AI Act Sales Tool Rules: Which Practices Are Prohibited and How to Stay Clear
An AI sales tool that drafts replies, quotes and follow-ups for your team is, in itself, very unlikely to fall under the AI Act's prohibited practices. Those bans target extreme uses: manipulating people, exploiting their vulnerabilities, scoring them socially. They are not aimed at drafting an email to a rental customer.
That said, a few features can drift toward the line, especially aggressive profiling and decisions taken without a person. Below you get a simplified breakdown, a table of sales activities against risk, and a checklist you can run through this week. This is general guidance, not legal advice. For your specific set-up, check with a lawyer.
What the AI Act prohibits, in plain terms
The AI Act sorts AI uses by risk. The top tier is a short list of practices that are banned outright. In everyday language, the banned categories include:
- Manipulation: techniques that work below a person's awareness or are deliberately deceptive, and that distort their decisions in a way that harms them.
- Exploiting vulnerabilities: targeting people because of age, disability or a difficult social or economic situation to influence their behaviour.
- Social scoring: rating people over time based on behaviour or personal traits, with unfair consequences.
- Certain biometric and emotion-related uses: for example inferring emotions in workplaces, or categorising people by sensitive traits from biometric data.
Details and exceptions matter here, and the text is updated through guidance. Treat this list as orientation and have a lawyer confirm anything borderline.
What this means for a B2B sales team
Most small B2B teams use AI for three things: reading enquiries, drafting replies and quotes, and reminding people to follow up. Here is how those activities usually sit against the banned categories.
| Sales activity | Risk to watch | Safe practice |
|---|---|---|
| Drafting a reply to an enquiry | Low. Only an issue if the draft is built to pressure or mislead | A person reads, edits and sends it |
| Summarising an enquiry and listing missing details | Low | Keep it to facts in the email |
| Prioritising which enquiry to answer first | Medium if it turns into a hidden score of people | Use business criteria (deadline, scope, stage), not personal traits |
| Profiling contacts from social media or inferred moods | Higher. Drifts toward emotion inference and manipulation | Avoid it. You rarely need it for B2B quoting |
| Urgency tricks such as fake deadlines in follow-ups | Manipulation concerns, plus consumer and unfair-practice rules | State real validity dates only |
| Declining or ranking a customer automatically | GDPR rules on solely automated decisions may apply | A person makes the decision |
Where ordinary sales AI really sits
Using an AI tool for enquiries and quotes generally makes you a deployer, not a provider. Your duties are practical: make sure the people using the tool understand it, be open with customers where relevant, and keep decisions with a person. We covered that role in detail in EU AI Act for small businesses, and the vendor side in the AI Act compliance checklist for sales assistant tools.
The general sales-AI question most people actually ask is the GDPR one. If that is you, start with the answer page Is an AI sales tool GDPR-compliant? and come back here for the AI Act angle.
Automated decisions in enquiry handling
The line that matters most in practice is between assisting a decision and taking it. GDPR has its own rules for decisions made solely by automated means that significantly affect a person. A B2B quote is usually not that, but the safest habit is the same either way: AI prepares, a person decides.
Things that should always stay with a person
- Whether to quote at all, or to decline an enquiry.
- The price, discounts and terms.
- Any statement about availability, dates or reservations.
- Anything sent to a customer.
Things AI can safely prepare
- A summary of what the customer asked for.
- A list of missing details to ask about.
- A draft reply, quote text or follow-up.
- A reminder that you promised to reply by Friday.
A six-step check for your sales tool
- List what the tool does. Drafting, summarising, prioritising, profiling, sending. Write each in one line.
- Mark anything that scores or profiles people. Ask what data feeds it and whether it includes personal traits or inferred emotions.
- Check that nothing leaves without approval. A person should review every outgoing email and quote.
- Review your templates and prompts for pressure tactics. No invented scarcity, no exploiting someone's situation.
- Train the people using it. A one-page internal rule on what AI may and may not do is enough to start.
- Ask the vendor in writing. Where is data stored, who processes the text, what is excluded. Our seven vendor questions give you a ready list.
A short email you can send to a vendor:
Hello, we are assessing your tool for handling sales enquiries and quotes. Please confirm in writing: (1) whether it profiles or scores individuals, and on what data; (2) whether anything is sent to customers without human approval; (3) where our data is stored and which provider processes the text; (4) how you support our obligations as a deployer under the AI Act. Thank you.
How BOTrento handles this
BOTrento Chief of Staff is a managed process for enquiries and quotes. It watches enquiries in your mailbox or uploaded data, shows a morning overview on working days, prepares drafts of replies, quotes and follow-ups, and records promises such as “I will get back to you by Friday” so they are not forgotten.
On the points above:
- Nothing is sent without a person. BOTrento does not send anything itself. Drafts of emails and quotes are approved by a human.
- No decisions on its own. It does not confirm dates or reservations, and it does not replace your CRM or accounting.
- Data location. Data is stored on servers in Germany (EU). The text needed for a task is processed by a commercial language model provider through its API, so an EU server is not the same as closed processing. If you want your own provider account, an optional own API key is possible (see the help page), currently Anthropic only.
- AI in the plan. AI use is included with a monthly usage limit, with a warning at 80%. You can compare the plans on the pricing page.
The easiest way to judge it is on your own enquiries: the 14-day free trial needs no card and no API key of your own, and the 14 days start with your first data upload. You can also see the flow in the process demo, and read more about what BOTrento is. If you run proposals for agencies or consultancies, the B2B services page shows the typical set-up. We do not claim that using any tool makes you automatically compliant. Your obligations stay with you, and a lawyer should review your specific case.
FAQ
Is an AI sales assistant banned under the AI Act?
No. The bans cover specific practices such as manipulation, exploiting vulnerabilities and social scoring. A tool that drafts emails and quotes for a person to approve is generally not among them, but check your exact use with a lawyer.
Can I use AI to prioritise enquiries?
Generally yes, if you rank by business criteria such as deadline, scope and deal stage, and a person makes the final call. Avoid building hidden scores about individuals from personal traits or inferred emotions.
Do I have to tell customers that an email was drafted with AI?
Transparency duties depend on the situation, and a person reviewing and sending the email changes the picture. Be open where it matters and ask your lawyer for the rule that fits your case.
Is my sales AI tool legally compliant?
Compliance depends on what the tool does, how you use it and what the vendor can show in writing. Use the six-step check above and ask the vendor your questions by email.
Summary
- The AI Act's banned practices target manipulation, exploiting vulnerabilities and social scoring, not ordinary email drafting.
- Risk grows with profiling people, inferring emotions and pressure tactics in follow-ups.
- Keep decisions and sending with a person: AI prepares, you approve.
- Run the six-step check and ask your vendor for written answers.
- This is general guidance, so confirm borderline cases with a lawyer.
Try BOTrento free for 14 days
Start the 14-day free trial on your own data. Our person will get in touch within 24 hours on working days and help you connect your mailbox. Nothing is sent without your approval. Fill in a few details and your access arrives by email right away. We are happy to help you connect your mailbox.
14 days free, no card and no API key of your own. The trial starts only once your data is first loaded. Then from €79 per month excl. VAT, AI included.
Nothing is sent without your approval. You add a card at the end, only if you continue; without payment access is blocked until you pay, we keep your data for 30 days and the export stays available. Pricing · See the process demo
Thank you, we have your sign-up.
We have sent a sign-in link and first steps to your email. If it does not arrive within a few minutes, please check your spam folder.