Will AI replace salespeople? IMF: 40% of jobs are exposed

AI is not going to replace salespeople en masse within the horizon the research can see, and the role losing the most ground is selling from a script, as in outbound telemarketing. The most quoted number comes from the IMF: almost 40% of global employment is exposed to AI, about 60% in advanced economies and 41% in Brazil. Exposure measures how many of an occupation’s tasks the technology can reach, without predicting layoffs, and the IMF itself estimates that in advanced economies about half of the exposed jobs stand to gain productivity from AI. The ILO, which measures generative AI only, finds one in four jobs with some exposure and 3.3% in the highest automation-risk band, where contact center salespeople appear. Commercial sales representatives sit in the middle band, with exposed tasks next to others that depend on judgment and relationships.

At Iasbeck&Co we implement CRM and AI agents in B2B operations, and I teach graduate courses for revenue professionals. The question in the title comes up in almost every class, asked by people afraid of losing their jobs or by people deciding how many to hire next year. I wrote this for sales, marketing and customer service leaders, for the board and for HR, covering what six primary studies say and how I read them for a commercial operation.

What does AI exposure mean, and why is it different from replacement?

The ILO and the IMF break each occupation into tasks and estimate how much of them an AI model could perform or speed up. The calculation says what AI can do in theory and leaves out whether the company will adopt it, what it costs, whether the customer accepts talking to a machine and what the company does with the hours freed up.

The distinction that matters to leaders separates automation from augmentation. When almost all of an occupation’s tasks are exposed, AI can take over the whole job, and the risk is automation. When the occupation mixes exposed tasks with tasks that depend on a person, AI takes part of the work and the person gains time, which researchers call augmentation or complementarity. The IMF gives the example of judges, highly exposed because they read and write text all day, and shielded because society does not accept delegating a ruling to unsupervised AI. The account executive who drafts a proposal with AI help and answers for the price offered is in the same situation.

How the ILO grades exposure

The ILO’s 2023 study, by Pawel Gmyrek, Janine Berg and David Bescond, used GPT-4 to score the tasks of the international occupational classification ISCO-08 and concluded that the main effect of generative AI would be to augment occupations, with automation concentrated in clerical work and reaching 5.5% of employment in high-income countries. The May 2025 update redid the scores with a sample of the 29,753 tasks in the Polish occupational classification, a survey of 1,640 workers in Poland, international experts and more recent AI models. The split between automation and augmentation became a scale of four gradients, from 1, with low exposure and many human tasks, to 4, where almost every task is exposed.

Worldwide, 24% of employment has some exposure, and 3.3% is in gradient 4, weighing more on female employment, which is more concentrated in clerical roles. In high-income countries exposure rises to 34% of employment, against 11% in low-income countries. For the authors, since most occupations have tasks that require human input, transformation of work is the most likely effect.

Why does the IMF say 40% and the ILO 24%?

The two numbers measure different things. The IMF Staff Discussion Note of January 2024, by Mauro Cazzaniga and six other economists, uses an index of exposure to AI in general that matches AI applications against the abilities each occupation requires. The IMF defines high exposure using the index median, and the note itself warns that the measure is relative: it says whether an occupation is more or less exposed than others. The ILO scores, task by task, what generative AI can do and only counts as exposed an occupation that passes an absolute threshold.

The IMF adds a complementarity index, which considers the social, ethical and physical context of the work. In the average advanced economy, 27% of employment is in high-exposure, high-complementarity occupations, where AI tends to support the person, and 33% in high-exposure, low-complementarity ones, where the chance of AI replacing human tasks is greater. The example the IMF picks for this second group is telemarketers. In Brazil, 41% of employment is in high-exposure occupations.

The article GPTs are GPTs, published in Science in June 2024 by Tyna Eloundou, Sam Manning, Pamela Mishkin and Daniel Rock, shows how much the estimate depends on software. With the language model used through a simple interface, about 1.8% of US jobs would have more than half of their tasks affected. When the authors include software that complements the model, the category that covers agents connected to company systems, the share rises to just over 46%. In a commercial operation, the effect on the team depends less on the chat window open in the browser and more on the agent connected to the CRM, which reads the history and carries out the task.

Where do sales, marketing and customer service sit in the ILO index?

The appendix of the 2025 study lists each occupation’s mean score, from 0 to 1, and its gradient. I gathered the revenue occupations below. The international classification has no code for SDR, account executive or CS analyst, and each company needs to map its own roles to these descriptions.

Occupation (ISCO-08)	Mean exposure score	ILO 2025 gradient
Contact centre salespersons (5244)	0.61	4, the highest
Contact centre information clerks (4222)	0.58	3
Inquiry clerks (4225)	0.57	3
Advertising and marketing professionals (2431)	0.55	3
ICT sales professionals (2434)	0.51	3
Technical and medical sales professionals (2433)	0.50	3
Commercial sales representatives (3322)	0.49	2
Sales and marketing managers (1221)	0.41	2
Shop sales assistants (5223)	0.38	1

Source: ILO, Working Paper 140, 2025, appendix table A1.

