The Paid Media Agency that was, the one that should be, and the one that will survive

Paid media agencies built their business on manual execution that platforms have already automated. What remains is real strategy: connecting paid media with business objectives in contexts where data is ambiguous. The viable model combines automated tools with human leadership, but the transition carries financial, talent, and governance costs that nobody is quantifying well.

In 2015, a paid media agency justified its fee with concrete, measurable capabilities. A senior media buyer mastered the mechanics of manual bidding in Google Ads, built granular campaign structures with hundreds of ad groups, and managed long-tail keyword segmentation that drove meaningful differences in CPA. Bid optimization was artisanal work: modifiers by device, time of day, geographic location, and audience, all adjusted manually. According to Forrester (2025), agencies under this model dedicated between 60% and 70% of their billable hours to execution tasks including campaign setup, bid adjustments, report creation, and budget monitoring.

The value was tangible because the complexity was real. Managing a Google Ads account with 200 campaigns, 3,000 ad groups, and 50,000 keywords required technical knowledge that an average in-house marketer did not have. The agency functioned as the necessary intermediary between platform complexity and the client's business objective. That complexity was its competitive moat.

That moat began to erode starting in 2018, when Google introduced Smart Bidding as a standard and pushed campaigns toward automated formats. The launch of Performance Max in 2021 and its mass adoption between 2022 and 2023 represented a structural shift: Google consolidated inventory from Search, Display, YouTube, Discovery, and Maps into a single campaign type where the advertiser provides creatives and audience signals, and the platform decides where, when, and to whom to show the ad. Meta followed a parallel path with Advantage+ Shopping Campaigns in 2022, reducing segmentation controls to a minimum and delegating most decisions to the algorithm.

According to eMarketer (2025), Google and Meta concentrate approximately 48% of global digital advertising spend, and both have invested in making their automation tools outperform manual management in most scenarios. Google reported (Q3 2025) that advertisers who migrated to Performance Max saw an average 18% increase in conversions at similar cost. These data points are self-reported by a company with a direct incentive to promote adoption, and the experience of independent advertisers is more varied: in media buyer communities, PMax results fluctuate depending on vertical, data volume, and creative asset quality. For accounts with high conversion volume, results tend to be positive; for niches with little history, performance sometimes worsens when manual controls are lost. But the general direction is clear: the manual work of bid optimization is being absorbed by the platform itself.

Three layers, three different fates

To understand what remains when manual execution is removed, you need to decompose what an agency did into three levels.

The execution layer (campaign setup, bid adjustments, daily budget management, reports) is the one that automates most quickly. Performance Max and Advantage+ eliminate much of the manual setup. Tools like Supermetrics, Funnel.io, and native dashboards automate reporting. According to Digiday (2024), 67% of media buyers reported spending less time on manual optimization than two years prior, and 41% indicated their reporting functions had been partially or fully automated. What previously justified 3 to 5 dedicated people at an agency now runs on tools that cost a fraction of those salaries.

The analysis layer (interpreting data, diagnosing performance drops, proposing data-driven adjustments) is being attacked by AI tools. Google with Gemini integrated into Ads and external assisted-analysis tools generate insights that previously required a dedicated senior analyst. The threshold of "what requires human judgment" constantly moves upward. What in 2020 required an analyst with 5 years of experience can today be identified by an automated system in most scenarios.

The strategy layer (which channels to attack, how to distribute budget, which audiences to prioritize, how to integrate paid media with the marketing mix) still requires contextualized human judgment. But there is an uncomfortable nuance: the strategy that many agencies sold was not always real strategy. There are legitimate exceptions: agencies that provide genuine inventory planning, advance rate negotiation, and cross-channel coordination with measurable value. But for a significant portion of the market, the "strategy" billed was execution packaged as a premium service. The agencies that did genuine strategic work were always the minority, and they are the ones best positioned today.

