Marketing Agency AI Adoption Statistics 2026

Census Bureau data released on October 22, 2026 showed 24.6% of US businesses using AI in the two weeks to October 18, the highest share since the Business Trends and Outlook Survey (BTOS) revised its AI question in November 2025.

Our analysts compiled data on marketing agency AI adoption, tasks and tools, staffing, pricing models, revenue and client demand in October 2026 from two Census Bureau BTOS products, Bureau of Labor Statistics (BLS) payroll employment for advertising and public relations firms, Indeed job posting data and six agency surveys fielded between 2025 and 2026. We coded each survey by population, respondent count, fieldwork date and question wording, built our own comparison tables from those codes and added the disclosed AI investment of three agency holding companies, converted to US dollars.

In the following sections, we measure adoption, compare AI tasks and tools, review staffing, pricing, revenue and client demand, break out AI use by firm size and set out the barriers agencies report to wider AI use.

How Many Marketing Agencies Use AI

This section measures how widely US agencies use AI, first with federal survey data and then with the published agency surveys, coded question by question.

The BTOS surveys about 200,000 businesses in each two-week collection, drawn from a sample of about 1.2 million US employer businesses, and asks whether they used AI in any business function in the previous two weeks. For agencies we used the professional, scientific and technical services sector (NAICS 54), which contains advertising and public relations firms. Table 1 compares that sector with all US businesses.

Data: btos_ai_use_trend_nov2025_oct2026.csv

• The NAICS 54 rate rose 8.3 points between the two six-period averages, against 5.4 points for all US businesses (our calculation, base: six BTOS periods in each average, NAICS 54 and all US businesses). The later NAICS 54 average was about 1.8 times the all-business average (our calculation, base: six BTOS periods, NAICS 54 vs all US businesses).
• The NAICS 54 rate has been above 40% for five straight periods, starting with the August 10 to August 23, 2026 collection. The last reading under 40% was 39.3% in the July 27 to August 9, 2026 period.
• The NAICS 54 rate stayed at least 14.8 points above the national rate in all 24 periods since November 2025 (our calculation, base: NAICS 54 vs all US businesses in each two-week period). A national AI rate therefore understates adoption in the part of the economy where agencies are counted.

The Census Bureau also ran a one-time AI supplement alongside the core survey from November 17, 2025 to February 8, 2026, which drew about 18,000 responses from NAICS 54 firms and about 117,000 nationally. In that supplement, 34.2% of NAICS 54 firms had used AI in the prior two weeks, against 17.9% of all firms. Employees used AI for work tasks at 39.4% of NAICS 54 firms (all firms 22.6%) and used generative AI at 34.9% (all firms 20.8%). Looking ahead, 37.5% of NAICS 54 firms in the supplement expected to use AI in the following six months, against 21.6% of all firms.

Agency surveys ask different questions of different people. Table 2 codes six of them, plus the two BTOS measures, on the same fields so each figure can be read against its population and wording.

Data: agency_survey_adoption_coding.csv

• Half of agencies in the 2026 Forrester and 4As survey use agentic AI (nearly 200 US agency leaders). Generative AI use in that survey is more common than agent use.
• Among Basis respondents, 39.9% use AI at work without using it daily (our calculation, base n=213). Daily use is now the majority answer in that survey.
• A headline agency AI adoption rate depends on whether the survey counts any use or regular use. Cited alone, a near-universal any-use figure says little about how much agency work AI reaches.

The industry surveys differ from the BTOS in method. They ask whether an agency or person uses AI at all, with no two-week window. Most of them poll individual professionals or senior leaders, and their respondents opted in to each survey. The Forrester and 4As series shows the change within one survey program, with generative AI use rising from 61% of agencies in 2024 to 75% in 2025. The Basis agency series shows daily use by agency professionals climbing from 9.9% in 2023 (n=101) to 15.9% in 2024 (n=113), 38.6% in 2025 (n=171) and 59.2% in 2026 (n=213). Over the same four waves, the share of Basis respondents who did not use AI at all fell from 32.7% to 0.9%.

Frequency of use trails breadth of use in the agency surveys. In the 2025 Forrester and 4As survey, nearly 80% of agencies used generative AI on less than half of their work, even as 89% cited employee productivity as a benefit and 69% cited production cost savings, up from 13% in 2024. By April 2026, 81% of US agency leaders in the Forrester and 4As survey said staff productivity is their primary objective for generative AI (nearly 200 respondents).

Which Tasks Agencies Use AI For

This section compares the business functions and tasks where firms apply AI, starting with federal data and then moving to agency surveys.

The BTOS AI supplement asked every firm which functions had used AI in the prior six months, and it asked firms whose employees used generative AI which tasks they used it for. Table 3 shows the function results for NAICS 54 firms (about 18,000 responses) and all US firms (about 117,000 responses).

Data: btos_ai_supplement_2026_naics54_vs_all_us.csv

• Research and development is the most common AI function among NAICS 54 firms. Its rate is about 2.1 times the all-firm rate (our calculation, base: about 18,000 NAICS 54 and about 117,000 all-US supplement responses).
• Human resources is the least common AI function in both groups. It also shows the smallest gap between them.
• In professional services firms, AI use centers on research and client delivery work. Back-office functions such as human resources and customer service trail by wide margins.

Provision of services shows the widest ratio in the table, with NAICS 54 firms using AI to deliver client work at 2.8 times the all-firm rate (our calculation, base: about 18,000 NAICS 54 and about 117,000 all-US supplement responses). Sales and marketing, the function where AI use is most common across all US firms, shows one of the smallest gaps. Among NAICS 54 firms planning to use AI in the next six months, research and development (60.3%), strategy (57.6%) and information technology (53.7%) led the planned functions, followed by sales and marketing (51.8%). Across all US firms planning to use AI, sales and marketing led the planned functions at 62.8%.

Data: btos_ai_supplement_2026_naics54_vs_all_us.csv

• Writing and editing leads both groups. It is one of two tasks, with customer support, where NAICS 54 users trail all users.
• NAICS 54 generative AI users apply it to software coding at about 1.9 times the all-firm rate (our calculation, base: NAICS 54 vs all US firms whose employees used generative AI). That is the widest gap in the table.
• Generative AI use in the sector extends past writing into search, analysis and project research. More than half of NAICS 54 generative AI users apply it to search and to analysis, so the sector uses it as a research tool as well as a writing aid.

Agency surveys use their own task lists. We grouped matching items into common task groups and kept each survey's base, as shown in Table 5.

Data: agency_ai_tasks_by_survey.csv

• Ideation is the only task group above 80% in all three surveys that asked about it. Basis records the highest share, at 86.9% of agency professionals (n=213).
• SAGA owners use AI for reporting more often than for image generation or SEO (n=62). Fewer than half use it for any of those three tasks.
• Agencies apply AI most to research, ideation and writing work. Media planning and buying are the least common task groups in the table.

Research, ideation, summarizing and first drafts top every survey that asked about them, and media work comes in lower. SAGA also asked owners about proposals (66%), new business prospecting (42%) and media monitoring (24%). SoDA agencies had adopted an average of four of its seven AI uses, and 47% of SoDA leaders use AI to build new service offerings for clients, while 37% develop custom AI models or agents (n=251).

Which AI Tools Agencies Use

This section covers the AI assistants, AI agents and in-house platforms that agencies report using.

Two of the surveys asked which assistants agency staff use, and Basis asked agency professionals whether their agency had built or deployed AI agents and what those agents do. Table 6 brings those items together with their bases.

Data: ai_tools_and_agents.csv

• Gemini is the third most used assistant among SAGA owners (n=62). It leads Microsoft Copilot by 33 points (our calculation, base n=62).
• Reporting is the most common job for AI agents among Basis agent users (n=98). Media buying is the least common, at fewer than 1 in 10 agent users.
• At least 52% of SAGA's owner-led agencies use both ChatGPT and Claude (our calculation, base n=62). Most owner-led agencies run more than one general assistant, so their AI work spans several vendors' terms and data policies.

Basis recorded Claude use rising from 17.9% to 72.3% of its agency and brand respondents in one year while ChatGPT use fell from 88.6% to 74.5% (n=184 in 2026). Image generation tools (16%) and Jasper (5%) trail the general assistants among SAGA respondents (n=62). In the Forrester and 4As survey, 63% of agency leaders said staff productivity is their primary objective for AI agents (nearly 200 respondents).

Large holding companies run their own platforms on top of these models. WPP reported over 70,000 monthly active users of its WPP Open platform in December 2025, about 90% of its client-facing staff, up from 33,000 users (about 40%) in December 2024. WPP's total technology spend in 2025 was £642 million ($846.9 million, our calculation, base: WPP disclosure at the Federal Reserve 2025 average rate), down from £684 million in 2024.

Tool adoption also shows up in federal data. Among NAICS 54 firms that used AI, 24.3% said AI replaced software or equipment they had used before, against 16.1% of all AI-using firms, and 12.5% bought new computing power or software to support AI, against 8.2% (BTOS AI supplement, about 18,000 NAICS 54 responses).

AI and Agency Staffing

This section sets federal employment counts for advertising and public relations firms beside what AI users and agencies report about jobs.

The BLS Current Employment Statistics program publishes monthly payroll employment for advertising, public relations and related services (NAICS 5418), seasonally adjusted. Table 7 shows four recent monthly points from that series.

Data: bls_ces_naics5418_employment_2019_2026.csv

• Employment fell 0.6% between September 2025 and September 2026 (our calculation, base: BLS payroll estimates for September 2025 and September 2026). That is 2,700 fewer jobs, from 481.3 thousand to 478.6 thousand.
• August 2026 was the low point of the four months shown, at 476.8 thousand. September 2026 employment was 1,800 jobs higher (our calculation, base: BLS payroll estimates for August and September 2026).
• The BLS series counts job changes from all causes, so payroll trends alone cannot show AI's effect on agency jobs. The August and September 2026 estimates are preliminary and subject to revision.

Firms that use AI report its direct effect in the BTOS supplement, shown in Table 8 for NAICS 54 and for all US firms.

Data: btos_ai_supplement_2026_naics54_vs_all_us.csv

• AI supplemented employee tasks at most NAICS 54 AI users. That rate is 16.8 points above all AI-using firms (our calculation, base: NAICS 54 vs all US AI-using firms, BTOS AI supplement).
• Net employment effects were close to zero in both groups. Among NAICS 54 AI users, job gains exceeded job losses by 0.9 points (our calculation, base: NAICS 54 AI-using firms, BTOS AI supplement).
• Just over half of NAICS 54 AI users (51.7%) made no organizational changes to adopt AI, and those that did most often trained current staff or developed new workflows. In professional services firms, AI has so far changed how existing staff work more than how many people firms employ.

Among NAICS 54 firms where AI replaced employee tasks, 61.6% said it replaced a small number of tasks, 27.6% a moderate number and 10.8% a large number. Hiring staff trained in AI was rare, at 2.1% of NAICS 54 AI users against 1.3% of all AI-using firms. Among all AI-using US firms, 64.3% made no organizational changes.

Agency surveys asked about staffing in their own terms, and Table 9 codes those items.

Data: agency_staffing_ai_items.csv

• Avoided hires outnumber AI-driven headcount cuts among SAGA's owner-led agencies (n=62). Together they cover 21% of those agencies (our calculation, base n=62).
• SoDA leaders plan to raise full-time labor spending more often than freelance spending in 2026 (n=251). The gap is 10 points (our calculation, base n=251).
• AI shows up in agency staffing through hiring choices and job requirements more than through AI-driven layoffs. The share of US marketing postings mentioning AI rose 6.5 points during 2025 (our calculation, base: Indeed US marketing postings, January and December 2025).

Indeed measured AI mentions in 4.2% of all US job postings in December 2025, and the marketing share was about 3.5 times that rate (our calculation, base: Indeed US marketing vs all US postings, December 2025). SoDA respondents also plan to raise training spending (50%) in 2026. In the Basis survey, 77.7% of senior agency leaders plan to increase AI investment over the next 12 months (n=94), and 73.7% of all respondents work at agencies planning new automation technology (n=213). SAGA found 3% of owner-led agencies had hired for AI roles (n=62).

AI Pricing Models at Agencies

This section tracks how agencies account for AI costs and whether they bill clients for AI work.

Forrester and the 4As have asked US agency leaders in two consecutive years whether AI is a cost of doing business or a line of business, and SAGA and SoDA asked related pricing questions. Table 10 codes those items.

Data: agency_ai_pricing_models.csv

• Nearly a third of US agency leaders in the 2026 Forrester and 4As survey plan to monetize AI agents within 24 months (nearly 200 respondents). Only 6% treat AI as a line of business today.
• Among SAGA owners, 28% charge clients directly for AI or plan to (our calculation, base n=62). Fewer disclose AI use in client contracts than charge for it.
• Most agencies treat AI as a cost of business. Agencies that bill clients for AI work are a minority in both surveys that asked.

The cost-of-business share fell 14 points between the 2025 and 2026 Forrester and 4As surveys (our calculation, base: 2025 and 2026 Forrester and 4As respondents, different groups in each year). In 2026, 6% treated AI as a line of business, after fewer than 15% expected a line-of-business model within two years in the 2025 survey. Most SAGA owners (61%) have no plan to charge clients directly for AI. In the SoDA study, 55% of agency leaders called client pricing pressure a growing challenge. Asked where they see room to deliver more value to clients, 49% of SoDA leaders chose exploring new pricing or service models and 70% chose using AI for speed, collaboration and output (n=251).

Agency Revenue and Client Demand

This section brings together what agencies report about client requests and revenue since AI entered their work.

SoDA asked agency leaders which AI-related requests they receive from clients, Basis asked agency professionals how much AI threatens agency revenue, and SoDA, SAGA and Forrester asked about growth and client goals. Table 11 shows the results with their bases.

Data: agency_revenue_and_client_demand.csv

• Help with AI strategy and faster delivery tie as the most common client requests in the SoDA study (n=251). Expectations of lower prices as a result of AI rank third.
• Most Basis respondents saw clients move work in-house in the past 12 months (n=213). About 1 in 7 rate AI as a significant threat to agency revenue.
• SoDA leaders also field client requests for AI training (30%) and custom brand models (28%, n=251). Agencies selling AI services are asked to widen the scope of their AI work while facing pressure on fees.

