AI in sales
AI is probably the most useful sales tool I’ve had access to in my career.
It can research an account in minutes, summarize months of calls, spot patterns across opportunities, help prepare for a CFO meeting, organize CRM notes, pressure-test a business case and remind you that somebody mentioned something important eleven weeks ago that everyone else completely forgot about.
That’s incredibly valuable.
It can also write an outbound email that sounds exactly like the other 400 AI-generated outbound emails sitting unread in someone’s inbox.
So there’s a balance.
I use AI constantly in sales. I also think relying on it too heavily can actively make a salesperson worse.
The rule I’ve landed on is pretty simple:
AI should give a salesperson more time and information to be human, not less reason to develop human skills.
Pre-call preparation is probably the clearest place where AI makes salespeople better.
Before a meeting, I want to understand the company, the industry, the person I’m talking to, previous conversations, likely business problems and anything else that helps me walk in with a useful point of view.
AI compresses that work dramatically.
If I’m meeting a CFO, I can think through what that person may care about differently than the VP of Sales.
If there are previous call transcripts, I can have AI review them before the next meeting.
If an executive publicly writes about their priorities or industry, I can synthesize that quickly.
The point isn’t to create some creepy digital clone of the person.
It’s preparation.
If someone has publicly told the world what they care about and I’m about to ask them for an hour of their time, reading it seems reasonable.
One of the best ways I’ve used AI in larger enterprise deals is as an external memory.
Complex opportunities can stretch across months and involve a lot of stakeholders.
Someone might mention something important in March that never comes up again until June.
I probably noticed it at the time.
Three months and fifty other sales conversations later, there’s a decent chance my brain has filed that information somewhere next to my childhood AOL password.
AI doesn’t have that problem.
It can surface things like:
“This stakeholder mentioned implementation risk two months ago and it hasn’t been addressed since.”
Or:
“The CFO has brought up budget structure multiple times.”
Or:
“This person cared heavily about an upcoming trade show, but that compelling event has disappeared from recent conversations.”
That can absolutely help save a deal.
But there’s an important distinction:
AI is very good at remembering everything. The salesperson still has to decide what matters.
AI is also useful when a deal starts slowing down.
It can review the account history and suggest that maybe the real issue is budget, implementation, internal consensus, timing or something else.
That’s useful.
But it’s still a hypothesis.
A good salesperson eventually needs to uncover the actual answer.
If AI thinks “timing” really means budget, great. Now go find out.
There’s a dangerous version of selling where the rep starts trusting the AI’s interpretation of the customer more than the customer.
I would avoid that.
Another way I use AI is as a sparring partner.
Usually I have the instinct first.
Maybe a deal is stuck because the scope is above the approved budget.
I might think:
“What if we stripped this down to an MVP, removed the things they don’t need right now and gave them three options that preserve the important value while getting closer to budget?”
Then I’ll run that through AI.
Review the transcripts.
What am I missing?
How might the customer react?
What are the risks?
How would you structure the options?
Now AI is helping me test the strategy.
It isn’t coming up with the entire strategy while I stare at the screen waiting for instructions.
That distinction matters.
This is where things start getting ugly.
AI is very good at producing average writing.
The problem is that everyone else has access to the same average.
Ask:
“Write me a good outbound sales email.”
So does everybody else.
Now the buyer’s inbox contains:
“Hey Alex, I noticed the innovative work you’re doing at…”
“I came across your website and was impressed by…”
“As a leader in the e-commerce space, you’re likely balancing growth with operational efficiency…”
Technically fine.
Spiritually dead.
There’s a point where the language becomes so polished and middle-of-the-road that the personalization itself starts proving nobody personalized it.
If the goal of personalization is to make someone feel like a person wrote specifically to them, sending something that immediately smells automated defeats the entire purpose.
For customer-facing writing, my preferred process is:
Human creates. AI improves. Human decides.
I’ll write the email first.
Then I might ask AI:
Is this too long?
Is anything unclear?
Can this sentence be tighter?
Am I missing a point?
Does the tone feel off?
Then I take what’s useful and make the final call myself.
Once you have enough examples of your own writing, AI becomes even more useful because you can train it around your actual style.
Now you’re not asking:
“What does a good salesperson sound like?”
You’re asking:
“How would I communicate this based on the way I already communicate?”
That’s much better.
This is where I draw the strongest line.
I don’t think AI should generally be driving a live sales conversation.
There may be exceptions for training or a new rep who needs help while ramping.
But as a permanent selling model?
No.
If somebody is talking and the rep is waiting for AI to tell them what to say next, they aren’t fully listening.
They’re waiting for instructions.
And that creates a shortcut around the part of becoming good at sales that unfortunately requires actually becoming good at sales.
Reps.
Conversations.
Mistakes.
Pattern recognition.
Learning when someone is hesitant.
Learning when to push.
Learning when to back off.
Learning when the question you planned to ask no longer matters because the buyer just said something much more interesting.
AI can help you review those moments afterward.
It shouldn’t replace developing the instinct to handle them live.
AI can be great for objection training.
Give it transcripts.
Ask it what patterns it sees.
Role-play.
Have it challenge your responses.
That’s useful.
What I don’t want is a rep hearing:
“We need to think about it.”
and immediately receiving:
RECOMMENDED RESPONSE 4B.
“We need to think about it” could mean price.
Implementation.
Another vendor.
Internal politics.
No urgency.
Or simply that they would like to leave the Zoom call without hurting your feelings.
Those are different conversations.
A good salesperson learns how to uncover the objection behind the objection.
AI can review the film afterward.
It shouldn’t play quarterback on every snap.
Once the conversation ends, AI becomes incredibly useful again.
Summarize the meeting.
Update CRM notes.
Compare the call to previous conversations.
Look for changes in priorities.
Review objections.
Identify unanswered questions.
Help prepare the next meeting.
That’s exactly the type of work I want AI absorbing.
CRM documentation is a good example.
Detailed account history is valuable, especially when opportunities last months or accounts change hands.
The problem is that most salespeople did not get into sales because they dreamed of spending Friday afternoon typing detailed CRM notes.
AI can capture much more context without creating nearly as much administrative work.
That’s a good trade.
The part I worry about most isn’t AI replacing salespeople.
It’s AI making mediocre salespeople LOOK more competent than they are.
Polished emails.
Perfect CRM notes.
Detailed account summaries.
Suggested discovery questions.
Suggested objection responses.
Every box checked.
Every dashboard green.
Revenue somehow confused.
A junior rep can repeat sophisticated language about an account because AI prepared it for them.
That doesn’t necessarily mean they understand the account.
And eventually something happens that the automation didn’t prepare for.
The buyer pushes back.
The messaging stops working.
The rep moves into larger deals.
The conversation goes somewhere unexpected.
If the salesperson never learned how to think critically, adapt or communicate in their own voice, eventually that gap shows up.
AI can standardize best practices.
It can also standardize mediocrity.
AI can help me know more about the customer.
It can help me prepare.
It can help me remember.
It can help me communicate more clearly.
But it can’t build the relationship for me.
If the buyer wanted nothing more than information, they could probably get most of it from AI themselves.
The salesperson has to bring something else.
Judgment.
Perspective.
Credibility.
Humor.
Curiosity.
Empathy.
Confidence.
Maybe even vibes, as scientifically rigorous as that sounds.
There are people you enjoy talking to and people whose name appears on your phone and you immediately consider throwing the device into traffic.
You do not want to become the second one.
The way I think about AI in sales is:
Use it aggressively.
Research. Prepare. Review history. Generate hypotheses. Rehearse. Pressure-test ideas.
Mostly put it away.
Listen. Think. Ask questions. Read the room. Be a person.
Turn it back on.
Analyze. Summarize. Document. Coach. Look for patterns. Prepare the next step.
And for customer-facing writing:
Human creates. AI improves. Human decides.
The tools will keep getting better.
The underlying principle probably won’t change.
The goal of AI shouldn’t be to create salespeople who no longer need to know how to sell.
It should remove enough research, administrative work and information overload that salespeople have more capacity to do the parts of the job humans are actually good at.
Think.
Listen.
Adapt.
Build trust.
Make judgment calls.
Have real conversations.
AI should give a salesperson more time and information to be human. It should never give them less reason to become one.