AI in marketing and sales: up to $1.2 trillion (McKinsey)

The most cited research on AI in marketing is McKinsey’s June 2023 report, which estimated the value of generative AI across 63 use cases in 16 business functions and arrived at a potential of $760 billion to $1.2 trillion a year in marketing and sales, the highest range among the functions analyzed (McKinsey, 2023). The number measures potential productivity calculated by a model, without implementation costs. Three years later, Gartner shows budgets chasing the promise: CMOs allocate an average of 15.3% of the marketing budget to AI initiatives, and only 30% say they have mature capabilities to scale those initiatives (Gartner, 2026). Academic experiments with consumers and the rules Google has published complete the picture: readers accept text made with AI, distrust it when they learn an emotional message came from a machine, and the search engine punishes scale without value, whether people or software produced it. For B2B marketing, my conclusion is that AI pays for itself when it grows the pipeline, and none of these studies measures that.

At Iasbeck&Co we implement CRM and AI agents in B2B operations, and in conversations with marketing leaders the McKinsey number tends to show up to justify budget, without the sentence that says what it measures. I checked every number in the original source and organized what the research says for marketing leadership, for sales and CS, and for HR. Day-to-day use of AI by the team is covered in AI in B2B marketing; here the subject is the data.

Which research on AI in marketing is worth reading?

I chose sources that publish their method: an economic estimate from a consulting firm, two editions of Gartner’s annual CMO survey, three peer-reviewed papers and Google’s official statements. The academic papers have smaller samples and experimental designs, which separate cause from correlation.

Source	Year	Sample and method	Main finding
McKinsey, The economic potential of generative AI	2023	Economic model of 63 use cases in 16 functions, built on cost bases per function	$760 billion to $1.2 trillion a year in marketing and sales; 5% to 15% of total marketing spend
Gartner, CMO Spend Survey 2026	2026	401 CMOs in North America, the UK and Europe, January to March 2026, most at companies with revenue above $1 billion	15.3% of the marketing budget goes to AI; only 30% have mature capabilities to scale
Gartner, CMO Spend Survey 2025	2025	402 CMOs in the same regions, February to March 2025	Generative AI returns reported as time efficiency (49%), cost efficiency (40%) and capacity to produce more content (27%)
Zhang and Gosline, Judgment and Decision Making	2023	Experiment with 1,203 participants rating ads and campaign messages written by professionals, by GPT-4 or jointly	Without knowing the source, participants rated AI content higher; knowing a human wrote it raises the rating
Kirk and Givi, Journal of Business Research	2025	Seven preregistered experiments with emotional marketing messages	People who believe an emotional message was written by AI recommend the brand less and show less loyalty
Reisenbichler and coauthors, Marketing Science	2022	Field experiments in two industries with machine-generated SEO texts revised by an editor	The revised texts ranked better in search than those written by experts, at lower cost
Google Search Central, 2023 and Google, 2024	2023 and 2024	Official guidance and the company’s own assessment of the March 2024 update	Appropriate use of AI is not against the guidelines; low-quality, unoriginal content fell 45% in results

How much is generative AI worth in marketing and sales, according to McKinsey?

McKinsey mapped 63 use cases in 16 business functions and estimated between $2.6 trillion and $4.4 trillion a year for the whole set. Four functions account for about 75% of that value: customer operations, marketing and sales, software engineering and research and development. In the exhibit that breaks value down by function, marketing and sales comes first, with a range of $760 billion to $1.2 trillion, and the footnote states that the estimate excludes implementation costs such as training and licenses.

Function	Estimated annual impact ($ billion)
Marketing and sales	760 to 1,200
Software engineering	580 to 1,200
Customer operations	340 to 470
Supply chain and operations	290 to 550
Product R&D	230 to 420
Risk and legal	180 to 260

Source: McKinsey (2023), exhibit on generative AI productivity impact by business function.

Where the marketing number comes from

The report separates the two areas. In marketing, the firm estimates that generative AI could increase productivity by a value between 5% and 15% of total marketing spend; in sales, by around 3% to 5% of global sales expenditure. The examples are messages personalized to customer behavior and first drafts of ads, headlines, slogans, posts and product descriptions. The report also warns of plagiarism and copyright risk.

What the estimate does not say

The 5% to 15% describe how much of current marketing spend a company could save or redeploy with AI doing part of the work. The calculation does not promise new revenue, does not deduct implementation costs and was done before the CRM-connected agents that platforms launched between 2025 and 2026. When someone brings the trillion to a budget meeting, ask how much of your company’s marketing spend goes to text, images and personalization, the base on which McKinsey’s range applies. The reports that measure actual use, such as McKinsey’s own annual survey, are covered in state of AI in companies.

How much do CMOs invest in AI, according to Gartner?

Gartner’s annual CMO survey shows the marketing budget standing still and the AI share growing inside it. In 2026, marketing budgets came to 7.8% of company revenue, against 7.7% in 2025, and 56% of CMOs said they lacked the budget to deliver the year’s strategy. Within that flat budget, 15.3% goes to AI initiatives, and Gartner describes the move as reallocating resources from other lines.

