Will AI Replace Sales Reps? What the Data Actually Shows

No, not the whole job, and not on current evidence. The US Bureau of Labor Statistics projects sales and related jobs to shrink 1.4% from 2025 to 2035 while total employment grows 3.5%. The steepest fall is in scripted work: telemarketers, down 21.4%. Sales managers grow 4%. AI is taking sales tasks, not closers.

• What has moved to AI: answering general product questions, scripted outbound calls on a known offer, quoting and order fulfilment, first-touch prospecting.
• What has not moved: judging whether a product fits the buyer’s situation, and closing complex or high-stakes deals.
• Jobs data: BLS Employment Projections 2025–2035 (released 27 August 2026) names “incorporation of AI tools into the sales process” as one cause of the 1.4% decline.
• The decision rule: the three-S test. A sales task that is Scripted, Small-stakes and Self-serve is moving to software now. A role with none of the three is growing.
• What changes the answer: whether the buyer knows it is AI, how much is at stake, and whether buyer preference keeps moving toward rep-free buying.

What does “replaced by AI” actually mean for a sales job?

A sales job is replaced by AI when the work it paid for is no longer done by a person at all. That is a different thing from a job being reshaped, where AI does some tasks and the person does the rest with less time on admin. Most of the evidence below describes reshaping. Real replacement shows up in only one place: occupations whose entire content was a single scripted task.

The boundary matters because the question is usually asked about the wrong unit. “Sales rep” is a bundle of tasks. In Salesforce’s State of Sales, 7th edition (4,050 sales professionals in 22 countries, surveyed August–September 2025), respondents said reps spend 40% of an average week selling and 60% not selling, on work such as data entry and prospecting. AI is eating into the 60%. A rep whose job is mostly the 40% is being given more time to sell. A rep whose job is mostly the 60% is being replaced.

Which parts of the sales job have already moved to AI?

This is the boundary, task by task, with the evidence for each row. “Moved” means buyers or firms already route the task away from a person. “Not moved” means the evidence points the other way.

Sources: Gartner, 25 June 2025; Luo, Tong, Fang and Qu (2019); Salesforce State of Sales, 7th ed.; Gartner, 25 August 2025.

The line runs between the top and the bottom of the funnel. Everything above the booked conversation is moving. Everything from “does this fit us?” onward is not. The hybrid AI and human SDR model is built on exactly that split, and our plain-English guide to AI sales agents covers what the software half does.

What do the jobs projections say about sales roles to 2035?

BLS projections are a neutral, occupation-level forecast that states its method. The table shows the 2025–35 projection for each sales occupation, the reason BLS gives, and our three-S reading (explained in the next section). The three-S column is our interpretation, not BLS’s.

Sources: BLS fastest-declining occupations table (telemarketers, 2025 and projected 2035); BLS Occupational Outlook Handbook pages for advertising sales agents, cashiers, wholesale and manufacturing reps, retail sales workers, financial services sales agents, insurance agents, sales engineers and sales managers.

Three cautions. Telemarketing was shrinking long before generative AI: a 2020 BLS analysis already projected a 16.6% fall for 2018–28 and blamed caller ID and the Do Not Call Registry. Advertising sales is the next-steepest decline and BLS puts it down mainly to falling print circulation, not AI, so not every shrinking sales job is an AI story. And BLS says in the release itself that it models technological progress “in line with historical patterns” and that uncertainty about AI’s impact “remains very high”. Read the BLS projection for sales as the no-surprises case, not a ceiling on change.

The three-S test: will my sales role be replaced by AI?

Score each task in the role, then the role as a whole, against three questions. Each yes is one point.

1. Scripted. Could the task be done from a script and a decision tree, without changing course on what the buyer says?
2. Small-stakes. If the answer is slightly wrong, is the cost to the buyer small?
3. Self-serve. Would the buyer happily complete this step without a person if the tool were good enough?

An SDR who spends the week on first-touch messages and calendar booking scores 3 on most of their tasks. An account executive who runs discovery on a multi-stakeholder deal scores 0 on the work that earns the commission. The three-S test says AI replaces the SDR task list well before it replaces the closer. The test is a heuristic for sorting tasks, not a forecast for any one person’s job.

