How can I use AI in sales?

AI can help with far more than writing sales emails.

It can research prospects, understand incoming enquiries, qualify leads, prepare meetings, follow up opportunities, create first drafts of proposals, update your CRM and help identify what needs attention next.

Some of that simply helps a salesperson work faster. Some can be automated. And increasingly, AI can take responsibility for moving defined parts of a sales process forward.

But that doesn't mean you should add AI everywhere.

The best place to start is usually much simpler:

Find the sales work that is taking too much time, getting missed or not happening properly.

Then decide whether AI is actually the right tool for it.

1. Respond to sales enquiries

Someone submits a form or sends an email.

Before anybody can respond, they may need to work out who the person is, what they want, whether you provide it, whether you've spoken before, what information is missing and what should happen next.

AI can help interpret the enquiry and bring together the information required to answer it.

At the simplest level, it can prepare a response for a person to check.

In a more agentic workflow, several steps could happen as part of the process:

• Enquiry received
• AI understands the request
• Checks relevant business information
• Identifies anything missing
• Prepares the response
• Human approves if required
• CRM updated and next action created

The interesting bit isn't that AI can write the email.

It's that it can potentially understand what has happened and help move the enquiry to the next appropriate step.

Good use when: you receive enough enquiries for response time and administration to matter.

Be careful when: enquiries regularly involve unusual pricing, sensitive situations or commitments requiring human judgement.

2. Qualify sales leads

Lead qualification is another area where AI can be useful because the information isn't always neatly structured.

A potential customer might tell you:

"We're looking at replacing our current system across three sites. Ideally we'd like something running before December, but we're still working out budget."

A rigid automation sees text.

AI can interpret what the person has actually told you.

It could identify:

• Requirement: replacement system
• Locations: three
• Timescale: before December
• Budget: not confirmed

It can then work out which important information is still missing.

Depending on your process, AI could read what the lead has already provided, ask appropriate follow-up questions, organise the answers and compare them with qualification criteria defined by your business.

It could then recommend what happens next.

The important part is defined by your business.

AI shouldn't quietly invent its own definition of a good customer.

3. Research prospects

Prospect research can consume a surprising amount of time.

A salesperson might check the company's website, previous CRM activity, old emails, account notes and other relevant sources before a conversation.

AI can help gather and organise that information.

But there's a trap here.

The goal isn't to turn ten minutes of research into a beautifully written 3,000-word AI report that takes 15 minutes to read.

A salesperson may simply need:

• Who are they?
• What does the company do?
• What do we already know about them?
• Have we spoken before?
• What appears relevant to this conversation?
• What don't we know yet?

Good AI research should reduce the amount of searching a person has to do.

Not create more material for them to process.

4. Prepare for sales meetings

This is one of the easier places to see how AI can create useful capacity without being given much authority.

Before a meeting, AI can potentially bring together customer history, previous conversations, open opportunities, outstanding actions and relevant account information.

Instead of somebody opening five systems before the call, they could receive something closer to this:

AI hasn't taken over the sale.

It has done the gathering.

For a salesperson, founder or account manager with several conversations in a day, that can be valuable in its own right.

5. Write sales emails with actual context

Yes, AI can write sales emails.

That's probably the least interesting thing on this list.

The improvement comes when AI has enough context to know what email needs writing.

There's a substantial difference between asking:

"Write a follow-up email."

and a system understanding who the customer is, what they asked for, what was previously discussed, what was promised, how long it has been and what should happen next.

The second isn't simply generating copy.

It's using business context to prepare part of the work.

Context matters more than clever copy.

And that context should come from information you understand and trust, not whatever the AI happens to infer.

6. Follow up sales leads

This is where AI can start moving beyond being a writing assistant.

A sales process might contain hundreds of future intentions:

"Come back to me next month."
"Send the quote and give me a few days."
"I need to speak to my business partner."
"Let's pick this up after the board meeting."

Someone then has to make sure those things happen.

An AI-enabled workflow could track when a follow-up is due, check whether the customer has replied, review the previous conversation and determine whether a follow-up is still appropriate.

For example:

• Quote sent
• Wait agreed period
• Has the customer replied?
• YES → Stop or continue appropriate workflow
• NO → Check previous context
• Is follow-up appropriate?
• Prepare message
• Approve / Send within limits / Escalate

The useful capability isn't merely generating:

"Just checking whether you've had a chance to look at my previous email..."

It's making sure the right next action actually happens.

7. Prepare quotes and proposals

A proposal often pulls information from several places.

• Customer requirements.
• Meeting notes.
• Pricing.
• Product or service information.
• Previous proposals.
• Templates.
• Case studies.
• Terms.

And, quite often, somebody's memory.

AI can help assemble those pieces into a first draft.

It might identify the customer's requirements, retrieve approved service information, populate standard sections, flag missing information and prepare the document for review.

But there's an important boundary.

Preparing a price is not necessarily the same as deciding a price.

The AI may be able to use approved pricing information to assemble a quote.

That doesn't mean it should independently decide that this particular customer receives a 20% discount.

One is preparation.