Telemarketing and scripted sales

The contact center salesperson is the only sales occupation in gradient 4, next to data entry clerks, bookkeeping clerks and general office clerks. The work already comes broken into a script, with a standard pitch, expected objections and system logging, and each of those steps fits a voice or messaging agent. The IMF reaches the same occupation through a different method and the World Economic Forum lists it among the fastest declining through 2030, a strong signal to plan the transition of the people in it.

Customer service and contact centers

Customer service roles sit in gradient 3, with significant exposure and some variation across tasks. Answering frequent questions and classifying tickets are exposed; calming an angry customer and deciding an exception depend on someone with context and authority. How to start with triage and drafts, with a written rule for handing the case to a human, is in AI in customer service.

Sales representatives and technical sales

The commercial sales representative is in gradient 2, with moderate exposure and uneven tasks. Researching the account, preparing the meeting and drafting the follow-up are exposed. Running a needs investigation with the decision maker, negotiating terms and sustaining the relationship after the contract stay with the person, and they are the core of consultative selling. Technical and ICT sellers sit in gradient 3, which fits work heavy on documentation and proposals.

Marketing

The advertising and marketing professional scores 0.55, among the highest in the revenue group, reflecting the weight of copy, assets, segmentation and campaign analysis in the job. For marketing leaders, production volume will depend less on team size, and the judgment of whoever approves each piece gains weight, because a smaller team publishes more and errs at scale if nobody checks.

What does the World Economic Forum project for sales and service through 2030?

The Future of Jobs Report 2025 surveyed, in late 2024, more than 1,000 employers representing more than 14 million workers in 55 economies. These employers project 170 million jobs created and 92 million displaced between 2025 and 2030, a net gain of 78 million, or 7% of the formal employment analyzed. AI accounts for part of this change, alongside cost of living, demographics and the green transition.

The fastest declining occupations in relative terms include telemarketers, door-to-door sales workers and cashiers. Client information and customer service workers appear among the largest declines in absolute numbers. Frontline salespersons are among the occupations growing most in volume, and the report points to slower economic growth as a driver of growth for business development professionals and sales representatives. The same slowdown appears among the causes of decline for sales and marketing professionals, a sign that the report mixes technology and the economic cycle in the same projection.

Among employers, 85% plan to prioritize upskilling, 70% expect to hire people with new skills and 40% plan to reduce staff whose skills are losing relevance; in the AI chapter, 41% expect to downsize as AI becomes able to replicate roles. The report estimates that 39% of a worker’s skills will change by 2030, down from 44% in the 2023 edition.

Is AI already taking jobs from people who sell and serve?

The most recent study with payroll data is Canaries in the Coal Mine, by Erik Brynjolfsson, Bharat Chandar and Ruyu Chen, from the Stanford Digital Economy Lab, revised in August 2026, using data from payroll processor ADP on millions of US workers through June 2026. The authors combine the exposure measure from Eloundou and coauthors with the index of real AI usage published by Anthropic. The most exposed quintile includes customer service representatives, wholesale and manufacturing sales representatives and telemarketers.

The authors find no widespread displacement: employment in the most exposed quintile grew about 4% since late 2022, against 6% for the sample. Employment of workers aged 22 to 25 in exposed occupations stands 19% below where it would be had it kept pace with less exposed peers, and the gap comes mostly from fewer hires. The decline concentrates where AI use substitutes for tasks; where it complements the work, employment is flat or rising. In a commercial operation, the adjustment shows up first at the career entry point, with the SDR, the junior support agent and the marketing assistant.

Summary of the research on AI and jobs

Source	Year	Sample and method	Main finding
ILO, Gmyrek, Berg and Bescond	2023	ISCO-08 tasks scored by GPT-4, employment in 189 countries	Main effect is augmentation; automation in 5.5% of employment in rich countries
ILO, Working Paper 140	2025	Polish classification tasks, 1,640 workers, experts and AI	24% of employment exposed, 3.3% in the highest gradient
IMF, Cazzaniga and others	2024	Exposure and complementarity indices, 142 countries	Almost 40% of employment exposed, 60% in advanced economies, 41% in Brazil
World Economic Forum, Future of Jobs	2025	Over 1,000 employers, 14 million workers, 55 economies	170 million jobs created and 92 million displaced by 2030
Eloundou, Manning, Mishkin and Rock, Science	2024	US O*NET tasks rated by humans and GPT-4	1.8% of jobs with half their tasks affected, 46% with software
Brynjolfsson, Chandar and Chen, Stanford	2026	ADP payroll data, US, through June 2026	No broad displacement; young workers in exposed jobs 19% below trend

What these data do not say

None of these studies measures B2B sales in Brazil. International classifications put the software seller with a six-month cycle and the representative who visits pharmacies under the same code, and each occupation’s score is an average over tasks that vary from company to company. The ILO’s 2025 update started from Polish tasks and a Polish survey, and the authors complemented the scores with international experts and AI models. The IMF’s 41% for Brazil depends on the cut at the index median and does not indicate how many Brazilian jobs will disappear.