If execution automates, analysis partially automates, and much of the "strategy" was execution in disguise, what remains is real strategy: deep knowledge of the client's business, the ability to connect paid media with business objectives (not just campaign metrics), and the skill to make decisions where data is ambiguous or incomplete.

The model that emerges (and the one that is possible)

In the ideal version, automated systems replace execution functions and first-level analysis. A system can monitor 50 accounts simultaneously, detect anomalies, implement budget adjustments within predefined parameters, and generate reports with narrative. Forrester (2025) estimates that an agency implementing these tools can reduce its execution team by 40% to 60% while maintaining similar operational capacity. That range is wide because it depends on client type (ecommerce with high SKU volume versus B2B services with long cycles), prior technological maturity, and investment in data infrastructure.

But the distance between the ideal model and reality is considerable. Most mid-sized agencies in LATAM (teams of 15 to 40 people) are not going to create an "AI operations lead" role or hire an ML engineer. What the faster movers are doing is something more pragmatic: reassigning existing people to hybrid functions. One representative, though anecdotal, case is a 25-person agency in Colombia that between 2024 and 2025 reduced its execution team from 12 to 6 by implementing reporting and monitoring automation. Of the 6 reassigned people, 4 moved to client strategy functions and 2 left the company because their skills did not transfer to the new model. It was not an elegant transformation: there was a temporary drop in client satisfaction lasting 3 months and an investment of $40,000 to $60,000 in tools and training in the first year.

The roles that disappear are the most repetitive: the junior media buyer who only does setup, the analyst who consolidates data in a spreadsheet, the account manager whose main function is sending weekly reports. The roles that mutate are the intermediate ones: the senior media buyer shifts to designing structures and parameters within which automated systems operate, defining escalation rules and supervising exceptions. The senior analyst stops extracting data and moves to interpreting complex patterns, connecting paid media insights with business context that no system has: conversations with sales, qualitative competitive intelligence, regulatory changes.

The pricing model also transforms. According to the ANA (2024), fixed retainer models represented 54% of contracts, but performance-based models grew from 12% to 23% between 2020 and 2024. The trend points toward hybrid models: a lower base fee for access to technological infrastructure, combined with a variable component tied to measurable business results.

What the client needs to know

If a company today pays a monthly retainer to a paid media agency, the operational question is direct: what percentage of the fee corresponds to execution that the platform already automates? If the agency charges the same as in 2020 but its media buyers spend half the hours on manual optimization, there is a gap between the service paid for and the service received. This does not imply dishonesty; it implies that the pricing model did not update at the pace of the platforms.

The signals that an agency delivers real value in 2026 are observable: it presents analysis connecting campaign metrics with business metrics (not just CPA and ROAS but impact on pipeline or lifetime value), proposes hypotheses about why something works or does not (not just reports what happened), and brings learnings from other clients that generate an advantage an internal team could not produce alone. If the reports received are dashboards the client could build with Supermetrics, that is a signal.

The alternative of building an internal team with automated tools is real but not trivial. At minimum, one senior person with paid media experience who understands how to configure and supervise automation tools is needed. The opportunity cost is not only financial: an internal team sees its own data but loses the cross-client perspective that a good agency can offer by observing patterns across multiple verticals. Agencies often argue that this "cross-client knowledge" is their main differentiator. In theory, an agency managing 10 ecommerce accounts should know more about what works in ecommerce than the internal team of any of those companies individually. In practice, the real transfer of learnings between accounts is more limited than agencies admit. NDAs, team compartmentalization, and staff turnover make cross-pollination of knowledge more of a potential than a constant. Agencies that do manage to systematize that knowledge have a real competitive advantage; those that mention it in their pitch but do not operationalize it, do not.

The decision depends on total paid media budget, channel complexity, and internal capacity to absorb the function. For companies with budgets below $50,000 per month, the fixed cost of a full internal team rarely justifies itself. For companies above $200,000 per month, the equation inverts. The midpoint is where the decision is hardest, and where agencies that truly deliver strategic value have their best opportunity.

