Will AI replace agencies? Not as a category, but it will replace a share of agency tasks, some roles and much of the hourly production fee. Forrester expects automation to remove about 7.5% of US agency jobs by 2030, concentrated in clerical, sales and research work. Agencies that move from producing assets to directing AI and explaining why work performs will keep their seat at the table.
Key takeaways
- AI is replacing agency tasks faster than it is replacing agencies; the best-known forecast puts automation at 7.5% of US agency jobs by 2030.
- AI enters advertising in three ways, replacing, assisting and explaining creative work, and each threatens a different part of the agency model.
- Platform automation squeezes production and media-buying fees first, because those are the services easiest to describe as inputs.
- The defensible ground is judgement: knowing why something works, for which audience, and taking responsibility for the decision.
- Agencies that keep training juniors and make every AI conclusion traceable to evidence will hold that ground.
What does the evidence say about AI replacing agency jobs?
It points to a large but partial shift: some roles shrink, some are redesigned and a few are created. None of the serious forecasts describes agencies disappearing.
The agency workforce forecast from Forrester expects about 32,000 US agency jobs, or 7.5% of the workforce, to be lost to automation by 2030, with generative AI accounting for nearly a third of them. The roles most affected are clerical (28% of the generative AI losses), sales (22%) and market research (18%). Forrester also names originality as the factor that most reduces a job's automation potential.
A survey of 1,303 senior leaders, run from January to April 2026, shows the same mix inside marketing teams. In a September 2026 release from Gartner, 18% of marketing leaders had eliminated roles because of automation, 16% had created new roles and nearly a third had redesigned existing ones.
The sharper threat is to revenue rather than to existence. In the 2026 Advertising Agency Report from Basis, nine in ten agency professionals said AI threatens their agency's primary revenue streams. The question for leaders is less whether agencies survive than which agency fees survive.
What are the three ways AI is entering advertising?
AI can replace creative work, assist it or explain it. I set out this comparison in my industry expert talk at ACM Multimedia 2025 in Dublin, "Will AI Make Agencies Obsolete? Rethinking the Future of Advertising" (the recording is on YouTube), which also looks at how agencies are adapting to each. The three modes do not threaten the same things.
| Mode | What AI does | Who captures the value | Agency exposure |
|---|---|---|---|
| Replacing | Generates, targets and optimises ads end to end from a product feed and a budget | The platform | High for production and buying fees |
| Assisting | Drafts, edits and versions work inside tools people already use | Shared between the tool and the team using it | Medium: faster output, pressure on hourly pricing |
| Explaining | Reads market and performance data and says why something works, for whom, with evidence | Whoever owns the question and the reasoning | Low, if the agency owns the reasoning |
The replacing mode is not hypothetical. EMARKETER reported in June 2025, citing The Wall Street Journal, that Meta intends to let brands create and target ads fully with AI by the end of 2026, from an uploaded product and a budget. When a platform can do that, the agency fee for building variants and setting up campaigns is the first to be questioned.
The explaining mode is where most of SOMIN's research sits. Our peer-reviewed papers, listed on the SOMIN research page, combine performance models with large language models so a marketer can see why a creative or a competitor message works, not only a score. Examples include "Against Opacity" at ACM Multimedia 2023 and SOEX, on explainable competitor analytics, at The Web Conference 2025.
Which agency work is most exposed to AI?
Work defined by volume rather than judgement. If a service is priced by the number of assets, reports or hours, a machine can usually deliver the volume more cheaply.
- Most exposed: versioning and resizing assets, routine performance reporting, desk research that collates public data, and campaign set-up on automated buying platforms.
- Least exposed: choosing which problem is worth solving, original ideas, reading culture and context across markets, and being accountable to a client for a recommendation.
SOMIN's own data shows clients already noticing the difference. In our ads-optimisation workspace, a September 2026 whitespace study of 156 posts grouped 23 under the theme "Navigating the Challenges of AI and Marketing Effectiveness". One of its pain points is the rise of AI agencies that prioritise speed over quality and deliver generic outputs that miss business goals.
A related SOMIN insight, "Authenticity is the antidote to homogenization", rests on the cultural-homogenisation angle: when everyone generates from similar templates, brands lose a recognisable voice. Our SIGIR 2026 paper on agentic AI that counters content homogenisation addresses the same problem from the research side. AI makes sameness cheap, so difference becomes the scarce thing clients pay for.
What do clients still need from an agency?
Clients need someone accountable for outcomes, with evidence they can check. AI tools give them more output; they do not give them a partner who will stand behind a decision.
SOMIN's moments-of-truth model, which generates the micro-moments a persona is likely to act on, produces the thought "I need to evaluate my agency based on real results, not just ROAS or CPA." It is model-generated, not a client quote, but it matches a SOMIN insight built on goal-setting theory: advertisers often find their objectives do not line up with the metrics agencies prioritise. The same model, in our audience-research workspace, generates a buyer looking for the right AI marketing agency, not for no agency at all.
The SAMY case study shows what that looks like in a pitch. Facing larger networks for the Nightcap Group account, SAMY led with a forensic competitive read built in SOMIN instead of a deck of concepts, moved the conversation from services to business impact and won a multi-million-pound retainer.
How should agency leaders respond?
Reposition the agency around the explaining mode, and price and staff it accordingly. Five moves matter most:
- Sell decisions, not hours. If AI halves the production time, an hourly fee halves with it. Price the recommendation and the outcome, and measure both; we show how in how to measure AI ROI in agencies.
- Move producers into orchestration. Briefing AI systems, checking their output and scoring creative before spend is skilled work that clients will pay for.
- Keep the bottom rung. Gartner's advice is to redesign entry-level roles, not remove them. Our playbook for training junior marketers on AI sets out how.
- Make every AI conclusion traceable. A recommendation that links back to the posts and ads behind it can be defended in a client meeting. A black-box answer can be replaced by the next black box.
- Own a data view the platforms will not give you. Each platform reports on itself. A cross-platform read of competitors and audiences is something an agency can own, which is the core of SOMIN for agencies.
FAQ
Will AI replace advertising agencies?
Not as a category. Forecasts such as Forrester's point to about 7.5% of US agency jobs being automated by 2030, with the biggest impact on production, clerical and research tasks and on hour-based fees.
Which agency jobs are most at risk from AI?
Roles built on repeatable volume: clerical work, routine reporting, desk research and asset versioning. Forrester found originality is the factor that most lowers a job's automation potential.
Will Meta's AI advertising tools replace agencies?
They will replace some agency services, mainly building ad variants and setting up campaigns on Meta. They do not replace cross-platform strategy, original ideas or accountability for business results.
How can agencies stay relevant as AI improves?
By shifting from producing assets to directing AI, explaining why work performs and taking responsibility for outcomes. Pricing, hiring and training need to change in the same direction.
Start with an audit of your revenue lines: mark each fee as replacing, assisting or explaining work, and estimate how much of it a platform could deliver within two years. Then pick one exposed service and redesign it around a decision you can explain and evidence, before a client asks you to.
