AI transparency

The assistant reads the ticket.
You read the answer.

BootDesk's AI layer is optional, identifies itself in every customer-facing reply, and leaves an attributable trail for every action it takes. This page says what it does, where it stops, and what is not shipped yet. Nothing here needs a PDF on request.

Start the 7-day trial Read the security page

Last updated 9 November 2026 · GRUPO OITO OÜ, Tallinn, Estonia

AI proposes. The operator decides. That is a configuration, not a slogan — a switch, a stop control, and a name on every action.

What the trail shows.

Every line below is readable inside your workspace, not in a monthly summary we send you.

01
Who acted
An assistant acting alone is labelled as such. A human action carries the user's name, not a shared login.
actor
assistant or agent
02
What it did
Each step renders in the thread, in the order it happened — no invisible searching, no silent edits.
tool call
search, fetch, write, run
03
Which model
The live catalogue is visible, and so is the region the request was routed to.
model + region
recorded per call
04

The assistant used tools3

View the callsHide the calls

✓lookup_order — ERP · order #8830

✓knowledge_base — delivery-policy.pdf

✓open_ticket — queue · support-tier-2

A tool call in the thread: what it looked for, where it looked, and what that call cost.

Good at the legwork.
Never at the accountability.

What the assistant doesthe job

01Reads the whole thread before it writes. Searches your knowledge base, your connectors and the live web, and cites every lookup it used.
02Drafts the reply, the weekly summary, the deck, the spreadsheet. Runs the flows you drew — the routing, the triage, the scheduled posts. Speaks to customers only where you put it in front of them, and only as the assistant.

The model catalogue is a product decision, not a configuration mistake: you pick the work, we maintain the models.

What it does not dothe boundary

01It does not decide on its own that a person is not needed, and it does not get that decision made for it by default. It does not conceal its work. It does not train on your conversations, and it does not speak as a human — every customer-facing message says it is the assistant. It does not reach outside the region you pinned or the permissions you granted.

If a limit here does not fit your operation, tell us before you sign — the fix is usually a rule, not a migration.

Controls that support the
transparency duties.

EU AI Act transparency and human-oversight rules apply to customer-service AI from 2 August 2026. These are the controls we ship that support them. We do not claim certification, because there is none to claim.

Disclosure

It says it is AI

Every customer-facing reply is labelled as the assistant. There is no mode where it passes for a person.

Oversight

A human can intervene

Takeover at any moment, handoff with context, pause and resume per assistant, review gates before a reply goes out.

Documentation

Purpose and limits written down

What each assistant is configured to do, where it may act, and when it must escalate — documented per assistant rather than implied by its name.

shipped

Data governance around it

Masking before the prompt, region pinning, retention you control, an open API and an export you can actually use.

Not shipped yet — and not in the trial.

We publish the gaps on the same page as the features, because a transparency page that hides them is a brochure.

AI voice agents and call transcription as a general release

Still in development. Calling is available as a channel today; an AI that handles the call end to end is not in the product.

A self-service data-subject portal

Deletion and export requests are handled by your own team through the API. A tenant-facing portal is not built yet.

Automated per-tenant retention and erasure

Retention windows are yours to set today; scheduling erasure automatically is on the list, not in the release.

On-premise or single-tenant deployment

BootDesk is an EU-hosted multi-tenant SaaS. EU hosting and EU-only AI routing are the answers available today.

The questions this page
has to answer first.

Something missing? Write to suporte@bootdesk.eu — an engineer answers, not a form.

Can I turn the AI off completely?

Yes, per tenant, in settings. The inbox, tickets, CRM, mail, files, calendar and boards keep working without it. When it is on, assistants can answer, act and run scheduled work — and they can stop, hand over and ask for review.

Which models can it use, and where do they run?

The live catalogue is visible in the workspace, along with the region options behind it. You pin inference to the regions your policy allows; EU-only is an ordinary setting. Coverage depends on which models a given feature needs — the product shows what is actually available rather than promising a catalogue it cannot route to.

Do you train on our conversations?

No. Not our models, not a provider's. With your own provider keys the usage bills to your accounts directly, and the training posture does not change.

Can I bring my own model keys?

Yes. Plug in your provider and the token usage bills against your accounts, with the split visible in your wallet. The floor still applies, and the abstraction stays the same.

Is BootDesk AI Act compliant?

We do not make that claim, because compliance is a legal conclusion and not a badge. What we ship are the controls the transparency and human-oversight duties ask for: people are told they are talking to AI, a person can always intervene, each assistant's purpose and limits are documented, and data governance sits around every call.

Does the assistant replace my team?

No, and it is not trying to. It takes the repeat work — the lookups, the drafts, the triage, the summaries — and leaves the negotiation, the exception and the empathy to the people. Every assistant is a seat on the board, with a level you set and a human at the root of the review chain.

Read the line before you sign.
Then break it on the trial.

The DPA is public on this site, with the subprocessor list and the transfer mechanisms behind it. Nothing on this page needs to be sent to you on request — it is either in the product, or we say so.

Read the DPA Start the 7-day trial

The assistant reads the ticket.
You read the answer.

What the trail shows.

Who acted

What it did

Which model

Good at the legwork.
Never at the accountability.

Controls that support the
transparency duties.

It says it is AI

A human can intervene

Purpose and limits written down

Data governance around it

Not shipped yet — and not in the trial.

AI voice agents and call transcription as a general release

A self-service data-subject portal

Automated per-tenant retention and erasure

On-premise or single-tenant deployment

The questions this page
has to answer first.

Can I turn the AI off completely?

Which models can it use, and where do they run?

Do you train on our conversations?

Can I bring my own model keys?

Is BootDesk AI Act compliant?

Does the assistant replace my team?

Read the line before you sign.
Then break it on the trial.

What it cost

What was masked

Who approved

The assistant reads the ticket.You read the answer.

What the trail shows.

Who acted

What it did

Which model

Good at the legwork.Never at the accountability.

Controls that support thetransparency duties.

It says it is AI

A human can intervene

Purpose and limits written down

Data governance around it

Not shipped yet — and not in the trial.

AI voice agents and call transcription as a general release

A self-service data-subject portal

Automated per-tenant retention and erasure

On-premise or single-tenant deployment

The questions this pagehas to answer first.

01Can I turn the AI off completely?

02Which models can it use, and where do they run?

03Do you train on our conversations?

04Can I bring my own model keys?

05Is BootDesk AI Act compliant?

06Does the assistant replace my team?

Read the line before you sign.Then break it on the trial.

What it cost

What was masked

Who approved

The assistant reads the ticket.
You read the answer.

Good at the legwork.
Never at the accountability.

Controls that support the
transparency duties.

The questions this page
has to answer first.

Can I turn the AI off completely?

Which models can it use, and where do they run?

Do you train on our conversations?

Can I bring my own model keys?

Is BootDesk AI Act compliant?

Does the assistant replace my team?

Read the line before you sign.
Then break it on the trial.