AI policy

Eventflow uses artificial intelligence in specific, narrow places where it earns its keep. AI in event tools is currently being marketed badly: oversold as the headline, underexplained when it actually matters. This page exists to be specific.

the principle

AI in eventflow is a draft layer, not a decision layer. It proposes; the human approves. Invoice checks, letter drafts and imported lineups are reviewed by your team before anyone relies on them, and AI never approves an invoice, sets a status or pays anything. Three outputs are shown without a person reviewing them first, and we name them here so you know where they are:

• the first reply in our support chat, written by an AI model; the conversation can be handed to a person at any time;
• the automatic pre-check an artist sees after uploading an invoice, before your team has looked at it; your team still reviews the invoice and decides its status;
• the translations of your briefing headlines, which are machine-translated into the other artist languages when you save them and shown to artists as they are; you can change or remove a headline at any time, and it is translated again when you save.

The scheduled invoice verification described below does not reach you or your artists: its results are reviewed by us, the operator of eventflow, to improve our prompts.

This is partly an ethical commitment and partly a craft argument: a venue's relationships with its artists are the thing that distinguishes one room from another. Eventflow uses AI to take the friction out of repetitive work; the booker still does the booking part.

where ai is used

1 · Invoice reading and checking. When an artist uploads an invoice, eventflow extracts the structured data (fee, dates, tax details) and checks it against what was agreed for the slot, VAT and KSK aware. When you write invoice instructions for your workspace, a model turns them into the list of requirements the check applies; your team sees that list. The same check runs as a pre-check on the artist side before submission, so fixable issues surface early. The AI never approves or pays an invoice; a human flips every status, and every extraction is stored for review. A second, independent model re-checks stored invoice checks on a schedule. Its results are reviewed by us, the operator of eventflow, in our internal AI dashboard to improve the prompts of the check; they are not shown to your team or your artists and never change a check or a status. The pre-check an artist sees after uploading is shown to the artist straight away, without prior review, and is labelled as an automatic check.

2 · Learning from your corrections. When your team corrects an invoice check, eventflow turns the correction into short guidance for future invoice checks in your workspace. That guidance is stored in your workspace, used only there, visible to your team and deleted with the workspace. It is not used for other customers. Your own written invoice instructions always outrank it.

3 · Briefing and communication drafting. Artist briefings and operational emails are generated from your event data and templates you control, rendered in the artist's language where configured, and rewritten to match your venue's voice. Sends default to human approval. Briefing headlines are the exception named above: their translations are made on save and shown without review. You can edit any draft, disable rewriting per template, or turn it off entirely.

4 · Event descriptions and platform copy. Eventflow can draft the public description of an event and paste-ready copy for platforms like RA, using your lineup data and your venue's voice configuration. You edit before anything is published; eventflow does not auto-submit to any platform.

5 · Support chat. Messages to our support chat are answered first by an AI model that draws on our published help articles, without a person reviewing the reply first; it can link articles and hands the conversation to a person when needed. We run the support chat as our own processing, as controller, and Anthropic processes these messages as our processor. Short guidance learned from support conversations is kept across customers to improve these first replies, and it is visible only to us.

Providers. The models are provided by Cloudflare (Workers AI), OpenAI and Anthropic, as listed on the sub-processor page. We reach OpenAI through Cloudflare's AI Gateway, and Anthropic through our gateway or direct API calls. When a new AI feature ships, we update this page.

what we don't use ai for

Booking decisions: who plays your room, what they are paid, and whether a request is accepted are human decisions, and eventflow will never score or rank artists for booking. Autonomous communication without your approval flow. Profiling artists or guests: no predictive scores, no demographic analysis of your door data. Generative audio or artwork that imitates a specific artist's work. And we never generate or alter financial records: the AI reads invoices, it does not write them.

training data

We do not use your data to train AI models, and we do not allow our AI providers to use your data to train their models. Requests through our AI gateway run with payload logging switched off and a zero data retention request. We do not run our own model training on customer data, and we do not sell, share, or otherwise hand your data to third parties for AI training purposes.

disclosure

Communications drafted with AI assistance and approved by your team are not marked as such on the artist-facing side, because the human approves and signs every draft; the message is yours. Outputs that nobody reviews first are different: the invoice page labels its pre-check as an automatic check, and the support chat will label its first replies as automated answers. Machine-translated briefing headlines are not marked yet. If a customer requests a system to mark AI-assisted communications, we will consider it as a feature.

bias, errors, and accountability

AI models can produce errors, hallucinations, or subtle bias. Eventflow's drafts and extractions are based on structured data you provide, which reduces but does not eliminate the risk. The human approval step is the safeguard wherever there is one, the three automated outputs named above can be corrected as described there, and invoice checks are deliberately conservative: a mismatch flags for review rather than auto-resolving. If the AI produces something inappropriate or wrong, report it via the contact details in the Impressum; we treat these as bugs and prioritize fixing them. The operator remains responsible for what gets sent under the venue's name. Eventflow is a tool, not a delegate.

eu ai act

Eventflow's AI features fall under the limited-risk and minimal-risk categories of the EU AI Act. The Act's transparency duty for AI systems that interact directly with people, such as chatbots (Art. 50), applies from 2 August 2026. The artist invoice page labels its automated pre-check as an automatic check, and the support chat will label its automated answers as such. We publish this page to be specific about the rest. As the Act's implementing rules and guidance evolve, we will update this page.

questions or concerns

If you have questions about how AI is used in eventflow, want to opt out of specific AI features, or want to raise an ethical concern about something the system did, reach us via the contact details in the Impressum. We respond personally.