What AI changes about how you sell, and what stays the same

Think about your best salesperson's Tuesday. They log yesterday's calls, update the CRM, rewrite the same follow-up email 4 times, chase 3 people who never replied, and fit in maybe 2 real conversations before lunch.

Salesforce's State of Sales survey puts a number on it. Sellers say 60% of their time goes to work other than selling.

AI is taking that part first. Here's what changes after that, inside the conversation itself, and what your buyer still needs from you.

TL;DR

• Sellers say 60% of their time goes to work other than selling (Salesforce State of Sales, 7th edition).
• AI takes that admin first: logging calls, drafting emails, booking meetings, chasing replies.
• The bigger change is inside the conversation, where AI now makes the judgment call: which objection is real, what to say next and when to follow up.
• Your buyer still needs the same things: to feel understood, to get a straight answer and to decide in their own time.
• Your customers are starting to shop with AI agents too, so your side of the conversation has to answer them well.

Where does a salesperson's week go?

Most of your team's week goes to work around the sale. In Salesforce's State of Sales report, 7th edition, 4,050 sales professionals in 22 countries said they spend 40% of their time selling and 60% on everything else. The survey ran in August and September 2025.

That "everything else" is the stuff you'd expect: data entry, research, hunting for the right deck, switching between tools. In the previous edition (July 2024) the non-selling share was 70%, so it's moving, slowly.

Picture a team of 10. At 40% selling time, you're paying for 10 people and getting 4 people's worth of conversations. (Example math, using Salesforce's average.)

What does AI take over first?

The admin. It's repetitive, easy to check, and your team won't miss doing it.

• Logging and notes. The call summary writes itself and lands in the CRM.
• Drafting. First drafts of follow-ups, quotes and recap emails.
• Scheduling. Finding a time and sending the invite.
• Chasing. Reminding people who said "send me something" and then didn't reply.

This part is useful, and it's where most AI in sales stops. It handles the tasks around the sale and leaves the decisions inside it to your team.

Can AI persuade as well as a person?

On written arguments, roughly yes. In April 2024, Anthropic measured how persuasive its models were and found that Claude 3 Opus wrote arguments with no statistically significant difference in persuasiveness from arguments written by people.

People still wrote the single most persuasive arguments in that study. Even so, a model that persuades about as well as a person, at any hour, changes what you can do in a sales conversation.

Why does a general AI assistant make a weak salesperson?

Because it's built to please whoever it's talking to, including your buyer. In October 2023, Anthropic's researchers studied sycophancy, the habit of telling people what they want to hear, and found it's a general behavior in AI assistants trained on human feedback.

In a sale, that habit shows up fast. Your buyer says "that's too expensive" and a general assistant agrees, apologizes, or offers a discount you never approved.

A good salesperson works out whether price is the real objection, then answers that.

What does a model built for selling do differently?

It makes the judgment call, inside your rules. Here's what that means in practice, in the 4 steps our persuasion model follows:

1. Reads the buyer. It works out what's holding them back: price, timing, still comparing, or a question they haven't asked yet.
2. Picks the move. It chooses from 420 persuasion moves, such as a price reframe, a coverage comparison or a straight answer, based on this buyer and this moment.
3. Says it right. It writes in your approved wording, so every rate, fee or feature comes from copy you signed off.
4. Remembers. The conversation keeps going for weeks, and day 12 picks up where day 1 left off.

Sellence's model is trained on 3.5M+ real sales conversations. That's where it learns which move tends to work on a buyer like yours.

What does that look like in a real sale?

Libra Insurance tested it on its own sales floor. Two teams of comparable size worked the same product, pricing, segments and leads for 90 days, and only one of them used Sellence.

Libra's team using Sellence reached 3x the sales per agent of the control team, with the same headcount. It also started the test behind.

What stays the same?

What your buyer needs. Someone renewing their car insurance with a cheaper quote in another tab wants 3 things: to feel understood, to get a straight answer about why your price is higher, and to decide in their own time.

Those needs haven't moved. The difference is that every buyer can now get that kind of attention, including the ones your team never had time to call back.

What changes next?

Your buyers are getting their own AI. In Experian's 2026 survey of 6,247 consumers in Europe and Asia-Pacific, 82% said they trust AI to compare loans across providers.

That means more of your conversations will be with an agent that compares you against everyone else at once, and you'll need an answer ready for it. We cover what that looks like in the future of agent-to-agent sales.

Key takeaways

• Most of a seller's week isn't selling. Sellers say 60% of their time goes to other work (Salesforce, 7th edition).
• AI takes the admin first. Logging, drafting, scheduling and chasing are the easy wins.
• The bigger change is the judgment call. A model built for selling reads the buyer, picks the move and writes in your approved wording.
• A general assistant gives in too easily. It's built to please, so it agrees with "too expensive" instead of finding the real objection.
• The buyer's needs stay the same. They want to feel understood, get a straight answer and decide in their own time.

Frequently asked questions

How is AI changing B2C sales?

It takes over the admin that fills most of a salesperson's week, and it now makes judgment calls inside the sales conversation: which objection is real, what to say next and when to follow up.

Will AI replace salespeople?

AI already does most of the admin, and models built for selling now handle the conversation itself. At Libra Insurance, the team using Sellence reached 3x sales per agent in a controlled 90-day test, with the same headcount.

Can AI be as persuasive as a person?

On written arguments, roughly. Anthropic's 2024 study found Claude 3 Opus wrote arguments with no statistically significant difference in persuasiveness from human-written ones, though people still wrote the single most persuasive arguments.

Why can't you sell with a general AI assistant?

General assistants tend to tell people what they want to hear, which Anthropic's researchers call sycophancy. In a sale, that means agreeing with a price objection instead of answering it.

What doesn't AI change in sales?

What the buyer needs: to feel understood, to get a straight answer to their objection and to decide in their own time.

Sources

1. Salesforce, "State of Sales," 7th edition (anonymous survey of 4,050 sales professionals in 22 countries, August to September 2025). https://www.salesforce.com/en-us/wp-content/uploads/sites/4/documents/reports/sales/salesforce-state-of-sales-report-2026.pdf
2. Salesforce, "Sales AI statistics 2024" (6th edition, 5,500 sales professionals in 27 countries), July 25, 2024. https://www.salesforce.com/news/stories/sales-ai-statistics-2024/
3. Anthropic, "Measuring the Persuasiveness of Language Models," April 9, 2024. https://www.anthropic.com/research/measuring-model-persuasiveness
4. Anthropic, "Towards Understanding Sycophancy in Language Models," October 23, 2023. https://www.anthropic.com/research/towards-understanding-sycophancy-in-language-models
5. Experian, "More than half of consumers are comfortable with AI agents applying for credit," September 16, 2026. https://www.experianplc.com/newsroom/press-releases/2026/consumers-comfortable-ai-agents-applying-for-credit