How Google Ads automation changed the agency value equation
The agency model for Google Ads was built on a specific labour exchange: the client could not run the account competently without dedicated expertise, so they paid the agency to provide it. That exchange is structurally different now. The expertise gap has narrowed sharply on the execution side, and widened on the strategic side.
Three developments drove this. First, Smart Bidding. Launched in its current form around 2018, it replaced the manual bid management that was a core agency service with auction-time machine learning across Target CPA (cost per acquisition), Target ROAS (return on ad spend), Maximize Conversions, and Maximize Conversion Value. An agency that spent billable hours adjusting bids by device, time, and keyword was doing work that Google's system does automatically and at higher speed with more signal.
Second, Responsive Search Ads (RSAs), which replaced the Expanded Text Ads format Google deprecated in 2022. RSAs take up to 15 headlines and 4 descriptions from the advertiser, then test combinations to find what performs. Creative optimisation, another traditional agency service, partially automated.
Third, Performance Max (PMax). Launched broadly in 2021, PMax is Google's all-in-one campaign type that serves ads across Search, Display, YouTube, Gmail, Discover, and Maps from a single campaign, with Google's AI making most placement, bidding, and creative decisions. Campaign architecture decisions, another traditional agency service, partially automated.
At Google Marketing Live in May 2026, Google announced AI Max for Search campaigns: extended AI-driven optimisation for Search that includes expanded keyword matching, creative personalisation from landing page content, and broader audience signals beyond the keywords the advertiser provides (blog.google, May 2026). Keyword research, another traditional agency service, further automated.
This does not mean agencies have no role. It means the role has shifted from execution to judgment, and not every agency has made that transition. The buyer's job is to identify which type of provider they are actually evaluating, not which label the provider uses.
Two roles remain genuinely human in 2026 Google Ads management. The first: deciding what objective to feed the machine. Performance Max and Smart Bidding optimise toward whatever conversion signal you give them. If you give them the wrong signal (form fills instead of qualified leads, or revenue instead of margin), the automation optimises aggressively toward the wrong thing. Setting the right objective requires business context the platform cannot access. The second: measuring whether the automated system is actually delivering incremental business value, or just capturing credit for conversions that would have happened anyway. Google's own measurement attributes those conversions to Google. An independent measurement layer is the only way to check that attribution claim.
Our post on AI agents for campaign optimisation covers the broader automation layer including what actually runs on autopilot and what still needs a human across the full channel stack, not just Search.
Three models for running Google Ads in 2026
The three viable management models in 2026 are traditional agency, in-house, and AI-native consultancy. They differ on strategy depth, how they handle platform automation, whether measurement is independent, and how transparent the engagement is. Use the tabs below to compare each model across six dimensions.
Traditional Google Ads agency
In-house management
AI-native consultancy
The comparison above covers qualitative dimensions deliberately. The cost of each model depends on scope, market, and account complexity in ways that make any published figure misleading. What holds across markets: in-house is competitive when the headcount already exists; agency is competitive when execution volume is high and strategic complexity is low; consultancy is competitive when the measurement problem or strategic complexity justifies a different engagement structure.
For the full Google Ads platform mechanics, the Google Ads platform page covers how leapbuzz runs campaigns within this consultancy model, including the specific Smart Bidding and Performance Max configurations we use. The broader performance marketing framework is on the performance marketing service page.
Which model fits your situation
Three variables narrow the decision more reliably than any other: spend maturity (the scale and complexity of the account), in-house skill (whether a dedicated search specialist already exists on the team), and regulated-sector requirements (whether compliance review is a non-negotiable part of the campaign workflow). The flowchart below maps those three variables to a model recommendation.
Decision variables: in-house specialist, independent measurement requirement, regulated industry. Traditional agency remains viable for high-volume, low-complexity, unregulated accounts with in-house oversight.
The flowchart deliberately excludes spend level as a primary decision variable. Volume alone does not determine which model is right. A regulated insurer running a modest Google Ads account in Singapore needs compliance-integrated management regardless of spend level. A direct-to-consumer brand with high volume and no regulatory complexity may do fine with in-house management. The variables that reliably discriminate are the ones in the flowchart.
One pattern worth naming: many organisations start with an agency, outgrow it (or find the measurement problem), and move to a consultancy engagement for the strategic and measurement layer while retaining or rebuilding in-house capability for day-to-day operations. That is not a failure of the agency model; it is a normal maturity progression. The AI performance marketing engagement is typically where organisations land when they have made this transition.
The measurement gap no model closes on its own
All three models have the same structural measurement problem: Google's reporting system is designed by Google, for Google's interests, and it attributes conversions to Google's channels. This is not a conspiracy; it is an incentive structure. When the platform both runs the campaigns and measures the outcomes, the measurement has a systematic bias toward showing the platform in a good light.
Performance Max made this problem acute. Because PMax runs across Search, Display, YouTube, Gmail, Discover, and Maps simultaneously, and because it optimises placements automatically, it is almost impossible to tell from the platform's own reporting which channel drove which conversion. The campaign reports conversions; the channel attribution is opaque.
Two independent measurement tools matter more in 2026 than they did three years ago, precisely because automation has made the platform's own measurement less transparent:
- Incrementality testing. A controlled experiment with a holdout group (people who see no ads) versus an exposed group. The difference in conversion rate between the two groups is the incremental lift the advertising actually caused, not the lift Google's attribution model assigns. This cannot run inside the platform's own tools without methodological compromises. The holdout must be managed externally, with the advertiser's first-party data as the ground truth.