When does the answer change?

Four conditions would move the line, and the evidence on each one is mixed.

• The buyer knows it is AI. In the Luo field experiment, telling customers up front that the caller was a bot cut the purchase rate from 23.7% to 4.8%, a fall of 79.7%. Late disclosure and buyers’ prior AI experience reduced the drop. The parity result applies only to undisclosed bots. If you disclose, the number that applies to you is the 79.7% drop, which makes the headline finding (“AI sells as well as a good rep”) the one you are least able to use.
• Buyer preference keeps moving. In Gartner’s own surveys, the share of B2B buyers who prefer a rep-free experience rose from 61% (632 buyers, 2024) to 67% (646 buyers, 2025). Yet Gartner also predicts that 75% will prefer human-first sales by 2030. These are a survey and a forecast, and they point different ways. If the survey trend wins, more of the fit-judging row moves.
• The stakes fall. A task scores Small-stakes when it becomes cheap to reverse: trials, monthly terms, easy returns. Lower the cost of a wrong purchase and a closer’s job starts to score like a cashier’s.
• AI stops raising output. Gartner predicts that AI agents will outnumber sellers ten to one by 2028, but fewer than 40% of sellers will say the agents improved their productivity. Agents that do not raise output do not reduce headcount.

Fewer reps or no reps? The capacity arithmetic

The realistic near-term outcome is fewer reps per unit of revenue, not zero reps. Here is the arithmetic, using Salesforce’s 40/60 split and an assumption you should replace with your own.

• A team of 10 reps selling 40% of the time delivers 4.0 full-time-equivalents of selling.
• Assume AI absorbs half the non-selling 60%. Each rep now sells 70% of the time (40 + 30).
• The same 4.0 FTE of selling now needs 4.0 ÷ 0.70 = 5.7 reps.
• Result: about 43% fewer seats for the same selling hours, or the same 10 seats with 1.75x the selling time.

That only holds if the freed hours actually go into selling, which has to be managed rather than assumed. Gartner’s analysts describe the same direction: “fewer but more specialized sales roles”. Running the AI half in-house takes an owner for prompts and sequences, calendar and CRM integration, deliverability and consent management, and someone to read the conversations every week. The AI SDR vs human SDR cost breakdown prices those pieces.

The split we run is the one the evidence supports: the AI works the scripted top of the funnel and a person closes. LeadsNow’s AI has booked 50,769+ sales appointments since 2017, and in that model the agent’s job ends at the booked appointment, where the client’s own closer takes over. We feed closers; we do not replace them. How that is set up for a larger sales organisation is on the corporate AI appointment setting page, and the full range of agent types is on the AI sales agents hub.

Frequently asked questions

Will AI replace SDRs before it replaces account executives?

Yes, on current evidence. SDR work is mostly first touch, follow-up and booking, which score high on the three-S test. In Salesforce’s 2025 survey, 34% of sales teams using AI agents already use them for prospecting. Discovery and closing have not moved.

How many sales jobs will AI eliminate?

Nobody has a reliable count. The best neutral forecast, the BLS 2025–2035 projections, has US sales and related employment falling 1.4% over ten years, with AI named as one cause alongside e-commerce. BLS itself says uncertainty about AI’s impact remains very high.

Can an AI sales call sell as well as a human?

On a scripted offer, yes, but only if the buyer does not know. A field experiment with more than 6,200 customers, published in Marketing Science, found undisclosed AI callers matched proficient workers. Disclosing the bot up front cut purchases by 79.7%.

Do B2B buyers want to buy without a salesperson?

For research, mostly yes. For deciding whether a product fits, mostly no. Gartner found 67% of buyers prefer a rep-free experience, but its 2024 survey showed buyers still want seller input on fit, and Gartner predicts 75% will prefer human-first sales by 2030.

Will my sales job be replaced by AI?

Run the three-S test on your week. If most of your hours go to tasks that are scripted, small-stakes and self-serve, those hours are moving to software. If most go to discovery, fit and negotiation, the BLS data puts you in the occupations still growing: sales managers at +4%, sales engineers at +3%.