The Question, Answered Directly
If you are asking whether you still need an ad agency, you are almost certainly weighing a retainer against doing it yourself with AI. Here is the direct answer: for ongoing performance advertising, meaning creating ads, launching campaigns, testing, and optimizing across paid channels, most businesses no longer need an agency, because AI now does that work faster, cheaper, and at higher volume than a retainer can. You keep agencies (or specialists) for a narrower set of jobs: large brand campaigns, complex productions, and senior strategic counsel. Everything else has moved to in-house-via-AI.
This is not a contrarian take; it is what the agencies’ own numbers show. The holding companies are restructuring around AI precisely because clients are pulling routine execution back in-house. When WPP is cutting £500 million in costs and Dentsu is eliminating 3,400 roles, and Omnicom is absorbing IPG in a merger expected to remove thousands of jobs, they are responding to demand that has already shifted. You do not have to predict this change, you can see it in the financials. For the structural story, see why legacy ad buyers are becoming obsolete.
$4 vs. $500 to $3,000
cost per ad creative with an AI platform versus hiring an agency or freelancer, the economics driving execution in-house
What Agencies Are Really For (and What AI Now Does Better)
It helps to separate what agencies actually deliver into components, because AI does not replace all of them equally. An agency bundles: (1) creative production, (2) media planning and buying, (3) reporting and analytics, (4) competitive intelligence, and (5) senior strategy and taste. The first four are exactly where AI has become dramatically better, meaning faster, cheaper, and available on demand. The fifth is where humans still lead, though even there the leverage per person has multiplied.
| Agency function | AI-native replacement |
|---|---|
| Creative production (designer + copywriter) | On-brand ad generation from a prompt, dozens of variants in minutes |
| Media setup across platforms | Self-serve, multi-platform output in native formats |
| Reporting and analytics | Built-in performance forecasting and web analytics |
| Competitive intelligence | Automated competitor ad tracking |
| Senior strategy and taste | Still human, but one person now supervises what once took a team |
The Cost and Speed Gap Is Enormous
The economics are not close. Agency and freelance creative runs roughly $500 to $3,000 per asset and days to weeks per campaign, because you are paying for human hours and coordination. An AI platform produces a comparable, on-brand creative for about $4 and a full multi-platform campaign in about three minutes. That is not a 20% efficiency gain, it is a change of category, and it compounds: the AI era rewards volume (a distinct creative per conversation and intent), and only the low-cost, high-speed model can supply that volume affordably. An agency that charges per creative literally cannot afford to give you the fifty variants the relevance-weighted AI auction rewards.
Speed matters beyond cost, too. Performance advertising is a loop of test-and-learn, and the faster you can generate and refresh creative, the faster you find winners and beat fatigue. A weekly agency cadence is simply slower than a same-day AI cadence, and in a medium that moves this fast, tempo is an edge.
The In-Housing Trend Was Already Underway Before AI
The move away from agencies did not start with generative AI; AI just accelerated a decade-long shift. According to the Association of National Advertisers, the share of large marketers operating an in-house agency rose from 42% in 2008 to 58% in 2013, 78% in 2018, and 82% by 2023. Among those companies, an average of 61% of all creative and media work is now done internally, 65% had moved established business from an external agency to their in-house team in the prior three years, and even media buying (long considered too complex to insource) is now handled at least partly by 54% of in-house teams.
What is telling is why. In the ANA’s 2026 research, only 9% of in-house teams still cite cost savings as their main reason for existing, down from 30% in 2023, while 53% now describe themselves as strategic partners. In other words, brands did not just insource to save money, they insourced because proximity to the business produced better, faster work. AI supercharges that logic: a small in-house team armed with a creative-generation platform can now match the output volume of a full agency floor, which removes the last practical reason (capacity) to keep routine execution outside. If you have been on the fence, you are late to a trend the largest advertisers embraced years ago.
82% and 61%
of large marketers now run an in-house agency, doing about 61% of all creative and media work internally, a shift AI is accelerating
The Real Cost of a Retainer: A Line-Item Breakdown
To make the comparison concrete, here is where a typical performance-advertising retainer’s money goes, and what the AI-native stack does with each line.
