AI & Performance Max: 28% CPL Drop in 2026

Pairing Performance Max with an AI agent completely changed how we hit conversion goals for a recent client. It’s a move away from the old-school manual tweaking and into a much more autonomous, data-first way of running campaigns. This case study breaks down what happened when we layered a strategic AI on a B2B SaaS campaign, showing how it boosted efficiency and ROAS and demonstrating the real power of this kind of intelligent automation.

Campaign Overview: B2B SaaS Lead Generation

We took on a client, a growing B2B SaaS firm in the project management space, who needed to seriously scale up leads for their enterprise product. Their old search and display campaigns just weren’t cutting it, with Cost Per Lead (CPL) all over the place and not enough reach. The goal was simple: get the CPL under $150 and hit a 3:1 ROAS. We were targeting decision-makers at North American companies with 500+ employees, and with a monthly budget of $75,000 for the three months between July and September 2026, our initial target was 500 qualified leads a month.

We built the campaign on Google’s Performance Max to get our ads across all its channels (Search, Display, Discover, Gmail, YouTube) from one place. The real secret sauce, though, was our proprietary AI agent overlay. This AI was an active manager, constantly in the weeds of the campaign’s daily operations and making tiny adjustments based on live performance data and its own predictive models. It was doing everything from tweaking bids on the fly and reallocating budget between asset groups to flagging creative that needed to be changed based on how users were engaging with it.

Creative Approach: Dynamic Storytelling with AI Insights

For the creative, we went with a classic problem/solution approach to show the software’s value. We built out a full asset library: five 30-second video ads for different use cases, ten high-quality image assets showing off features, and twenty headline and description variants. The key here was having our custom AI agent plugged in to optimize all of it. It was constantly crunching user interaction data like video view-through rates, image CTRs, and conversion rates for every single headline/description combo we threw at it, breaking it all down by audience segment.

Here’s a real example. We had one video, “Simplify Your Workflow,” that was killing it with audiences in the manufacturing sector. It was getting a CTR of 1.8%, way above our 1.2% campaign average. The AI spotted that trend and automatically pushed more budget to that video specifically for the manufacturing asset groups. On the flip side, an image of a dashboard was bombing with our tech sector audience. The agent flagged it and told us to swap it for something showing a benefit, not just a feature. This kind of feedback loop meant we were iterating on creative constantly, something a human team could never do at this speed or scale.

Targeting Strategy: Beyond Basic Demographics

We used Google’s standard audience signals in Performance Max as our foundation, but our AI agent is what really sharpened the targeting. We fed PMax a ton of signals to work with, custom segments built from competitor websites, our client’s first-party CRM data on their best customers, and basic website visitor remarketing lists. The AI then got to work, connecting those signals to actual conversion data to find patterns a person would likely never spot. For instance, it figured out that execs who had been on certain industry forums in the last 30 days had a 2.5x higher conversion rate compared to people who had just visited a competitor’s site.

Once it had that insight, the AI immediately started shifting more budget toward those high-intent forum visitors. It didn’t stop there. It also found small geographical pockets, like the financial district in Toronto, Ontario, where one of our ad copy variants was working much better than anywhere else because it clicked with the local business culture. Getting that granular, which was all powered by the AI bidding agents, is what made our ad spend so much more efficient. We were hitting the right people, in the right place, with the right message.

Performance Analysis: What Worked, What Didn’t

Over three months, the campaign pulled in 18.5 million impressions and brought in 1,620 qualified leads. The final numbers were solid. Our overall CPL landed at $138, well under the $150 target we’d set. We also hit a 3.8:1 ROAS, beating our 3:1 goal. An average CTR of 1.45% across all placements showed that the creative and targeting were really connecting, which isn’t always easy in a crowded B2B market.

The single biggest reason for our success was the AI agent’s real-time budget reallocation. It let us jump on opportunities instantly and cut our losses on underperformers just as fast. For example, in the second month, the AI noticed a sudden spike in conversions coming from YouTube In-Stream ads aimed at a few software review channels. In just a couple of hours, it automatically moved daily budget away from some Display placements where CTR was dipping and pumped it into those hot YouTube spots. That kind of agility is something you just don’t get with manual campaign management, where you might be checking performance once a day and completely missing these critical intra-day shifts.

So what didn’t work right out of the gate? The client’s generic image assets. They weren’t made for digital ads, and engagement was low. The AI flagged this almost immediately, which led us to pause those assets and get to work creating better, ad-specific visuals. This just showed that the quality of your inputs matters. The AI won’t turn bad creative into good creative, but it will prove it’s underperforming much faster than a human could. We saw a similar thing with some of our initial headlines that were too focused on tech specs instead of business benefits. They had terrible CTRs, and the agent’s reporting loop pushed us to rewrite them to focus on outcomes.

Optimization Steps and AI Agent’s Role

The AI agent influenced our optimization process constantly. A big piece of this was how we handled negative keywords. Performance Max is meant to be broad, but our AI analyzed search queries that got impressions but zero conversions and found a lot of junk. For instance, it flagged searches like “free project management templates” and “student project software”, obviously unqualified traffic we didn’t want to pay for. We took those terms and added them to the account-level negative list, which immediately improved impression quality and cut down wasted spend. PMax has its own logic for this, sure, but the agent acted as an extra layer of smart filtering that really paid off in a specific B2B environment like this.

Another big win was the AI-driven sequencing of our video assets. The agent figured out that people who saw a short, problem-focused video first (like our “Are Your Projects Falling Behind?” ad) were much more likely to convert if they later saw a longer, solution-focused video (like “Our Software: Your Project’s Success Story”). So, it started dynamically changing the ad serving order for different audience segments to create this mini narrative funnel. Is this something you can do with standard campaign settings? Not a chance. It’s a level of user journey mapping that requires serious processing power.

The agent was also indispensable for anomaly detection. At one point, it caught a sudden, weird spike in impressions coming from a geographic area we weren’t even targeting. It shot an alert to the team and, following our pre-set rules, automatically paused ads in that region until we could manually check it out. It turned out to be an anomaly, not some hidden market. That quick action prevented a major budget leak on junk traffic and probably saved us around $2,500 over a 48-hour window. With how complex PMax can be, this kind of proactive, automated monitoring is becoming fundamental for running campaigns effectively in 2026.

A recent IAB report says 72% of marketers think AI will have a major effect on their campaign optimization in the next two years. Our experience here backs that up completely. These AI agents are practical tools that get real results, right now. They let campaign managers stop worrying about constant, tiny manual adjustments and start focusing on high-level strategy, where they should be spending their time.

Conclusion

Putting a smart AI agent on top of Performance Max was a winning formula for this B2B SaaS campaign. The agent automated the hard stuff, bids, creative choices, and audience targeting, and delivered a CPL of $138 and a 3.8:1 ROAS. For digital marketers, it’s becoming clear that using AI solutions like this is necessary to wring more efficiency and performance out of campaigns and keep your strategies competitive.