- Marketing mix modelling (MMM). Statistical modelling that attributes business outcomes to marketing channels using historical data and economic priors, without relying on user-level tracking. In an environment where PMax and Smart Bidding are making placement and pacing decisions across multiple channels simultaneously, MMM gives you a channel-agnostic view of what is actually moving the business metric. Our earlier post on the marketing technology stack covers how MMM fits into a broader measurement architecture.
The analytics and insights engagement is where this measurement infrastructure gets built. It is also the engagement that tends to change the most about how clients read their Google Ads reports. The numbers do not usually look worse after independent measurement is introduced. They look different. Some campaigns that appeared flat in platform reporting show genuine incremental lift. Some campaigns with strong platform-reported conversions show low or zero incremental lift once holdout data is available. Both findings are useful; only independent measurement can produce them.
For search advertising specifically, the measurement problem has a particular shape. Search captures intent that already exists. A user searching for your brand or product may have converted anyway, without the ad. Brand keywords especially are prone to this: the ad shows, the user clicks, the platform records a conversion, but the user was already going to your site. Incrementality testing separates captured intent from created intent. It is the only method that answers whether your Search campaign is generating revenue or just taxing it.
One note on transparency and billing models: the measurement problem is worse when an agency's billing is tied to media spend (percentage of spend or cost-plus models), because the agency has a structural incentive to report good performance and increase budgets. A consultancy billed on fixed scope has no such incentive; the scope does not expand with spend, so the measurement finding can be honest about what is working and what is not. This structural alignment between measurement and billing model is worth examining when evaluating any provider, not just leapbuzz.
Multi-market considerations: Singapore, Australia, Canada, USA, Malaysia
Google Ads management in a single market with a clear regulatory context is genuinely simpler than managing across five markets with differing rules. leapbuzz operates across Singapore, Australia, Canada, USA, and Malaysia, and the regulatory and operational differences across these markets are substantive enough to affect the model choice.
| Market | Key regulatory factor | Labour market factor | Implication for model choice |
|---|---|---|---|
| Singapore | MAS notices apply to financial services advertising (FAA-N03, FAA-N06). PDPA governs data use in campaign targeting. MCI-IMDA digital advertising guidelines apply to certain categories. | Tight specialist talent pool. Senior search practitioners are expensive to hire and retain. Agency and consultancy density is high; quality is variable. | Regulated verticals strongly favour consultancy engagement. In-house is viable for consumer sectors with dedicated headcount. |
| Australia | ASIC regulates financial services advertising. RG 234 governs advertising for financial products. ACCC consumer law applies to claims and comparisons. State-level gambling advertising restrictions vary. | Strong in-house marketing culture in large enterprises. Agency market is mature; commoditisation is advanced. Consultancy model is gaining traction in financial services. | Financial services and regulated verticals favour consultancy. Consumer and retail are competitive in-house or agency markets. |
| Canada | Provincial securities commissions govern investment product advertising. FINTRAC rules apply to financial messaging. CASL governs digital outreach linked to ad campaigns. Bilingual requirements (English/French) in Quebec affect creative. | B2B and financial services sectors have strong in-house search capability in Toronto and Vancouver. Quebec requires bilingual creative at scale. | Multi-province campaigns with regulatory variation favour consultancy for compliance architecture. Single-province consumer campaigns are viable in-house. |
| USA | FTC Guides govern advertising claims and endorsements. SEC rules govern financial product advertising. State AG offices actively enforce deceptive advertising. Healthcare advertising has HIPAA implications for tracking. | Largest English-language search advertising market. Deep agency market and strong in-house specialist availability at scale. Consultancy model well-established for mid-market and enterprise. | All three models viable depending on vertical and scale. Healthcare and finance favour consultancy for measurement independence and compliance. |
| Malaysia | PDPA Malaysia governs data use. Bank Negara Malaysia regulates financial product advertising. Dual-language market (Malay/English) affects creative and targeting parameters. | Growing in-house capability in Kuala Lumpur. Agency market developing. Senior search talent more limited than Singapore. | Regional campaigns spanning MY and SG benefit from a single consultancy engagement covering both markets. In-house for local consumer brands is growing. |
The multi-market complexity argument for a consultancy over a traditional agency is specific: a consultancy that has operated in all five markets can design a campaign architecture that accounts for the regulatory and cultural parameters of each market from the start, rather than discovering them after the campaign goes live. A traditional agency typically specialises in one or two markets; multi-market scope requires either multiple agencies (coordination overhead) or a single agency with genuine multi-market depth (rare).
For regulated financial services specifically, the compliance review requirement is not optional in any of these five markets. The question is whether compliance review is built into the campaign workflow, or added after the creative is already live. An AI-native consultancy builds it in. Google's AI Max for Search, which can personalise ad creative from landing page content, requires a review step to verify that the generated creative does not drift from compliant language. That review step cannot be automated by the platform.
One practical note: Google's Performance Max and AI Max features are available in all five markets but their behaviour is not identical. Audience expansion signals, the inventory Google can access for PMax Display and Video placements, and the creative personalisation depth of AI Max all vary by market and account history. An account that has been running in Singapore for three years has more Signal than a new account in Malaysia. That historical signal gap affects how much of the automation's claimed capability is actually available to a given advertiser in a given market.
