A first-of-its kind, expert-verified AI for better Google Ads performance

A new AI agent enables deeper performance optimization at the Norwegian marketing agency Brist, by scaling senior expertise and shifting time from analysis to impact. In pilot use, the solution contributed to around a 20 percent improvement in core KPIs for one client, with a potential value impact in the multi-million NOK range for larger advertisers.

For Brist, innovation is about more than efficiency. It is about protecting and scaling expertise in a market where complexity is increasing and time is always limited. Together with Antire, the performance marketing agency has developed bristIQ, an expert-verified AI agent designed to support better decisions in Google Ads based on real experience, not generic AI assumptions.

Founded in 2018, Brist is a digital agency built entirely on senior expertise. All specialists have long experience in their respective channels and work closely with clients who demand measurable results and clear accountability. This senior-only model ensures high quality, but also creates a structural challenge: experienced expertise is scarce, and much of the available time is spent on analysis rather than improvement.

"Even with senior specialists, 70 to 80 percent of the time goes into analysis, and only 20 percent into actual changes. That’s a paradox, because those changes are what create value for clients." says Thor Christian Valle, co-founder of Brist.

When generic AI is not good enough

AI has rapidly entered the marketing industry, and many agencies now rely on generic AI tools to analyze data and suggest optimizations. After testing and quality-checking several of these tools, Brist took a more cautious approach.

"I asked the same Google Ads questions to different AI bots and got completely different answers. Some failed as soon as I asked follow-up questions. Others suggested actions that aren’t even possible." says Valle.

Most generic AI tools pull unverified information from the internet and lack context, making them unreliable in complex performance environments. For Valle, this reinforced an idea he had been thinking about for years: using AI to systematize and scale the senior expertise he had built up over nearly two decades.

"AI is clearly better than humans at processing large amounts of data. But that doesn’t mean it understands how decisions should be made. The challenge wasn’t competence, it was capacity. Expertise had to be systematized in order to scale." says Valle.

Brist chooses Antire to build an expert-verified AI agent

Instead of connecting to existing AI platforms, Brist chose to build their own AI agent together with Antire.

The goal was clear: create an AI that works like a senior advisor.

"This is not just a wrapper around ChatGPT or other language models. It is a fully fledged AI agent with access to Google Ads performance data, a verified knowledge bank of Thor Christian, and the latest Google Ads rules and regulations, where all components are orchestrated so that it can play as a brainstorming partner for senior analysts." says Peyman Kor, AI consultant at Antire, who holds a PhD in Artificial Intelligence.

The project initially started as part of Antire’s AI Strike Team, a pilot program designed to rapidly test and validate AI concepts. As the solution proved its value, the work transitioned into Antire’s AI Value Center, where Peyman Kor is now part of the core team further developing and scaling the solution.

"Translating what Thor Christian wanted into a working AI solution was not straightforward." Kor explains.

The final solution is built using a deep agent architecture. When the agent is asked a question, it first determines which data it needs to examine, retrieves both structured performance data and relevant expert knowledge, and then iterates between them until it concludes.

The agent is completely disconnected from the internet.

"All analysis is based solely on verified documents and data provided by Brist and Thor Christian. No web searches, no external sources." says Kor.

One expert’s knowledge systematized

At the core of bristIQ is Thor Christian Valle’s documented experience from nearly 20 years working with paid search for some of Norway’s and the world’s largest brands.

The knowledge base includes more than 1,000 pages of expert documentation, hundreds of decision trees and cause–effect models, clearly defined performance rules, and customer-specific context such as KPIs, margins, and return rates.

"After so many years, you know where to look when performance changes. That experience is now built into the agent." Valle says.

In one customer account, the agent analyzes more than 25 million data points, which is far beyond what a human specialist can process manually.

"It can identify issues even inside campaigns that appear to perform well on the surface. That’s where some of the biggest improvements often hide." Valle adds.

Human-in-the-loop by design

Brist deliberately chose not to automate changes in ad accounts.

"This is a hybrid model. AI is excellent at processing data. Humans are better at context, judgment, and customer understanding." says Kor.

The agent analyzes, prioritizes, and explains. The specialist decides.

"For example, the AI might suggest increasing YouTube ads based on data. But the customer may not want that. Human judgment is essential." Valle explains.

The goal is trust. bristIQ is designed to feel like a senior advisor, calm, precise, and reliable. Like an extension of the specialist, and a colleague you can trust.

Better use of time, and better results

The impact on Brist’s daily work has been significant. Time spent on analysis has dropped from around 70 percent to 20 percent, freeing up more time for value-creating improvements. In test accounts, Brist has seen 20–30 percent improvements in key performance indicators, with a clear correlation between AI recommendations, implemented changes, and results. For larger accounts, performance improvements at this level can represent several million kroner in potential annual value.

"In a large customer account where we have been testing the agent for a full month, 80 percent of recent account changes were implemented based on the agent’s recommendations. During the same period, spending increased by just 3 percent, while leads grew by 30 percent and cost per lead fell by 30 percent." says Valle.

For Brist, this means specialists can work deeper in each account, customers receive more value, and the agency can operate more sustainably.

A close and dedicated collaboration

Developing bristIQ required intense collaboration between Brist and Antire.

"I’m very exposed in this process. We’re essentially cloning my expertise, which makes me extremely demanding." Valle says.

He emphasizes the role Antire and Kor have played, both in the development and in maintaining a clear focus on business value throughout the process.

"Peyman has been incredibly invested. He goes deep into the material, challenges assumptions, and constantly pushes the solution forward. It has exceeded all expectations. He has practically become a performance marketing specialist along the way." says Valle.

For Brist, it was important to work with a partner that could handle sensitive data responsibly, ensure ownership of both code and knowledge structures, and build a solution that could evolve over time.

Bringing execution into the workflow

The next phase is to further integrate analysis and execution into a single interface, while keeping the human firmly in control.

"This is not about letting AI take over. It’s about building a tool you can trust" says Valle.

While bristIQ was built specifically for performance marketing, the architecture behind the solution could also inspire AI use in other knowledge-intensive industries where trust and expert judgment are critical.

For Brist, bristIQ is a way to preserve what matters most in performance marketing: expertise, judgment, and accountability - supported by AI where it truly adds value.