Three requirements to set for a sales AI tool before you run a pilot

You have sat through demos. You have heard the pitch about saved time and more booked meetings. Maybe you have already tried ChatGPT or Copilot and felt it was a step in the right direction, but that something was missing.

Before you put your team into a pilot, there are three questions you need to be able to answer yes to. If a vendor cannot clear them, the risk is high that the pilot becomes yet another tool nobody uses after week three.

Requirement 1: Context must be there from day one

The most common reason generic AI tools fail in sales is that the tool has no knowledge of who you are, what you sell, or where each customer sits in the pipe.

A salesperson about to call a customer who viewed your pricing page and had a meeting 14 days ago does not need a generic company summary. They need to know the ball has been in the customer's court for 14 days, what was said last time, and what the logical next step looks like right now.

Questions to ask the vendor:

• Does the system pull data from your CRM, calendar, and telephony automatically, or does the salesperson have to paste information in manually?
• Can the system read the history of a deal and give a concrete suggestion for the next step, rather than a general piece of advice?
• Does it work with the CRM you already use, whether that is HubSpot, Salesforce, Upsales, or Lime?

If the answer requires you to export data, reconfigure something, or train the system for months before it delivers value: do not expect to see results during the pilot.

Requirement 2: The salesperson must work exactly as they do today

Friction is the biggest enemy of AI adoption in sales. Every time a new tool demands a new login, a new view, or that the salesperson explains their intention from scratch, the likelihood of it being used the next day drops.

The majority of software developers use AI every week. Salespeople sit at the lowest figure among white-collar roles. The gap comes down to tools that have required the salesperson to adapt to the tool, rather than the tool adapting to the salesperson.

Questions to ask the vendor:

• Can the salesperson keep using their existing CRM, calendar, and telephony without changing how they work?
• Does the implementation require migration, training days, or replacing anything in your current stack?
• What happens to your existing systems if the tool stops working?

An answer you should expect: that master data is never touched and that existing systems continue to operate independently if the AI layer were switched off. If the vendor cannot answer the third question without hesitating, that is a warning sign.

Requirement 3: Follow-up must be built on real conversations

This is the requirement most vendors avoid addressing. The sales manager does not need scores, dashboards, or a ranking list. What the sales manager needs is concrete knowledge of what was said in a call with a customer, which objections came up, and how the salesperson handled them.

That is what call listening has always been about. The problem is that it takes time most sales managers do not have. AI solves that equation, but only if the system transcribes and analyses in the language the conversations actually happen in.

Swedish B2B calls frequently mix regional accents, code-switching between Swedish and English, and industry-specific terms. A system trained on English and calibrated for American sales processes will give you a transcription, but not a useful one.

Questions to ask the vendor:

• Does the system transcribe and analyse Swedish calls including accents and code-switching?
• Can the sales manager follow each individual call and get real context, not just a numerical score?
• Confirm what is stored, where it is stored, and who has access to it.

On that last question: a serious answer includes a data processing agreement, a clear description of where data is processed, and how you handle the Swedish Data Protection Authority's guidelines. If the vendor says it is GDPR-compliant without being able to specify how, ask for the agreement before you continue.

A note on privacy

Listening to recorded calls raises questions in most teams, and those questions deserve to be taken seriously. The employer owns the information in existing systems in line with the employment agreement, but that does not release you from handling it transparently with your team.

Ask the vendor to explain exactly what is logged, who sees it, and whether anything ends up in the CRM without the salesperson having approved it. Answers that do not hold up against a reading of your employment agreements and GDPR will not hold up against scrutiny from the Swedish Data Protection Authority either.

Test against your own day-to-day reality

Take a concrete situation from last week. A customer who has not replied in three weeks. A deal that stalled after a demo. A new-business outreach that nobody on the team had time to follow up.

Ask the vendor: what would the system have done in that scenario, step by step? If the answer is abstract or stops at the tool helping you prioritise, ask for a demo using your actual data.

The three requirements take ten minutes to go through. They can save you three months of a pilot that never delivered what was promised.