Most marketing teams are chasing the wrong milestone. The LinkedIn feed is full of announcements about AI agents managing campaigns autonomously, and the instinct is to build one immediately. That instinct will cost you six figures and six months before you realize the foundation wasn't there.
Robert Simpkins at Search Engine Land published a roadmap yesterday that cuts through the noise. After a year building agentic systems for Google Ads, he's identified a pattern: the organizations that see commercial value follow roughly the same journey, while the ones that struggle skip straight to the expensive part. His framework deserves a CFO-grade breakdown because the sequencing determines whether you're buying time-to-learning or just buying toys.
The Foundation Problem Nobody Wants to Solve
Before any AI touches your campaigns, you need two things in place: a knowledge base and clean data. This is the stage most teams skip because it's unglamorous. But here's the math that should change your mind: Ryze AI's 2026 data shows advertisers using AI agents see an average 34% improvement in ROAS compared to manual management. That lift doesn't come from the model. It comes from the context you feed it.
Your knowledge base needs to document products, services, business rules, tone of voice, campaign structure, and internal processes in a format AI can parse. At the same time, your marketing data needs to be accurate, connected, and accessible. Whether you use BigQuery or another centralized warehouse matters less than eliminating the silos that stop AI from seeing the full picture. An LLM can't make sensible decisions if your business knowledge is scattered across Notion pages, Slack threads, and someone's head.
The misconception that AI compensates for poor processes is expensive. In reality, it automates those processes faster. If your attribution is broken, AI will optimize toward broken signals at scale.
Exhaust Off-the-Shelf Before You Build
You don't need developers to start benefiting from AI. Most Google Ads teams haven't exhausted what today's off-the-shelf tools can already do. Stormy AI's 2026 playbook notes that small businesses using Google's conversational experience are 63% more likely to publish campaigns with "Good" or "Excellent" Ad Strength. Improving Ad Strength from "Poor" to "Excellent" yields an average 12% lift in conversions.
Start by exporting campaign data into ChatGPT or Claude. Ask it to audit account structure, identify wasted spend, surface search term opportunities, or review your shopping feed. The results won't be perfect, but they'll reveal whether your data is clean enough to support more sophisticated automation.
Fluency's 2026 advertiser survey shows Performance Max adoption jumped from 60% to 71% in a single year. That's not because PMax is magic. It's because teams finally had the data hygiene to let automation work. The same principle applies to AI agents: the tool is only as good as the inputs.
Custom Automation Comes Third, Not First
Once you've exhausted off-the-shelf capabilities and identified specific gaps, you can start building custom automation. The key word is specific. You're not building an agent because agents are trendy. You're building one because you've identified a repeatable workflow where human judgment adds minimal value and the cost of errors is low.
Alexander Perleman at Groas draws a useful distinction: an AI agent has four defining characteristics. Perception (reading campaign data, analyzing search terms, detecting anomalies). Reasoning (determining why CPA rose 30% over 48 hours). Action (making the change directly). And continuous operation without waiting for permission. That's a specific technical definition, not marketing language.
The gap between Smart Bidding and a true AI agent is significant. Smart Bidding optimizes bids within predefined parameters. An AI agent decides whether to use automated bidding at all, evaluates performance against your business goals, manages budgets across campaigns, handles keyword strategy, and coordinates with other marketing channels. If you can't articulate which of those capabilities you need and why, you're not ready to build.
Full Agentic Systems Require Governance
The final stage is deploying AI agents that operate with genuine autonomy. This is where the CFO conversation gets serious, because autonomy without governance is a liability.
Google's May 2025 announcement introduced Marketing Advisor, an AI agent that lives in the Chrome browser and can recommend and apply strategies across multiple lines of business. The agentic expert in Google Ads offers personalized recommendations for new and existing campaigns and can implement them on the advertiser's behalf. That's real autonomy, and it requires real controls.
Omnibound's 2026 adoption statistics reveal the execution gap: 87% of marketers now use generative AI in at least one recurring workflow, but only 6-30% of marketing organizations have fully integrated AI across their workflows. 74% of companies struggle to achieve and scale value from AI initiatives. Adoption is high. Execution maturity is not.
The governance question isn't theoretical. When an agent reallocates budget at 2 AM based on signals you didn't anticipate, who's accountable? What's the rollback procedure? How do you audit decisions that happened faster than any human could review? These aren't edge cases. They're the operating reality of agentic systems.
The Sequencing Is the Strategy
The roadmap isn't four steps because four is a nice number. It's four steps because each stage creates the preconditions for the next. Clean data enables off-the-shelf tools to work. Off-the-shelf tools reveal which custom automation is worth building. Custom automation teaches you the failure modes you'll need to govern when agents operate autonomously.
Hooked Marketing's 2026 benchmarks show the average conversion rate on Google Ads is 7.52%, more than 3x the global PPC average. Demand Gen campaigns saw a 26% increase in conversions per dollar thanks to AI-powered optimizations. The lift is real. But it accrues to teams that did the foundation work, not teams that skipped to the agent announcement.
The CFO question isn't whether AI agents will transform Google Ads management. They will. The question is whether your organization has the data hygiene, process documentation, and governance frameworks to capture that value, or whether you're about to automate your existing problems at machine speed.