The 2025 edition shows where the money comes from. At the time, 39% of CMOs planned to cut agency budgets, 39% wanted to reduce labor spending, and 22% said generative AI had reduced their reliance on external agencies for creativity and strategy (Gartner, 2025). When Gartner asked where generative AI delivered returns, CMOs cited time efficiency (49%), cost efficiency (40%) and capacity to produce more content or handle more business (27%). None of the three answers measures revenue, opportunities or conversion.

The gap between ambition and readiness

In 2026, 70% of CMOs said becoming an AI leader is a critical goal for the year, and the same 70% admitted their internal processes are not yet mature enough to implement and scale the technology. The group with mature capabilities allocates 21.3% of the budget to AI and has a marketing budget of 8.9% of revenue, above average. The survey shows association: companies with larger budgets can invest more in AI and in process at the same time, without one causing the other. For Gartner, the risk lies in buying tools faster than the company builds the data, processes, governance and people to use them.

Indicator	2025	2026
Marketing budget as % of revenue	7.7%	7.8%
CMOs without enough budget for the strategy	59%	56%
Share of marketing budget allocated to AI	Not reported	15.3%
CMOs with mature capabilities to scale AI	Not reported	30%

Source: Gartner, CMO Spend Survey 2025 and 2026.

Do consumers reject AI-generated content?

The experiments say it depends on what the consumer knows and on the type of message. Yunhao Zhang and Renée Gosline, from MIT Sloan, asked content professionals at a large consulting firm and GPT-4 to write product ads and campaign messages in four arrangements: human only, AI only, a human finalizing from the AI draft and the AI finalizing from the human text (Zhang and Gosline, 2023). Among the 1,203 participants, those who rated the material without knowing the source gave higher scores to content the AI wrote alone or finalized. Revealing the source narrowed the gap without reversing it, and the effect came from the human side: knowing an expert wrote it raised the score, and knowing AI took part did not lower it.

Colleen Kirk and Julian Givi reached a result that looks contrary and measures something else. In seven preregistered experiments, people who believed an emotional marketing message had been written by AI were less willing to recommend the brand and showed less loyalty (Kirk and Givi, 2025). The effect runs through perceived authenticity and a feeling the authors call moral disgust. It weakens when the message is factual, when AI only edits the text, when the AI itself signs and when consumers believe almost all marketing already comes from machines.

Read together, the two studies separate what AI writes well from what it should not sign. Ads, product descriptions and informational text pass the test, and a message that sells closeness, such as a founder’s thank-you note, loses value when the customer finds out nobody wrote it.

Does Google penalize content made with AI?

Google assesses quality and purpose without looking at how the text was produced. In February 2023, Google’s search quality team published that its ranking systems reward original, high-quality content however it is produced, and that appropriate use of AI or automation is not against its guidelines (Google Search Central, 2023). The violation lies in generating content with the primary purpose of manipulating rankings. In March 2024, Google expanded its policy against scaled content abuse to cover text produced to climb rankings by automation, by people or by a combination of both, and reported in April that the update reduced low-quality, unoriginal content in results by 45%, above the 40% it expected (Google, 2024).

The most cited field study on the subject predates ChatGPT. Martin Reisenbichler and coauthors used natural language generation to draft SEO landing page texts, with a human editor revising each piece, and tested the result in field experiments in two industries (Reisenbichler and coauthors, 2022). The revised texts were nearly indistinguishable from those written by SEO experts on the perceptual dimensions evaluated, outranked the human texts in search and cost less to produce. The authors note that the human editor’s role remained essential. Combined with Google’s policy, the study indicates that human review decides whether a page made with AI ends up being useful.

What do these numbers mean for B2B marketing?

The research covers large companies, many of them in consumer markets, and markets outside Brazil. Even so, four readings apply to the B2B operations I work with.

Content at scale without losing the voice

The Zhang and Gosline experiment shows that text made with AI tends to pass the writing test. In B2B, the risk appears when the whole team asks the same model for the same topics and publishes correct pieces that look like the competitor’s, something the productivity experiments also record, as I discuss in generative AI and productivity. The 5% to 15% gain on marketing spend that McKinsey estimates only shows up if the freed time goes to what the machine does not do alone, such as customer interviews, proprietary data and the company’s point of view.

Review proportional to risk

The SEO study and Google’s policy say review is the part of the process that separates useful content from content at scale. Kirk and Givi add a rule for leadership: relationship messages that depend on authenticity, such as those sent in the name of a director or account executive, need a real author, even if AI helps with editing.

CRM data to personalize

Personalization is the first example McKinsey cites, and it depends on what the company knows about the contact. Without interaction history, an updated job title and the account stage in the CRM, AI personalizes with the name and the company, which buyers spot from a distance. The numbers that measure the effect of personalization are in personalization in data.

Measure in pipeline instead of volume

Gartner’s CMOs report AI returns in time, cost and content volume. These are the three measures easiest to inflate with AI and the three furthest from revenue. In B2B marketing, I ask leadership to track sourced and influenced opportunities, MQL to SQL conversion and revenue by channel in the CRM, with the attribution I describe in marketing attribution, and to add the cost of AI to that calculation, as in AI ROI.

What these data do not say

McKinsey’s estimate is a 2023 model-based projection of global spend by function. It measures potential productivity, excludes implementation costs and does not break down marketing value by country or company size.

Both Gartner surveys hear from CMOs in North America, the UK and Europe, almost all at companies with revenue above $1 billion, reporting on their own work. The release does not detail what counts as “AI initiatives”, and the high-maturity group shows a correlation between budget and readiness without proving that more AI spending produces results.

The academic experiments use online panel participants rating consumer pieces, and none of them tested a B2B buyer reading a proposal or a nurture email within a months-long buying process. The SEO study used text generation that predates current models and covers two industries. Google’s 45% comes from the company’s internal assessment, without independent audit, and the post does not say how much of the removed content was produced with AI.

None of the sources measures the effect of AI on a B2B operation’s pipeline. Your company produces that number, with a baseline before the pilot.

What leadership does with these numbers

This is the roadmap I propose for marketing leaders who will discuss AI budgets with this research on the table.

1. Calculate the base McKinsey’s range applies to. Add up marketing spend on text production, images, ad variations and personalization, including agencies. The 5% to 15% range applies to that base, and the result is usually much smaller than the trillion slide.
2. Compare your AI share with Gartner’s 15.3%. List licenses, credits, agent consumption and implementation hours. If your share is close to average and the process is not ready, the next investment goes to data and training.
3. Swap efficiency for pipeline on the AI dashboard. Production time and cost stay as operating indicators, and sourced opportunities, MQL to SQL conversion and revenue by channel become the outcome indicators.
4. Separate what AI writes from what it does not sign. Release AI for informational content, variations and descriptions, with sample review, and keep a human author on relationship messages sent in people’s names.
5. Audit published content against Google’s policy. Review the pages created in volume over the last 12 months and remove those that do not answer a real buyer question.
6. Write the disclosure policy. Decide with legal when the company discloses the use of AI, bearing in mind that in Kirk and Givi’s experiment the negative effect diminished when the message was signed by the AI itself.

What HR and L&D should take from this research

Gartner points to the lack of processes, governance and talent as what separates the 30% that are ready from the other 70%, and McKinsey leaves training out of the potential, which leaves that cost for the company to add. For HR, the need for training shows up when marketing produces more pieces and conversion into opportunities does not rise. The track that works brings together marketing leadership, analysts and sales representatives, because the lead definition and the pipeline metric are shared. The effect of training is measured in behavior over the following weeks, such as the share of pieces approved without rewriting and the evolution of MQLs accepted by sales.

The research shows large potential, growing budgets and few people prepared to carry the production gain all the way to pipeline. That is the work of the in-company applied AI training, built on the company’s process and CRM, with marketing, sales and CS cohorts. It delivers the prompt library by area and the agent architecture, with what each agent reads, what it executes on its own and who approves the rest.

Frequently asked questions

How much could generative AI generate in marketing and sales?

McKinsey estimated in 2023 a potential of $760 billion to $1.2 trillion a year in marketing and sales, the highest range among the 16 functions analyzed. In marketing, that equals 5% to 15% of the function’s total spend. It is a productivity projection, without implementation costs, and not a measured result.

How much do CMOs invest in AI in 2026?

In Gartner’s survey of 401 CMOs in North America, the UK and Europe, the average was 15.3% of the marketing budget allocated to AI initiatives. The total budget came to 7.8% of revenue, and only 30% of CMOs said they had mature capabilities to scale AI.

Do consumers trust AI-generated content?

In an experiment with 1,203 participants, ads written by AI received higher scores when the source was not revealed, and knowing AI took part did not lower the score. Another set of seven experiments showed a drop in recommendation and loyalty when consumers believe an emotional message was written by AI.

Does Google penalize text written with AI?

Google states that appropriate use of AI is not against its guidelines and that its systems reward useful content however it is produced. The penalty falls on content produced at scale to manipulate rankings, by machines or by people, and the 2024 update reduced low-quality, unoriginal content in results by 45%, according to the company itself.

How do you measure the results of AI in B2B marketing?

Use production time and cost as operating indicators and measure results in the CRM, with sourced opportunities, MQL to SQL conversion and revenue by channel, against a baseline recorded before the pilot. Add licenses, credits and implementation to the cost.

How do you prepare the marketing team to use AI?

Start with leadership, which defines the use cases, who reviews each type of piece and the pipeline indicators. Then train analysts with the company’s context and CRM history, with sales present in the lead criteria sessions, and track behavior over the following weeks.