Exposure measures technical potential, without the cost of deploying an agent or the quality of the data it will read. The Science estimate goes from 1.8% to over 46% depending on the software built on top of the model. Two of the article’s four authors were OpenAI researchers, a company that develops the models evaluated, which does not invalidate the method and belongs in the reading.

The World Economic Forum publishes expectations from large global employers, and stated expectations tend to misjudge the pace of change. The Stanford study is the only one with observed employment data, and the authors treat it as a descriptive indicator, not a causal estimate, limited to the United States and with a smaller effect when controlling for education. The line that AI will not take your job but someone using AI will circulates in talks and posts without any study behind it; it may describe the direction of the research, and nobody has measured it.

What leaders do with these numbers

The studies help decide about tasks, and the role comes after. I use this roadmap with sales and service directors who need to answer the board and the team about AI’s effect on headcount.

1. List the tasks of each revenue role. From the SDR to the CS analyst, write down the week’s tasks and the time each takes, using calendars and the CRM.
2. Classify each task using the ILO logic. Mark what AI can already perform with human review, what it speeds up and what depends on judgment, relationships or decision authority.
3. Separate roles at risk of automation from roles that gain from augmentation. A role with almost every task exposed, such as scripted telemarketing, needs a transition plan with a deadline; a role with mixed tasks needs its work reorganized.
4. Decide in advance what to do with the freed hours. More accounts per executive, prospecting a new segment or a postponed hire are leadership choices, and hours nobody reallocates vanish into the routine. Measuring that gain is covered in generative AI and productivity.
5. Redesign the entry point. If AI takes over account research and first drafts, juniors need to learn earlier how to lead the conversation and qualify, and sales onboarding changes with it.
6. Review the map every six months. The ILO scores changed between 2023 and 2025, and agents connected to the CRM move faster than the indices.

What HR and L&D need to consider about reskilling

According to the World Economic Forum, in a group of 100 workers, 59 will need training by 2030, 29 should be upskilled in their current role, 19 should be reskilled and redeployed within the organization and 11 are unlikely to receive the training they need. For HR, the 19 redeployed weigh the most: people leaving outbound telemarketing or the first-level queue need a designed path into roles such as SDR, CS or agent operations. The IMF shows, with data from Brazil and the UK, that college-educated workers move more easily into occupations complementary to AI, and that older workers tend to struggle more.

The skills gaining value, according to the Forum’s employers, combine technology and human relationships: AI and big data top the list of fastest growing, followed by networks and cybersecurity and technological literacy. Among those that gained relevance since 2023 are leadership and social influence, talent management and service orientation and customer service, and empathy and active listening appear in the list of core skills. In revenue teams, I translate this into training to use and check AI in each role’s tasks and to lead the conversations AI does not lead, such as discovery, negotiation and renewal. How to build that training by role is in AI training for companies.

AI changes tasks and the entry point first, and the effect on headcount depends on decisions leaders and HR make now. The in-company applied AI training works through this map with sales, marketing and service teams on the company’s own processes, and delivers the prompt library by area and the agent architecture, covering what each agent reads, what it executes on its own and who approves the rest.

Frequently asked questions

Will AI replace salespeople?

The research does not point to mass replacement of salespeople through 2030. Automation risk concentrates in telemarketing and scripted selling, which the ILO places in the highest exposure gradient. Commercial sales representatives sit in the moderate band, with part of their tasks automated and negotiation and relationship work still with the person.

What percentage of jobs is exposed to AI?

The IMF estimates that almost 40% of global employment is exposed to AI, 60% in advanced economies and 41% in Brazil, using a relative measure of exposure. The ILO, which measures generative AI only and task by task, finds 24% of global employment with some exposure and 3.3% in the highest automation-risk band.

What is the difference between AI exposure and replacement?

Exposure measures how many of an occupation’s tasks AI can perform or speed up. Replacement depends on adoption, cost, regulation and the company’s decision about what to do with the freed hours. A highly exposed occupation can gain productivity, as the IMF estimates for about half of exposed jobs in advanced economies.

Will customer service be handled by AI alone?

Customer service roles are in the ILO’s gradient 3, with significant exposure, and the World Economic Forum projects a decline in absolute numbers through 2030. Cases that require judgment, such as an angry customer, a commercial exception or churn risk, stay with people, and the operation needs a written rule for passing the case from AI to a human.

How do you prepare the sales and service team for AI?

Start by mapping each role’s tasks and deciding what AI performs, what it speeds up and what stays with the person. Training covers using AI in the role’s tasks, checking what it produces and the conversations that remain human, such as discovery and negotiation. For people in roles at risk of automation, HR needs to design a redeployment path with a deadline.

Which sales and marketing jobs are growing with AI?

The World Economic Forum projects growth for frontline salespersons, business development professionals and sales representatives, and decline for telemarketing and door-to-door sales. Among skills, AI and big data lead growth, with leadership, social influence and customer orientation gaining weight.