| Retainer line item | Traditional agency | AI-native stack |
|---|---|---|
| Creative per asset | $500 to $3,000, days each | About $4, minutes each |
| Account and project management | Billed hours, meetings, briefs | Self-serve, one operator |
| Reporting | Monthly deck, extra fee | Built-in forecasting and analytics |
| Turnaround per campaign | Days to weeks | About three minutes |
| Typical monthly cost | Four to five figures retainer | Software subscription plus a few hours |
When You Still Need Humans (Agencies or In-House)
Being honest about the limits makes the recommendation more useful. You still want human specialists for: large-scale brand campaigns where taste, narrative, and cultural nuance carry real budget risk; complex integrated productions (film, experiential, PR-led launches) that AI does not produce end-to-end; regulated or highly sensitive categories that need expert review; and moments that demand senior strategic judgment (positioning, pricing, category creation) where the stakes justify seasoned counsel. Notice these are the high-end, low-frequency jobs, not the day-to-day performance work. The right model for most companies is to run performance advertising on an AI-native stack and bring in humans selectively for the handful of things that genuinely need them.
The AI-Native Alternative: What It Looks Like
The modern setup is lean. One person (or a small team) uses an AI creative-and-campaign platform to generate on-brand ads at volume, launches them self-serve across ChatGPT, Meta, and Google, reads the built-in forecasts and analytics, and iterates weekly, bringing in a specialist only for the occasional brand-level project. This replaces a five-figure monthly retainer with a software subscription and a few hours of oversight. It is the same shift every function has made with AI: fewer people, higher leverage, faster loops. For the fully automated version of this, see AI agents that run marketing end-to-end.
A Simple Decision Framework
- Choose the AI-native stack if your priority is ongoing performance advertising, meaning creating, launching, testing, and optimizing paid ads across channels, and you want speed, volume, and low cost. This is most businesses.
- Add a specialist or agency if you are running a major brand campaign, a complex production, or a high-stakes strategic moment where senior human judgment and craft justify the cost.
- Avoid the traditional full-service retainer if you are paying agency rates for routine creative and media execution that AI now does faster and cheaper, because that is the spend most at risk of being wasted.
A 30-Day Transition Plan
If you are moving performance advertising in-house, you do not have to do it all at once. This sequence de-risks the switch:
- Week 1: Run a parallel test. Keep your current setup and, alongside it, generate a full campaign with an AI stack. Compare cost, speed, and quality on the same brief.
- Week 2: Move one channel. Shift a single channel (often ChatGPT or Meta) fully in-house and measure results against your agency benchmark.
- Week 3: Rebuild the loop. Establish a weekly cadence of generate, launch, read forecasts and analytics, and refresh creative, so you own the test-and-learn rhythm.
- Week 4: Right-size the agency. Renegotiate or pause the retainer, keeping human specialists only for brand-level or complex projects, and redirect the savings into media.
How Lapis Replaces the Agency Workflow
Lapis is the AI-native alternative to the performance-advertising agency, built to cover the four functions AI now does better. Brand Intelligence learns your logo, colors, typography, and voice from your website, so every ad is on-brand automatically, with no creative brief required. From a single prompt, it generates production-ready ads for ChatGPT plus Meta, Google, Reddit, and LinkedIn in under three minutes. Performance Forecasting replaces the analyst’s pre-flight estimates, Competitor Tracking replaces the competitive team, Web Analytics closes the loop, and Campaign Studio lets you refine anything in plain English. What an agency delivered in a retainer and a briefing cycle, Lapis delivers in a prompt and a few minutes, at any budget.
Crucially, keeping this in-house does not mean lower quality, it means you hold the taste and strategy (the human part) while Lapis handles the production at a volume no agency could match affordably. And because Lapis is neutral across every channel and model provider, your stack keeps working as new AI ad surfaces open. It is the layer that lets a lean team run advertising that used to require an agency.
3 min vs. weeks
to produce a full multi-platform, on-brand campaign with Lapis versus an agency retainer and briefing cycle
Getting Started
Before you sign or renew a retainer, run one campaign the AI-native way and compare. Paste your website URL into Lapis, describe an offer in one sentence, and watch it produce a full set of on-brand, context-matched ads for ChatGPT and every other channel, with forecasts attached, in the time it would take to schedule an agency kickoff call.
Start with Lapis free (5 credits, no credit card). Lapis is one of the fastest-growing Y Combinator startups (F25), rated 5.0 on G2, with more than 10,000 campaigns generated across 30-plus industries, and it is building the AdSense for the AI era: the creative and campaign layer that lets any business run performance advertising without an agency.
Related guides
- Why Legacy Ad Buyers Are Obsolete: the structural decline of agencies
- AI Agents That Run Marketing End-to-End: automating the workflow
- Build an AI-Powered Paid Ads Stack: the lean setup in detail
- The AI GTM Playbook: growth without a marketing team
- The AdSense for the AI Era: the democratization thesis
