Where Should a Sales Team Start Using AI?

Start with the work that is highly repetitive and low in value, and keep the trust-building and closing human. AI fixes nothing unless you tell it exactly what to fix, so be prescriptive. Pick a few small, low-risk tasks, get quick wins, and let confidence build before you hand it anything that touches a customer deal.

How do I decide what to give AI?

I use a four-box grid. One axis is how repetitive the work is. The other is how much value it creates.

• Highly repetitive, low value: give it to AI. Sending connection requests, building a weekly market report, CRM hygiene. Things that happen the same way over and over.
• Low volume, high value: don't touch it. Custom conversations and custom deal quoting are where customers value your people.
• The middle: repetitive but moderate to high value. Weigh the risk of an agent getting it wrong before you automate anything here.
• Low volume, low value, low risk: clear it out of the way too.

Here's a concrete one from my own week. Clicking on 40 people a day to send connection requests is a waste of my time. So an agent looks at the person's profile, likes a post, follows them, and a couple of days later sends the request. I don't think about it. It just happens every day.

The same grid works in a medical practice. Ask the front desk to sort its own tasks into the four boxes, and the low-risk starting points show up fast.

Why does a team need rules before tools?

In most sales organizations there's no guidance. Everybody is just using AI in their own personal way. That's how you get inconsistent messages going to customers. Decide as a team which boxes AI is allowed to touch, who checks the output, and what never goes out without a human. I still write my own outreach templates. I've tested AI that drafts the opening message from someone's website and profile, and I just don't trust it yet.

Where does a sales team get the most time back?

Research and preparation. Reps and managers can spend 20, 30, 40 or 50 hours on things like account research, quarterly business reviews and business plans that repeat every cycle. About 80% of that heavy lifting can be done by automation and agents. A person still has to go in and check it. And honestly, nobody loves building 70 slides.

Call prep is another big one. With the right questions and a clear discovery method, AI can build a call brief for a rep in about three minutes. AI also does follow-up more consistently than people do, once you train it and make it sound human.

What does a good first win look like?

Small and visible. In my own work, an agent reads my call notes at the end of each day and leaves me a to-do list for the morning. It flags emails I haven't answered and drafts replies. I approve, reject or change each one. After three or four weeks you forget you ever did those tasks by hand. That's when people start trusting it with the next thing.

Used the right way, AI should turn one person into 1.5. At a big company that might cost $10,000 to $15,000 per person a year, and return something like $80,000 to $100,000 in benefit. The point is augmentation, not replacement. People still want to buy from people.

Where to start

Get your team in a room for an hour. Have each person list the tasks they repeat every week, then place them on the four-box grid. Pick two from the repetitive, low-value box. Write exact instructions for each, run them for a month with a human check, and measure the hours saved before you move to the next box.