There is another AI agent announcement every week. Google is building them, vendors are selling them, and LinkedIn would have you believe every marketing team will soon have an autonomous employee running campaigns around the clock.
It is easy to conclude the next competitive advantage is building an AI agent as fast as possible. Robert Simpkins, co-founder of Propel, argues from a year of building agentic systems for Google Ads that this is the wrong frame. The businesses that see genuine commercial value follow roughly the same journey — and the ones that struggle usually skip straight to the expensive part.
Before experimenting with AI, invest in two things: your knowledge base and your data. This is the least exciting stage, which is probably why it is the one most often skipped.
One of the biggest misconceptions about AI is that it compensates for poor processes. In reality it automates those processes faster.
The quality of any AI system depends far less on the model you choose than on the context you give it. Even the most capable LLM cannot make sensible decisions when business knowledge is inaccessible and data is fragmented across platforms.
In a format AI can consume:
In parallel, make marketing data 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.
You do not need developers to start benefiting from AI. Simpkins is blunt about it: most Google Ads teams have not yet exhausted what today's off-the-shelf tools already do.
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. Current models are remarkably capable at analyzing large datasets and routinely uncover patterns that used to take hours in spreadsheets.
Next, connect those tools to Google Ads, Google Analytics or Google Merchant Center through pre-built Model Context Protocol (MCP) connectors. Instead of exporting spreadsheets weekly, you query live account data while retaining persistent business context through projects or custom GPTs.
For many organizations this combination delivers the majority of the value they will ever need. Build only when you have genuinely reached the limits of this setup.
Eventually you may outgrow off-the-shelf tools — when requirements get specific.
You might need to combine advertising performance with stock availability, pricing, margin and CRM data. Or have AI continuously monitor accounts rather than waiting for prompts. Or automate approval workflows while retaining appropriate controls.
That is when custom development becomes worthwhile. Developers make AI systems more reliable: custom MCPs, guardrails, orchestration, scheduling and cost optimization together turn an interesting demo into a system dependable enough to use every day.
Ironically, the biggest obstacle to AI adoption rarely has anything to do with technology. It is people.
Organizations that progress fastest do not expect every employee to become an AI expert overnight. They identify enthusiastic early adopters and give them space to experiment, then have those people share what works so the workflows gradually embed across the wider team.
Rather than replacing marketers, AI changes how and where marketers create value. The last decade already handed execution to algorithms through Smart Bidding, broad match and Performance Max. Agentic AI is the next stage of that evolution, and the marketer's role keeps shifting from manual execution toward strategy and judgment.
The biggest mistake Simpkins sees is businesses trying to automate every aspect of Google Ads. That is not where the value is.
Agentic AI is exceptionally good at repetitive, data-heavy work: auditing accounts, monitoring performance, analyzing trends, surfacing optimization opportunities. Take those off experienced marketers' lists and they spend more time on strategy, creative problem-solving and business objectives.
The teams that outperform over the next few years will not necessarily have the most sophisticated agentic AI. They will be the ones who understand where AI creates leverage, where human judgment still matters, and how to build the foundations that let the two work together.
The gap worth sitting with is the one between stage one and stage two.
Most advertising operations run across fragmented platforms with different reporting structures, stitched together by a person and a spreadsheet. Attach an agent to that and, in Simpkins' framing, you have simply automated a broken process faster.
The advice is also well timed. Google has stacked the AI Max migration and the removal of Search language targeting into the same month, and each one strips out another manual control. As that continues, the leverage available to operators moves away from platform settings and toward the context and data you feed those platforms. See Google Ads Is Removing Language Targeting From Search Campaigns for the same pattern.
For the same logic applied to search visibility, Most AI Visibility Gains Are Just Technical Debt Repayment makes the parallel case: finishing the deferred homework is usually cheaper and faster than buying a new tactic.
The knowledge base and the data. Document products, services, business rules, tone of voice, campaign structure and internal processes in a format AI can consume, and eliminate the data silos that block a full picture.
Export campaign data into ChatGPT or Claude and ask for account structure audits, wasted spend analysis, search term opportunities and shopping feed reviews. Pre-built MCP connectors then let you query live Google Ads, Analytics and Merchant Center data directly.
When you need to combine ad performance with stock, pricing, margin or CRM data, have AI monitor accounts continuously without prompting, or automate approval workflows while keeping controls in place.
The argument here is that it shifts the role. Repetitive, data-heavy work moves to AI, and marketers move toward strategy and judgment. Trying to automate everything is presented as the most common mistake.
To apply what you just read to your own site, start with a free audit of where things are now.
A strategist replies within 24 hours on business days.
A new Enhanced matching option is appearing in Customer Match settings, extending first-party audience reach by matching consented users with consented publisher users.
Google is adding multi-campaign A/B testing, experiments that keep brand and location controls on, and expanded Performance Planner forecasting to AI Max for Search campaigns.
Demand Gen starts where there is no keyword and no existing interest. As that protection disappears, PPC specialists need creative strategy, audience research and full-journey measurement.