The SoDA and SAGA fee items come from different questions asked of different groups. SoDA asked leaders whether clients are expecting lower prices as a result of AI, and SAGA asked owners whether any client had requested lower fees because of AI. In the Forrester and 4As survey, 11% of agency leaders said clients want revenue growth as the main benefit of AI agents. US agencies in the SoDA study reported 2025 revenue growth less often than European agencies, at 69% against 88%.

Agency leaders expect the effects to grow. In the SoDA study, 24% said AI already has a significant impact on their business and 48% expect a significant impact within 12 to 18 months (n=251). Basis asked whether the traditional agency model is broken, and 29.1% said yes, 35.7% said somewhat and 22.5% said it will be within three to five years, leaving 12.7% who said no (n=213).

AI Adoption by Agency Size

This section compares AI use across firm sizes with federal employment size classes, then adds the size and revenue splits that agency surveys publish.

BTOS publishes AI use for NAICS 54 by employment size class. Single two-week estimates for the larger classes move several points from period to period, so Table 12 averages the six periods from July 27 to October 18, 2026.

Data: btos_ai_use_by_employment_size_jul_oct2026.csv

• NAICS 54 firms used AI above the all-business rate in every size class. Firms with 100 or more employees used it 16.9 points more often than firms with one to four employees (our calculation, base: six BTOS periods, NAICS 54 size classes).
• The widest gap with all US businesses is in the 50 to 99 employee class. The narrowest is in the 250 or more class.
• NAICS 54 firms with one to four employees used AI at 41.3%, against 40.7% for all US businesses with 250 or more employees. A professional services firm with a handful of staff uses AI at a slightly higher rate than the largest US businesses.

Agency surveys split by revenue. In the SoDA study, 63% of agencies under $10 million in revenue grew in 2025, against 86% of agencies above $10 million. The SAGA sample is mostly micro agencies, with 64% at five or fewer full-time employees and 71% under $1 million in fees, and 74% of its owners use AI daily (n=62). Half of SAGA respondents (50%) had used AI meaningfully for more than a year, and 88% reported a positive effect on productivity (n=62). The SoDA sample spans a wider size range, with 39% of its 251 agencies under $5 million in revenue.

Data: agency_size_survey_items.csv

The largest agency groups report AI spending directly. Table 13 converts their disclosed commitments to US dollars at one fixed rate per currency.

Data: holdco_ai_investment_usd.csv

• WPP's 2025 AI investment exceeds the per-year amounts of the Publicis and Havas commitments combined (our calculation, base: company disclosures at Federal Reserve 2025 rates). WPP's figure covers one year of spending.
• Publicis and Havas each commit about $113.1 million a year (our calculation, base: USD total divided evenly across stated years). Their totals differ because the Havas commitment runs one year longer.
• The three commitments work out to more than $100 million a year at each group (our calculation, base: company disclosures at Federal Reserve 2025 rates, divided evenly across stated years). The commitments differ in scope and period, so the per-year amounts compare stated budgets.

Publicis planned to split the first €100 million of its program, budgeted for 2024, evenly between people and technology.

Barriers to Wider AI Use at Agencies

This section compares the risks and concerns that hold back wider AI use at agencies and at professional services firms with no plans to use AI.

Forrester and the 4As asked US agency leaders which barriers hold back generative AI and AI agents, and SAGA asked owners about their concerns. The BTOS supplement asked firms with no plans to use AI for their reasons. Table 14 shows those items with their bases.

Data: agency_ai_barriers.csv

• Accuracy leads the concerns in both agency surveys that asked about it. Legal risk follows closely among the US agency leaders in the Forrester and 4As survey (nearly 200 respondents).
• Among NAICS 54 firms with no plans to use AI, privacy or security concerns run 8.0 points above the all-firm rate (our calculation, base: NAICS 54 vs all US firms not planning to use AI). The gap for AI not being mature enough is 8.5 points (our calculation, base: NAICS 54 vs all US firms not planning to use AI).
• The limits on wider AI use center on output quality and data handling more than on price. Among NAICS 54 firms with no plans to use AI, too expensive (6.7%) ranks below privacy or security (28.7%) and AI not being mature enough (21.5%).

Conclusion

Agencies whose clients want help with AI strategy are being asked for work that few agencies treat as a line of business. An agency that starts billing for that work would be pricing it while some clients expect AI to lower what they pay.

For a company hiring a marketing agency, AI work on its account is, at most agencies, a cost of business, and few agencies charge clients for it directly. That work is concentrated in research, ideation and drafting, and few agencies disclose it in their client contracts.