turned off.

Conversations are stored in Staffbite's database in UpCloud's Helsinki data centre (Location and data). The model can be changed through configuration, as we explain in our EU-first commitments.

The architect has evals: imagined conversations that check, for example, that the answers contain no prices and that the links lead to pages on this site. The evals use only imagined conversations, never visitors' conversations.

What is stored from a conversation

Conversations are always stored, and Staffbite can read them. We read them to improve the architect's answers and the site's content, and to investigate misuse.

We store:

  • your messages and the architect's answers
  • the page of this site from which you opened the architect (for example AI visibility)
  • technical details of each answer: the model name, token counts, cost and response time
  • the cards shown with the answer, including their content: contact card, content suggestions, solution sketch or EU readiness assessment
  • the solution sketches

If you write an email address or a phone number in a message, the server replaces it with the marker [sähköposti] or [puhelin] (Finnish for email and phone) before the message is stored or sent to the model. The email address is prefilled in the contact card at the same time, but it is stored only if you send the card. Other information, such as names, is not masked.

The conversation does not store your IP address, your browser's user agent or the external page you came from (Privacy). The server uses your IP address to limit the number of messages, but the address is kept only in the server's memory.

A conversation from which you do not send a contact card is deleted automatically 6 months after the last message (Privacy). If you send a card, the conversation is attached to your contact request and kept with it for 24 months. Solution sketches are kept with the conversation for the same time. The privacy policy has the details.

Contact card

With the contact card you send your question and contact details to Staffbite. Only the email address is required. Name, company, phone number and message are optional, and you can say whether you would like a reply by email, a call or an online meeting.

The architect can prefill the message field with a summary of the conversation, and you can edit it before sending. The card is sent only when you send it yourself. The card is always available below the conversation under Leave your contact details, also when the architect is taking a break.

The card the architect offers may also include a section called “What I understood about your needs”. It shows your need, company size, timeline and role as the architect understood them from the conversation. The architect has been instructed to fill them in only from what you have said, and the section shows only the details it has inferred. You see them before sending, but you cannot edit them. They are sent with the card when you send it. If they are wrong, you can say so in the card's message. The message does not change the stored details, but you can ask for incorrect information to be corrected or deleted at karo@staffbite.fi. If you do not want them included at all, you can use the contact form. The card below the conversation does not have this section.

A sent card is stored in the same way as a message sent with the contact form, and Staffbite is notified of it by email. The notification includes the details from “What I understood about your needs”, the conversation (at most its last 8,000 characters) and links to the conversation's solution sketches. No email is sent to you; after sending, the page shows a confirmation.

After you have sent the card, Staffbite may make an internal summary of it and the conversation with the Gemini model in Vertex AI's EU multi-region. No summary is made if the architect is taking a break, the daily limit for summaries is reached or the conversation is very short. The summary helps us prepare our reply, and it assesses, among other things, how well your need fits our services. The card's name and contact fields are not sent to the model, but the card's message and the conversation are, and names in them are not masked. The summary has no price, only Staffbite reads it, and no automated decisions are made on its basis. The summary is kept with the contact request for 24 months. The privacy policy has the details.

Frequently asked questions

Is the AI architect a person?

No. The architect is an AI, and its answers are produced by Google's Gemini model. It can be wrong. If you want to talk to a person, send the contact card or write to karo@staffbite.fi.

Which language does the architect answer in?

In the language of your latest message. On the English pages it starts in English, and if you write in Finnish, it answers in Finnish.

Why does the architect not give a price?

The price depends on the scope, so we do not publish prices. When you ask about the price, the architect explains what the scope depends on and offers the contact card. The solution sketch has no price either, only the scope and its duration as a range.

What is a solution sketch?

A structured sketch that the AI drafts from the conversation: situation, goal, approach, steps, scope, risks, stop criteria and first step. It is not an offer or a commitment. The architect can draft a sketch once you have described your situation and goal, and you can also ask for one yourself.

Is the EU readiness assessment legal advice?

No. The assessment is produced by AI and is indicative, and the card always says: Indicative assessment, not legal advice. It is based on what you have said in the conversation and on the site's public content. If the use case may be high-risk or prohibited, the architect recommends that a person reviews it.

Who can see my solution sketch?

Anyone who has the sketch's share link. The link contains a random identifier, and the page is not in search engines, the sitemap or the llms.txt files. Staffbite can read the sketch, just like the conversation. The sketch is based on your conversation, so share the link only with people you want to see it.

How do I delete a solution sketch?

Choose “Delete conversation” below the conversation. The conversation's solution sketches are deleted at the same time, and their share links stop working. A single sketch cannot be deleted separately. The Start a new conversation button does not delete sketches.

Does someone read my conversation?

Yes. Conversations are stored, and Staffbite can read them. Email addresses and phone numbers you write are masked before storage, but other information is not. So do not write sensitive information in the conversation.

Where is my conversation processed?

The Gemini model runs in Vertex AI's EU multi-region in EU member states, and conversations are stored in UpCloud's Helsinki data centre. Google is a US company, so the US CLOUD Act applies to it regardless of where the data is.

How do I delete a conversation?

Choose “Delete conversation” below the conversation. The conversation is deleted from the database immediately, and the cookie from your browser. If you have already started a new conversation or the cookie has expired, ask for deletion at karo@staffbite.fi.

What information about my needs is sent with the contact card?

The card the architect offers may include a section called “What I understood about your needs”: your need, company size, timeline and role as the AI understood them from the conversation. You see them before sending, but you cannot edit them. They are sent only when you send the card yourself.

Does the architect work without JavaScript?

Yes. Without JavaScript your message is sent as an ordinary form, and the answer appears when the page reloads. With JavaScript the answer appears sentence by sentence as it is completed.