The other is commercial authority.

The technology being capable of both doesn't mean you should treat them as the same job.

8. Keep the CRM updated

CRM administration is one of the obvious places to look for work that surrounds selling rather than being the valuable sales conversation itself.

After an interaction, AI could potentially summarise what happened, log notes, create the next action, update agreed fields and identify information that appears to be missing.

But not every CRM field needs the same level of authority.

For example:

The exact boundary will vary by business.

The principle doesn't:

Don't give AI blanket permission when the job only requires specific permission.

9. Tell you what needs attention

This could be particularly useful for a small sales team.

Most business software expects the person to open it and work out what's happening.

AI gives us the possibility of reversing that relationship.

Instead of inspecting an entire pipeline, imagine beginning the day with:

The AI hasn't necessarily contacted anybody.

It hasn't closed a sale.

It may not even have changed anything.

But it has watched the process, interpreted what is happening and brought the exceptions to the person.

That can be an extremely useful job for an agent.

10. Move sales work between systems

This is where AI starts becoming more interesting than a standalone chat window.

Most businesses already have software.

The problem is often the human work required between the software.

An enquiry arrives through the website.

Somebody checks the CRM.

Searches email.

Looks at a document.

Creates a meeting.

Writes a proposal.

Updates the CRM again.

Creates a reminder.

AI, integrations and ordinary automation can potentially coordinate more of that work:

• Website enquiry
• AI understands
• CRM checked
• Relevant business information retrieved
• Response prepared
• Email sent or approved
• Meeting booked
• Meeting brief prepared
• Human conversation
• CRM updated
• Next action created

This is where AI begins to move beyond assisting with isolated tasks.

It starts participating in the workflow.

That's the idea behind agentic selling.

AI or automation: which do you actually need?

Not every sales task needs AI.

A useful way to think about it is:

Use it when the rule is predictable.

• Meeting booked → Send confirmation
• Form submitted → Create CRM record
• Deal marked won → Create onboarding task

You already know exactly what should happen.

Use it when something needs interpreting.

• What is this customer asking?
• Which information applies?
• What's missing?
• What kind of response is appropriate?
• Does this situation need a person?

Keep people where consequence, judgement, relationships or uncertainty make them valuable.

• Should we offer this discount?
• Should we agree to these terms?
• How should we handle this unhappy customer?
• What commercial commitment should we make?

The strongest sales workflow may contain all three:

AUTOMATION + AI + HUMAN

Not:

AI DOES EVERYTHING.

Where should you start?

Don't start by opening an AI tool.

Pick one piece of your sales process and draw what happens.

For example:

• Enquiry arrives
• Read
• Check customer
• Find information
• Respond
• Update CRM
• Create next action

Now look at every step.

Ask:

Does this genuinely need a person?

If it doesn't, ask another question.

Could ordinary automation do it?

If the answer is no because something needs understanding, context or judgement:

Could AI help?

And then:

Should AI recommend it, prepare it or actually do it?

That last question matters.

AI doesn't need full autonomy to be useful

There's a tendency to assume an AI agent becomes more sophisticated as more human involvement is removed.

That's not necessarily true.

A useful progression is:

A meeting-preparation agent may never need to progress beyond READ + PREPARE.

A routine CRM workflow may eventually be allowed to make certain changes automatically.

A pricing exception may always go to a person.

Different jobs require different levels of authority.

The objective isn't maximum autonomy. It's useful autonomy.

What should you not use AI for in sales?

There isn't one universal list.

But be considerably more cautious when the work involves sensitive conversations, unusual commercial decisions, contractual commitments, important relationships, high-consequence actions or information the AI cannot reliably access.

And don't automate something simply because the existing process is frustrating.

Sometimes the correct order is:

If a process contains three unnecessary steps, don't proudly build an AI agent to perform all three faster.

Remove them.

Will AI replace salespeople?

It's more useful to look at the work than the job title.

Selling contains research, administration, communication, coordination, judgement, negotiation, relationship-building and commercial decision-making.

Those activities aren't equally suited to AI.

Some are already becoming easier to automate.

Some can be heavily assisted.

Some still depend greatly on human judgement and relationships.

For many businesses, the immediate opportunity is therefore less dramatic than replacing a sales team.

It's:

Remove more of the work surrounding selling.

That can still create substantial capacity.

So, how should you use AI in sales?

• Start with a real sales problem.
• Find the repeated work.
• Understand how it happens now.
• Remove anything unnecessary.
• Use ordinary automation where the rules are predictable.
• Use AI where interpretation or context adds something useful.
• Give it access only to the information the job requires.
• Set clear boundaries around what it can do.
• Keep people involved where judgement matters.
• Then measure whether the process actually got better.

Because the goal isn't:

We use AI in sales.

It's:

Our sales process works better than it did before.

Quick answers

Want to see what this looks like in practice?

AI becomes much easier to understand when you stop thinking about the technology and look at the workflow.

See examples of AI handling enquiries, qualification, follow-up, meeting preparation, proposals, CRM administration and pipeline attention.

Or start with the bigger idea: