AI in sales: 8 use cases and the conditions without which they will not work

“Will AI help us with sales?” is a question to which the honest answer is: it depends on what with. A tool that summarises a call brilliantly can be useless for preparing a proposal - not because it is weak, but because the company does not have an organised price list.

That is why for each of the use cases below I describe not only what it does, but above all which condition has to be met. That condition decides whether the tool saves time or adds to it.

First, about risk

For each use case I give one of three levels:

• green - working material, stays inside the company,
• yellow - the content reaches a client, so a person always reads it,
• red - it contains numbers, deadlines or commitments, so a person checks them against the source.

I write more about how to check answers at each level in the article on checking AI answers.

1. Call summary and task list

Risk: green · Effort: low

What goes in: a transcript or notes from a call with a client. What comes out: a summary, agreements, tasks with deadlines and threads that were left open.

Condition: someone has to record the call or write it down. Without input material there is nothing to summarise.

Why it is worth starting with this: low risk, a benefit from the first use and the team sees the point straight away. It is a good choice to begin with in a company where people are sceptical about AI.

2. Preparing for a client meeting

Risk: green · Effort: low

What goes in: contact history, previous agreements, publicly available information about the client’s company. What comes out: a short note before the meeting - where we left off, what to ask, what we can propose.

Condition: agreements are recorded in one place, not in individual people’s inboxes. If the client’s history is scattered, the tool will not piece it together.

3. Tidying up notes in the CRM

Risk: green · Effort: low

What goes in: raw notes, emails, messages. What comes out: a consistent entry with the stage of the conversation, the next step and a deadline.

Condition: a CRM exists and has an agreed entry structure.

Note: if the company does not use a CRM, this task is really a CRM implementation, not an AI implementation. It is worth saying so plainly before buying anything.

4. First draft of a reply to an enquiry

Risk: yellow · Effort: low

What goes in: the enquiry, service descriptions, the price list and a few good replies from the past. What comes out: a draft reply to read and send.

Condition: organised service descriptions and an up-to-date price list. Without them the draft will be generic and it will be quicker to write the reply yourself.

What to measure: time to the first substantive reply and the number of corrections that have to be made.

5. Answers to common questions and objections

Risk: yellow · Effort: medium

What goes in: a list of recurring questions, the good answers given so far, rules for the tone the company uses with clients. What comes out: a library of ready answers to adapt in a specific conversation.

Condition: collecting the real questions from the team - dozens of them, not a handful. This is work to do before the implementation, not during it. Questions invented at a desk differ from the ones that actually get asked.

6. Qualifying and sorting enquiries

Risk: yellow · Effort: medium

What goes in: incoming enquiries and the criteria for a good client. What comes out: an enquiry described, assigned to a person and marked with a priority.

Condition: written criteria for what counts as a valuable enquiry. Companies usually do not have them - every salesperson judges in their own way. Agreeing on these criteria alone is often half the value of the whole exercise.

Caveat: the decision to reject an enquiry is made by a person. The tool sorts, but does not cross out clients.

7. Follow-up after a proposal has been sent

Risk: yellow · Effort: low

What goes in: the proposal, the date it was sent, the contact history. What comes out: a draft reminder message matched to the stage of the conversation.

Condition: someone keeps track of the dates. The tool will not remind you of them on its own unless it is connected to a calendar or a CRM. A draft message achieves nothing if nobody knows it is time to send it.

8. Draft proposal based on what was agreed

Risk: red · Effort: medium

What goes in: notes from the call, the price list, the service catalogue, a proposal template. What comes out: a completed draft proposal with the places where a person decides clearly marked.

Condition: a standardised proposal template and clear pricing rules. Prices are always checked by a person against the source.

Note: this use case saves the most time and at the same time carries the biggest risk. A proposal contains prices and deadlines, and a mistake in them goes straight to the client. That is why it is not a good first step - especially in a company that does not have an organised price list.

Which one to start with

If I had to suggest an order, it would be:

1. To begin with a call summary or preparing for a meeting. Low risk, a quick effect, no need to tidy up data.
2. Then a first draft of a reply to an enquiry - once the service descriptions and price list are in order.
3. Last a draft proposal - once the team has the habit of checking answers and the proposal template and pricing rules are written down.

What all eight have in common: the tool processes what the company already has. It will not organise the price list, write down the criteria or set up a CRM. If these things are missing, the first step is putting them in order - and it is worth counting that honestly as separate work.

I write more about the order of the whole process - from talking to the team to the pilot - in the article on where to start with AI in your company. If you want a chosen use case built and tested on your team’s tasks, I do that as part of AI implementation.