Skip to content

Buying guide

Conversation intelligence for European sales teams: what to check before you buy

A buyer's checklist for European sales leaders: data residency, Dutch, French and German transcripts, the EU AI Act, training, scorecards and pricing.

5 min read

A train pulling into a busy platform at Amsterdam Centraal station

Before you buy conversation intelligence for a European sales team, check six things: where your recordings and transcripts are processed and by which subprocessors, how well it transcribes Dutch, French and German, how the vendor reads the EU AI Act, whether it trains on your data, whether scorecards cost extra, and how the price is built.

Where are your calls stored and processed?

Ask for the region of every step, not only of storage. A call passes through recording, transcription, a language model that writes summaries and scores, and a database. A vendor can store data in Frankfurt and still send audio to a transcription service outside the EU.

Three documents answer the question:

  • The subprocessor list. Under GDPR a processor needs your authorisation before it brings in another processor (GDPR, Article 28). The list shows every company that touches your calls, and where.
  • The data processing agreement. It should be available before you sign, not after.
  • The retention settings. Can you set how long recordings, transcripts and scores are kept, per team?

Drummer processes everything in the EU: AWS Frankfurt, EU transcription and EU-hosted models. The details are on our security page.

Does it understand Dutch, French and German?

Most serious tools now transcribe these languages; the real questions are accuracy on your customers’ accents, and whether the rest of the product works in them too. Gong, for example, lists Dutch (Netherlands and Belgium), French and German among its supported languages for calls, and also lists capabilities that are available in English only (Gong help centre, 2026).

Test with your own material. Give each vendor three recordings: a Flemish rep with a Dutch customer, a call in French with a Walloon or Swiss buyer, and a call that switches language halfway. Then check three things beyond the transcript: does the summary come out in the call’s language, is the follow-up email written in the customer’s language, and can your scorecard and its criteria be in the language your managers coach in?

Drummer transcribes 60+ languages, including Dutch, French and German, and drafts the follow-up in the language of the call.

What does the EU AI Act ask of a conversation intelligence tool?

At minimum, transparency. Article 50 of the Act says people must be informed when they interact with an AI system, and that AI-generated content must be detectable as such (EU AI Act, Article 50). A meeting bot that joins a call and records it should say so; content the AI wrote should be marked.

Software that evaluates employees can raise further questions under the Act and under labour law. Rather than rely on a vendor’s badge, ask for its written assessment of how the Act applies to its product and to your use: scoring your own people’s calls. In many European countries a monitoring tool also needs to go past the works council, so involve yours early.

Drummer’s bot announces itself when it joins a call, and everything the AI wrote is marked as such in the product.

Will the vendor train its models on your calls?

Ask for the answer in the contract, not on the website. Three things can differ: whether the vendor trains its own models on your calls, whether its model providers may, and what a clause like “to improve our services” allows in practice.

Sales calls carry prices, margins, customer names and strategy. A clear “no training on customer data” commitment, covering subprocessors, is the only safe answer. Ours is short: we don’t train our models on your data, ever.

Are scorecards included or a paid add-on?

In much of the market, scoring is an add-on or sits in a higher plan. Avoma sells AI scorecards in its Conversation Intelligence add-on at $29 per user per month billed annually (Avoma pricing, 2026). Leexi sells scorecards in its Leexi Coach add-on at €16 per licence per month billed annually (Leexi pricing, 2026). Fathom includes “Coaching metrics & AI scorecards” in its Business plan at $25 per user per month billed annually (Fathom pricing, 2026).

Beyond the price, check what a score contains. Can you build your own scorecard from your playbook, such as MEDDIC or your own discovery criteria? Is every call scored automatically, or only the ones a manager picks? Does each criterion come with the reasoning and the timestamp that proves it? Our coaching page shows what we mean by evidence.

How is the price built?

Two models dominate: a price per seat, or a price per seat plus a platform fee. Gong does not publish its prices. Sybill estimates Gong’s core licence at $1,400 to $1,600 per user per year, plus a mandatory platform fee from $5,000 to over $50,000 a year depending on team size (Sybill, 2026). tl;dv reports the platform fee as usually a minimum of $5,000 a year, rising to $50,000 or more (tl;dv, 2026).

Ask every vendor for the first-year total for your team: seats, platform fee, add-ons, onboarding, the minimum term, and whether managers and viewers need a paid seat. Our compare pages set the main tools side by side.

The checklist

Question to ask Why it matters What good looks like
Where are recording, transcription and models run? Storage in the EU says nothing about the processing Every step in the EU, named per subprocessor
Can I see the subprocessor list and the DPA now? GDPR needs your authorisation for each subprocessor Both available before you sign
How does it handle our languages and accents? A transcript in the wrong language breaks every score after it Tested on your own Flemish, French and mixed-language calls
How do you read the EU AI Act for this product? Transparency duties apply, and your use is employee evaluation A written assessment, a bot that announces itself, AI content marked
Do you or your providers train on our data? Calls carry prices, margins and strategy “No”, in the contract, covering subprocessors
Are custom scorecards in the plan you quote? Scoring is often a paid add-on or a higher tier Your own scorecards, every call scored, evidence per criterion
What is the first-year total for our team? Platform fees, add-ons and minimum terms hide in the quote One number covering seats, fees, add-ons and onboarding
Can we set retention and delete on request? GDPR storage limitation and the rights of the people recorded Retention per team, deletion that reaches backups and subprocessors

Take the table into your next vendor call. If you want to see how Drummer answers each line, start with the security page.

Questions readers ask

Is EU hosting enough for GDPR compliance?

No. Hosting is one part. Check where transcription and the language models run, which subprocessors touch your recordings, how long data is kept, and whether you have a signed data processing agreement. Recording consent and informing the people on the call are your side of the work, and the tool should make that easy.

Which languages should a European conversation intelligence tool handle?

At least the languages your team sells in, with the accents your customers speak. For Benelux that usually means Dutch, including Flemish, French and English, often German. Test the tool on your own recordings, and check that summaries, follow-up emails and scorecards work in those languages, not only the transcript.

Are scorecards usually included in conversation intelligence pricing?

Often not. Several vendors sell call scoring as a paid add-on or only in a higher plan, for example Avoma, Leexi and Fathom on their public pricing pages. Ask whether custom scorecards, automatic scoring of every call and evidence per criterion are in the plan you are quoted, or extra.

Give every rep a day back, starting next week.

Thirty minutes, on your own calls and your own CRM. No migration, nothing for your reps to learn.