Skip to main content
Guides

How to onboard your first AI salesperson: from knowledge to pilot

Onboarding an AI salesperson in a Danish sales team: knowledge, tone, systems, pilot and rules. The mistakes to avoid, and the numbers that show if it works.

Author: Agent360 editorial team
Published: 30 January 2026
Last updated: 7 November 2026
Reading time: 12 min read

How long does it take to put an AI salesperson to work?

Many sales managers hesitate over AI, not because they doubt it can help, but because they fear the process. "It will surely take months to train." "We don't have time for a big IT project." The fear is understandable, and it is partly justified: doing it well takes longer than the sales brochures say. At the same time, it is quicker and less dramatic than many think.

The honest answer is that a first small pilot can run within days, while a solution you can rely on takes weeks. The time depends on how well your knowledge is gathered, how many systems need to be connected, and how much you dare let the AI do on its own from the start. Talk of "ready in 24 hours" is rarely anything more than a demo. On the other hand, an onboarding does not have to be a big IT project if it is taken in the right steps.

The key points in 30 seconds:

  • Onboarding an AI salesperson is mostly a question of knowledge, boundaries and patience. The technology takes up less of it.
  • Start with a small pilot, read the results, and only then scale.
  • Your colleagues need to be involved from day one. Many failed AI projects founder on resistance, not on technology.
  • The rules on marketing, personal data and AI must be settled before anything goes live.

AI salesperson onboarding is the structured process that turns an AI tool into a useful colleague in sales: it gets your knowledge, your tone and your boundaries, is connected to the systems you work in, and is tested on a small scale before it is allowed to do more.

Why do so many sales managers hesitate?

There are four reasons, and all of them are worth taking seriously.

  1. Fear of complexity. Many associate AI with long IT projects. A delimited task can be started much faster.
  2. Distrust of quality. "What if it says something wrong to a customer?" That question is right, and the answer is boundaries and approval, not blind trust.
  3. Law. Calls, recordings and personal data are regulated. It requires that you know what you are doing.
  4. Colleagues. If the sales team discovers the AI without having heard about it, resistance arises.

The positive side is that, according to Eurostat, Danish businesses led the EU in 2025 when it comes to using AI. Competitors have not been left behind by the technology, so it pays to learn it thoroughly.

What is an AI salesperson, and what is it not?

The word covers three quite different things:

  • An assistant to the salesperson. Prepares meetings, writes email drafts, makes notes and updates the CRM. The salesperson is still the one who talks to the customer.
  • A follow-up agent. Follows up on offers, reminds people of agreements and sorts incoming enquiries. A person approves what is important.
  • A voice agent. Makes or receives calls on its own. This is the most demanding variant, both technically and legally.

Most businesses get the most out of starting with the first two. They give time back without the same risks, and they teach the team to work with AI.

Phase 1: What does the AI salesperson need to know?

An AI without knowledge of your business answers generically. All effectiveness starts with the quality of what you feed it.

A minimal set of material:

  1. Your best sales presentation, with the messages and examples you yourselves use.
  2. Your website, so that language and product descriptions match.
  3. The fifteen questions customers ask most often, with your standard answers.
  4. Your known objections and the answers that usually work.
  5. Examples of good conversations, preferably from your best salesperson. These are optional, but they calibrate tone and rhythm.

The more precise the material, the sharper the result. If the material is thin, the answers will be thin. It is also important that you own the knowledge and the rules, not a tool you cannot take with you.

Phase 2: Tone, boundaries and handover

What the customer experiences is tone and content. Write it down in plain language, for example:

"Be direct when the customer says they are short of time." "Never mention prices without asking a colleague." "When the customer asks whether you are a human, tell them the truth."

Three things must be settled:

  • What must it never say? Prices, promises, legal assessments and anything that binds the business.
  • When does it hand over to a person? In cases of anger, complex questions, complaints and anything not covered by the material.
  • Who approves? You decide what the system may send on its own, and what waits for an approval.

Involve at least one experienced salesperson. Not only because they know what works, but because the team then has ownership from day one. Test internally: let colleagues play customers, and see how it responds to "Who are you?", to rejection and to questions it cannot answer.

Phase 3: Connections to your systems

An AI salesperson that does not know your CRM, calendar and mail creates more work, not less. The aim is for the information to flow where you would otherwise have had to type it yourself.

Typical connections and what you need to clarify

System What is typically connected Questions you need answered
CRM Contacts, activities, agreements Is there access through an interface? Which licence does it require?
Calendar Free slots, meeting appointments Is the system only allowed to look, or also to create?
Mail Drafts, follow-up Must everything be approved, or may something be sent on its own?
Telephony Calls, recordings, transcripts Who is the data processor, and where is the audio processed?

Each system is different. What is simple with one may require an extra licence or help from the supplier with another. That is why it is wise to clarify access early, before the rest of the plan is laid.

Phase 4: Pilot, not big bang

This is where many make the classic mistake: they send the AI straight out to their warmest customers. Instead, start with something where the risk is low and the learning is large.

A good first test group is contacts already in the CRM that you have given up following up on. There is little to lose, and the feedback is realistic: real objections, real rejections, real answers.

Expect the picture to look like any other cold contact: many without an answer, some rejections, a small group of real conversations. The most important thing is not the first booked meetings, but what the rejections say about tone, timing and message. Adjust, and run another small round.

What do the rules require?

This area calls for care, and you should get a legal assessment of your specific setup.