The transition has costs that optimistic projections tend to omit. For a mid-sized agency in LATAM, the order of magnitude is between $30,000 and $80,000 in the first year across licenses, automated workflow development, and training. This does not include the temporary productivity drop during the transition, which can last 2 to 4 months. The payroll savings can offset the investment in 12 to 18 months, but that assumes the agency maintains its client base during the process, which is not guaranteed.

The talent problem runs deeper than the "reskilling" narrative suggests. There is a real gap between knowing how to manage Google Ads and knowing how to interpret how a macroeconomic trend affects the strategy of a retail client. Available experience suggests that roughly a third of current execution talent at agencies may not be retrainable, with human and operational implications that macro projections do not capture.

The client trust tension is particularly acute in LATAM. A CMO paying $15,000 per month for an agency with 4 assigned people reacts differently when the agency communicates that they now have 1 person and 3 automated systems. The perception that "fewer people = less service" is deeply rooted in markets where personal relationships weigh significantly in client retention. Agencies that manage this transition well will need to invest as much in client communication as in technology.

The governance and risk question is the least developed and perhaps the most important for companies with compliance processes. When an automated system manages campaigns, who is responsible if an ad appears in an unwanted context? Performance Max already presents this problem with its lack of control over placements. More automation without guardrails amplifies it. There are no industry standards for governance of automated systems in paid media, which leaves a vacuum that each agency and client is filling on an ad hoc basis. For regulated sectors or those with high brand sensitivity, this absence of a framework is a real brake on adoption, and rightfully so.

Finally, the existential tension: why doesn't the client just do this internally? There is no structural barrier, only operational advantage. Agencies need to maintain speed of adoption of new tools and capitalize on cross-client knowledge. This last differentiator is real when it materializes (seeing patterns across multiple verticals, identifying trends because they appear in several accounts), but team compartmentalization, NDAs, and staff turnover mean many agencies do not capitalize on it in practice. It is a potential advantage, not a guaranteed one.

This analysis makes sense for two groups. The first is founders and directors of paid media agencies, particularly in LATAM, who feel pressure from clients asking why they still pay the same while platforms automate more and more. The three-layer framework serves to diagnose which part of the current business is vulnerable and which part has a future. If the agency still bills primarily for execution hours, the urgency is high.

The second group is CMOs, Heads of Growth, and Marketing Directors who hire agencies and evaluate whether the value received justifies the cost. The analysis offers a language for that conversation: does my agency operate at the execution layer, the analysis layer, or the strategy layer? Do the reports I receive add value I could not generate internally? If the answer is ambiguous, that is already information.

Where the analysis has less utility is in enterprise contexts with marketing teams of more than 50 people and holding group agencies. Those relationships operate under procurement, compliance, and scale dynamics that this diagnostic does not address in depth. It also applies in limited fashion to agencies that have already completed the transition or to paid media freelancers who face a different dynamic.

There is no broad quantitative study on the adoption rate of automated tools in paid media agencies in LATAM. Available data comes from markets in the United States and Europe, where the operational reality differs significantly in team scale, access to technology, and budget levels.

We do not know how long it will take platforms to make the human analysis layer redundant for most scenarios. The speed of improvement of AI models suggests it could be faster than expected, but real implementations are always slower than demos.

We also do not know whether the agency model with automation and human leadership will prevail, or whether something different will emerge: a marketplace of automated tools that companies contract directly, a model where the platforms themselves offer the strategy layer, or a polarization where only very large and very small agencies survive, eliminating the middle market.

The governance questions are completely open. Who is legally responsible when an automated system spends $50,000 on inappropriate placements? What audit standards should apply? Until these questions have answers, adoption in regulated sectors will be slower than automation enthusiasts project.

What is clear: the current model of charging for manual execution has an expiration date. Agencies that do not redefine their value proposition will discover that their clients have already redefined it for them.

Content assisted by AI and critiqued by three people who cannot agree on anything. This is the result.