AI in sales: what really helps a calling team today
AI in sales helps most around the call: researching accounts, summarising notes, drafting follow-up emails, grouping reasons for loss and transcribing calls for coaching. It is weak or risky where trust and judgement matter, such as fully automated cold calls and forecasts built on thin data. Start with clean, structured call data, clear rules on which data may go into which tool, and a person who checks every output.
AI in sales helps a calling team most with the work around the call: preparing accounts, summarising notes, drafting follow-up emails, spotting patterns in lost deals and transcribing calls for coaching. It helps least, and carries the most risk, where trust and judgement matter: the cold call itself and forecasts built on thin data. And every AI tool is only as good as the call data you feed it.
This guide to AI for sales teams covers what works, what does not, the GDPR and EU AI Act basics, and a checklist to start without regrets.
How many companies use AI in sales today
Plenty of teams are experimenting. According to Eurostat, 19.95% of EU enterprises with ten or more employees used AI in 2025. Among those users, marketing or sales was the most common purpose (34.70%).
The same survey shows what holds companies back. Of the enterprises that considered AI but did not use it, 70.89% pointed to a lack of expertise, 52.52% to unclear legal consequences and 48.83% to data protection and privacy concerns. Those three topics run through this article.
What research says about AI at work
Two well-known studies give a balanced picture. In a study of 5,172 customer support agents, access to an AI assistant raised issues resolved per hour by 15% on average. Less experienced and lower-skilled workers improved in both speed and quality, while the most experienced saw small gains in speed and small declines in quality (Brynjolfsson, Li and Raymond).
A field experiment with 758 consultants found a "jagged frontier". On tasks within AI's capabilities, consultants using AI completed 12.2% more tasks, 25.1% faster and with better quality. On a task outside that frontier, they were 19% less likely to reach a correct solution than colleagues without AI (Dell'Acqua et al.).
The lesson for sales: AI can get a new caller up to speed faster, but on judgement calls it can make confident people wrong.
Where AI helps a calling team today
Six common uses at a glance. The first four are the safest place to start; the last two need more care, as the next section explains.
| Task | What AI does well | What to check |
|---|---|---|
| Account research and list preparation | Summarises a company website, suggests segments and opening questions | Invented or outdated facts |
| Summarising notes | Turns rough notes or a transcript into a summary and a next step | Which personal data goes into which tool |
| Follow-up emails | Writes a first draft from your notes | Generic tone, wrong details |
| Reasons for loss | Groups "not interested" reasons and notes into themes | Only works if reasons are logged consistently |
| Transcription and coaching | Transcribes calls and finds moments worth reviewing | Informing people, rules on monitoring staff |
| Forecasting | Spots trends in pipeline history | Needs a lot of consistent data |
Research and list preparation
Before a call, AI can read a company's website and give the caller three lines: what the company does, a likely problem and a good opening question. That saves minutes per lead. The risk is a summary that sounds right but is wrong, so the caller checks the source before quoting it.
Do not let AI guess personal details such as names or direct numbers. Under the GDPR, personal data must be accurate (Article 5(1)(d)), and you need to know where your data came from. Our guide to B2B lead lists and the GDPR covers sourcing.
Summarising notes and drafting follow-up emails
This is the quickest win. A caller types five rough lines after a conversation, and AI turns them into a clean summary, a next step and a draft email that refers to what the prospect actually said. That beats a template.
Read every email before it goes out. You are the sender, not the tool. Our guide to sales follow-up explains what a good follow-up contains.
Analysing reasons for loss
If every "not interested" has a reason from a fixed list plus a short note, AI can group hundreds of notes into themes in minutes: price, timing, an existing supplier, the wrong person. That tells you whether to change your list, your pitch or your offer. Without logged reasons, there is nothing to analyse.
Where AI is weaker or riskier
Call transcription and coaching
Transcripts make coaching easier: a manager can review the opening and the close of ten calls in the time one full call used to take. But recording and transcribing come with duties. The EDPB says clients must be told about the purposes of the recording, who receives it, their right to object and their right to access it.
The AI Act adds limits for your own staff. Emotion recognition in the workplace is prohibited. The Commission's guidelines on prohibited practices give a call centre as the example: tracking employees' emotions with voice recognition is prohibited, with a narrow exception for personal training whose results never reach HR or affect assessments. Tracking customers' emotions is not covered by that ban.
AI that monitors and evaluates the performance and behaviour of workers is listed as high-risk in Annex III. Use transcripts to help people improve, not to score them in secret. Our article on coaching a sales team with data shows how.
Forecasting
Forecasting models learn from history: stages, deal sizes, wins and losses. A small team with a few dozen deals a year rarely gives a model enough consistent data. A weekly look at your own conversion rates is often more reliable, and you can explain it to anyone.
AI cold calling
Voice agents that dial prospects and hold a conversation now exist. For B2B cold calling in Europe they are a risky choice, for three reasons.
- Transparency: under Article 50 of the AI Act, AI systems that interact directly with people must be designed so those people know they are dealing with an AI.
- Consent: the ePrivacy Directive (Article 13) only allows direct marketing through automated calling systems without human intervention to subscribers who gave prior consent. That rule covers natural persons, such as sole traders. National calling rules apply on top, see our overview of cold calling rules in Belgium and the Netherlands.