  • The Danish Marketing Practices Act. The rules on telephone and electronic marketing (including section 10) depend on whether the recipient is a business or a private individual, and on the channel. The use of AI voices in outbound calls is an area under development.
  • Personal data. Processing of contact details and recordings must have a lawful basis. Information to the data subject, data minimisation, retention periods and a data processing agreement with external suppliers all belong here. Also check where the data is processed.
  • The AI Act. It requires transparency when people communicate with an AI system. Check the applicable deadlines for when the requirements take effect in practice.
  • Openness towards the customer. Regardless of legal requirements: never let a customer believe they are speaking with a human if they are not.

The three most common mistakes

Mistake 1: Knowledge that is too thin. The sales manager puts in a three-page brochure and expects the AI to understand the whole business. The result is answers that sound right but miss the mark. Solution: gather the material, and write down the fifteen most important questions and answers.

Mistake 2: No internal communication. The AI is introduced without the team knowing. Two weeks later a salesperson discovers that "there is a robot contacting my customers". Solution: talk to the sales team before anything goes live. Show what the system does, what it does not do, and what time it frees up.

Mistake 3: Scaling too quickly. The pilot went well, so several hundred are sent out the next day, without anyone having read what the pilot actually showed. Solution: scale step by step, and only once the previous step has been read and understood.

Which numbers show whether it works?

Five numbers are enough for the first 30 days:

  1. Contact rate. How many attempts lead to a real conversation?
  2. Agreement rate. How many conversations end in an agreement?
  3. The objections. Which come up most often? This steers the next round of material.
  4. Length of conversations. Very short ones suggest the opening is not working. Very long ones suggest the process is too complicated.
  5. Data quality. How many contacts end up with complete fields in the CRM?

Compare with your own numbers from before, not with industry averages. It is your funnel that must get better.

What does the sales team look like afterwards?

AI does not replace salespeople. It takes over the part of working time that is not selling. The team typically gets three roles:

  • The AI tool. Prepares, follows up, notes and updates.
  • The human salesperson. Focuses on meetings, negotiation and relationships, and arrives at the meeting better prepared.
  • The sales manager. Sees what works without guessing, and can coach on what was actually said.

Which industries suit it best?

It depends less on the industry than on the form of the sale. Easiest are sales with clear products, many enquiries and fixed questions: insurance, energy, installation and software with a simple message. Hardest are complex B2B processes with many decision-makers and long sales cycles, where trust and understanding count most. Here it is often best that the AI supports the salesperson rather than taking over.

What can Agent360 help with?

Agent360 does not sell a finished AI salesperson, and does not promise that it will be ready in a certain number of hours. We build the part that fits you, on top of the systems you already use, as part of Your AI department:

  • We start by understanding how your sales work and what takes the time.
  • We finish one thing properly before we move on. One example is our sales cockpit: a call list for the salespeople, where the conversation is written out in Danish, AI suggests the outcome and drafts the follow-up email, and the salesperson sends it themselves. We have built that for our own sales.
  • We can build a meeting recorder that finds the promises in the minutes and turns them into tasks with an owner.
  • You decide what the system may do on its own. Everything else waits for approval.
  • The price depends on the task, and you receive a quote before you commit.

See examples of tasks we can help with on the use cases page. If you want to learn to use AI on your own tasks, there is the personal AI mentoring programme: three individual meetings, fixed price and no subscription. Read about the programme.

For AI-based calls, we mention our partner JesperAI. It is their product, and we do not promise anything on their behalf.

FAQ: Bringing an AI salesperson into use

Q: How long does it take to onboard an AI salesperson? A small pilot can be set up within days. A solution you trust and can let work on its own takes weeks, because knowledge has to be gathered, systems connected, and the results read and corrected. Be sceptical of promises of a short time from start to full operation.

Q: Can an AI speak Danish naturally? It has become markedly better, but the quality varies with the task, the sound and the tool. Always test on your own examples and listen to the results before you let anything go out to customers.

Q: Are AI calls lawful in Denmark? It depends on whom you call, how, and what is said. The Danish Marketing Practices Act, the data protection rules and the AI Act are all relevant, and the area is in motion. Get a legal assessment of your specific setup.

Q: What happens when the AI meets a question it cannot answer? You must decide that in advance. A sensible rule is that it says it does not know, and either hands over to a colleague or notes the question, so your knowledge can be expanded.

Q: What does it cost? It depends on what needs to be built. Agent360 gives a quote after looking at the task, and you commit to nothing before then.

Q: How do you handle resistance in the sales team? By bringing people along from the beginning. Explain what the AI takes over (administration, preparation, follow-up) and what it does not take over (relationships, negotiation). Let them see the results and point out errors.

Q: Which systems can be connected? It depends on your CRM, calendar, mail and telephony, and on which licences and access you have. We clarify that first in a review.

Next steps

You do not need an IT department, but you do need three things: willingness to try, material to start with, and access to the systems that must be connected. The rest can be built in small steps.

Would you like to hear what a pilot could look like at your company?

Book a free 20-minute consultation, and get an honest view of where it makes sense to begin.

Read on:

Should your sales reps spend their time talking?

We have built it for our own sales: the call is transcribed, the AI suggests the outcome, and the email is ready. We look at how your sales run today.

How we built it · Sales & follow-up

Sales cockpit with AI

A call list for salespeople. The call is transcribed in Danish, AI suggests the outcome (e.g. “Call again”) and drafts the email based on what the customer said.

  1. The salesperson calls
  2. AI writes transcript and email
  3. The salesperson sends
See the call and the follow-up →