- Trust: a cold call works because a real person took the time to prepare and listen. Remove the person and you remove the reason to stay on the line.
This is general information, not legal advice. Check your own situation with a lawyer before you automate any outbound calls.
The risks: data protection, hallucinations, generic messages and trust
The legal points below are general information, not legal advice. The linked official sources have the details.
Data protection under the GDPR
Pasting a prospect's name, number and call notes into an AI tool is processing personal data. The usual GDPR rules apply: a legal basis, transparency, and a contract with the provider if it processes data on your behalf (Article 28).
The Dutch Data Protection Authority has received data breach notifications caused by employees entering customer or patient data into chatbots. It notes that most companies behind chatbots store all data entered, and that using a chatbot against company agreements with personal data is a data breach. Its advice: agree with staff which data may and may not go in, and consider arranging with the provider that inputs are not stored.
EU AI Act basics
The AI Act entered into force on 1 August 2024 and applies in phases, according to the European Commission:
- Since 2 February 2025: prohibited practices and AI literacy.
- Since August 2026: transparency rules, for example for chatbots.
- From 2 December 2027: rules for high-risk areas such as employment, moved back by the 2026 AI Omnibus.
AI literacy matters to every sales team. According to the Commission's AI literacy Q&A, a company whose staff use ChatGPT to write texts must comply, and staff should be informed about specific risks such as hallucination. No certificate is needed; an internal record of training is enough.
Hallucinations
NIST describes confabulation, better known as hallucination, as AI producing "confidently stated but erroneous or false content" (NIST AI 600-1). In sales that looks like an invented customer reference, a wrong revenue figure or a feature you do not offer. Repeat one of those on a call and the trust is gone.
Generic messages
AI makes it cheap to send a thousand emails that all sound the same. Prospects notice, and so do mailbox providers. Google's sender guidelines require all senders to keep their spam rate in Postmaster Tools below 0.3%. Fewer, more personal messages protect both your reply rate and your domain.
Trust
Prospects give you a few minutes because they expect a person who has done their homework. If a message or a call feels machine-made, that goodwill disappears and is hard to win back. Use AI to prepare better, not to fake personal attention.
AI is only as good as what you log
Every useful AI task above depends on structured data. A summary needs notes. Loss analysis needs reasons. A forecast needs consistent stages. Before you buy an AI sales tool, make sure your team logs these basics on every call:
- One outcome from a short, fixed list.
- One next step with a date.
- A reason from a fixed list for every "not interested".
- Two or three lines of notes after every real conversation.
A calling workspace such as Ringlane does this without any AI: each call is logged with one of six fixed outcomes and notes, the outcome plans the follow-up, and the reports show the reasons for "not interested". That kind of structured history is the foundation any AI tool needs later.
Questions to ask before you buy AI sales tools
Most AI sales tools look impressive in a demo. Ask these questions before your team uses one with real prospect data:
- Where is our data stored and processed, and is there a data processing agreement?
- Are our inputs or call recordings used to train the provider's models, and can we switch that off?
- Can we see where each fact in a summary or email came from?
- Does the tool decide things about people, or only suggest things that a person approves?
- Does it analyse the emotions or performance of our own staff? Then check the AI Act limits above first.
- What happens to our data when we stop using it?
Checklist: how to start with AI in sales
- Pick one task to start with, for example drafting follow-up emails.
- Write down which data may and may not go into which tool.
- Use a business plan with a data processing agreement, and check where data is stored and whether inputs are used for training.
- Fix your logging first: fixed outcomes, fixed reasons, short notes.
- Keep a person in the loop: every AI text is read, every fact checked.
- Tell people when you record or transcribe a call.
- Do not use emotion recognition on your own team.
- Train the team on what the tool does and where it fails, and keep a short record of that training.
- Measure one number before and after, for four weeks.
- After a month, decide: keep it, change it or stop.
Frequently asked questions
How can AI be used in sales?
Can AI make cold calls for me?
Is it allowed to put customer data into ChatGPT?
What does the EU AI Act mean for a sales team?
Will AI replace sales reps?
Sources
- Eurostat: Use of artificial intelligence in enterprises (data extracted December 2025)
- Brynjolfsson, Li and Raymond: Generative AI at Work (arXiv, revised November 2024)
- Dell'Acqua et al.: Navigating the Jagged Technological Frontier, Harvard Business School Working Paper 24-013
- Regulation (EU) 2016/679 (GDPR), EUR-Lex
- Dutch Data Protection Authority: Caution, use of AI chatbot may lead to data breaches (6 August 2024)
- EDPB Data Protection Guide for Small Business: FAQ on recording telephone conversations
- European Commission: AI Act, regulatory framework and timeline
- European Commission: AI literacy, questions and answers
- European Commission: Guidelines on prohibited AI practices under the AI Act
- AI Act Service Desk: Annex III, high-risk AI systems
- AI Act Service Desk: Article 50, transparency obligations
- Directive 2002/58/EC (ePrivacy Directive), Publications Office of the EU
- NIST AI 600-1: Generative Artificial Intelligence Profile (July 2024)
- Google Workspace Admin Help: Email sender guidelines