Filing taxes is one of the more begrudged tasks in many countries. The market environment is correspondingly challenging – with complex legal requirements, high localization demands, tight margins, and government-provided options shaping the framework.
In the midst of this challenging environment, Taxfix, a leading European fintech company for digital tax solutions, wants to position itself with maximum efficiency and is therefore pursuing a clear strategy: "We are an AI-first company," says Alex Beresford, Chief Growth Officer and CMO at Taxfix. "This means that we implement our mission, vision, and value creation with AI solutions. Our focus is always on creating even better offerings for our users." In practice, this means: Taxfix first tries to think about everything with AI and then refutes it if it doesn't work. But, according to Beresford, "60 to 70 percent of tasks in marketing, sales, and partnerships can be done AI-first today – in some cases, even better."
Taxfix already saw initial success with Google AI in 2022, when the team tested Performance Max for customer acquisition. The automated campaigns achieved nine-percent more tax returns during the highly competitive tax season – at a 44-percent-lower cost per conversion compared to pure search campaigns. Since then, Performance Max has been an integral part of the marketing mix.
An order of magnitude in creativity
Since 2025, Taxfix has also been testing the AI-first approach in the production of creatives. In video production especially, iterative work, as is necessary for data-driven campaigns, has been hardly feasible until now. The classic agency process has always been complex, costly, and comparatively slow: Weeks often passed from the first storyboard and detailed styleframes until audio mixing in post-production. The necessary speed to systematically test creative variants – for example, as part of Brand Lift Surveys – was simply lacking.
Taxfix therefore asked itself the question: Can creative ideas be implemented faster and just as effectively with AI compared to classically produced videos? To find out, the company, in collaboration with Google, started a large-scale experiment at the beginning of 2025 with Veo 2 – a novel, generative video tool. The marketers ran human-produced stories, AI-generated clips, and hybrid variants, in which AI supports the marketers, against each other – including in different languages. "We have many multilingual customers and we didn't know, for example, whether a creative in their native language would work better than in their second language – or vice versa," says Beresford. "So we tested that out."
Taxfix tested its videos created with Veo 2 on YouTube – with great success
In the first test run on YouTube, the financial company was primarily concerned with analyzing the effects of the various commercials. As part of different Brand Lift Surveys, the effects on the metrics Video Ad Recall and Consideration were examined. For this reason, Taxfix chose Video View Campaigns (VVCs) as the campaign type, because this allows the targeted maximization of genuine views and the effects on brand perception and purchase intent to be measured with particular validity. Four weeks after the start of the test balloon, Beresford took stock: "Large parts of our audience actually preferred the AI-first approach, but just as many found hybrid formats to be the most convincing," he says. "That is fascinating and was completely new for us."
Veo 2 reduces production time from eight weeks to a few days
It was also clear that the marketers using Veo 2 could implement their creative content at a fraction of the usual time and production costs – and in several language versions at the same time. "What used to take six to eight weeks, we can now do in a few days with Veo 2," says Beresford. In doing so, AI not only helps in implementing existing ideas, but also develops new motifs and narratives that go beyond the original campaign concept. "That gives us a lot of room to test," says the CMO. At the same time, the marketers also found that the AI does not perfectly stage every idea. "In the first script, someone was juggling chainsaws – that was so physically complex that Veo was overwhelmed by it," recalls Nishant John, Creative Director at Taxfix. The solution: simpler, surprising motifs that are based on familiar genres. The system worked significantly better following this principle.
The team had barely warmed up with Veo 2 by the time the next stage opened up: When an ad for the NBA finals, produced entirely with Veo 3, appeared at the end of June, it was immediately clear to CMO Alex Beresford: Taxfix should be one of the first companies in Europe to test the new tool. "I told our creative team: You will be the first in Europe to take commercials live with Veo 3 – on connected TVs and ideally also on linear television," says Beresford. Two days later, his employees presented the first video.
Veo 3: Why generative AI requires creative flexibility
However, the team had to rethink its approach to producing the creatives. Veo 2 still worked along classic production steps – with a script, scene breakdown, styleframes, separately generated video, and subsequent audio production. Veo 3, on the other hand, works with complex and detailed prompts: Image and sound are created simultaneously from text descriptions. "Unlike conventional video shoots, where everything is determined in minute detail in advance, video production with generative AI requires a more agile and adaptive mindset," says Creative Director John. "We had agreed on the script and scene plan, and then Veo 3 still did things that nobody expected. Creative flexibility wasn't optional; it was mandatory."
The team also learned new things about prompting. It proved particularly helpful to format inputs in JSON format or to have existing text prompts translated into machine-readable structures by an LLM like Gemini 2.5 Pro. CMO Beresford also recommends a multi-stage approach in the conception phase, where ideas are first sketched out cost-effectively with Imagen 4 and Veo 2 and then brought to final implementation in Veo 3. "Many creative processes start as a vague idea in your head. Imagen 4 and Veo 2 help to solidify these ideas before producing in Veo 3," says Beresford. Anyone who is still unsure about prompting should feel free to get help from the AI: "A model like Gemini 2.5 can help in formulating one's own prompts even more effectively."
Taxfix is currently testing the commercials generated with Veo 3 against other alternatives again: commercials created with Veo 2, commercials implemented by humans alone, and hybrid versions. This time, the marketers do not want to measure the effects on Video Ad Recall and Consideration, but on performance metrics like Website Action or App Action. They therefore chose Demand Gen as the campaign type. Even if the results are not yet available, one thing is clear for Beresford: Veo 3 is raising the bar in video production in terms of quality, speed, and cost-effectiveness. "If Veo 2 was a turning point, then Veo 3 is a complete paradigm shift," he concludes.
YouTube as a testing ground, CTV as a stage
But even the best video only works where real attention is possible. "I don't just want a view, I want a Value View," says Beresford. That means real attention instead of fleeting visual contact. To achieve this effect, Taxfix relied on YouTube for distribution. The platform offered ideal conditions for testing different video approaches against each other in a controlled manner – with measurable impact through Brand Lift studies.
Furthermore, YouTube offers the largest reach on connected TV worldwide – including in Germany. In the eyes of the CMO, this is another benefit of the platform. "If you show a video in the right context – on YouTube, in the living room, on the big screen – and your ideal customer is watching at that exact moment: That's the sweet spot," he says. This is where attention, relevance, and target group precision meet, making YouTube and CTV the perfect stage for one's own creation.
Plan consolidated, scale targeted – why Taxfix relies on DV360
In the past, however, consolidated CTV campaigns were difficult to implement given the enormous fragmentation of the market. Here, too, Taxfix is now pursuing a radical course. "You can't do clean impact measurement if your teams are booking differently or working with different strategies," says the CMO. "Therefore, every video view we buy now runs through DV360. This allows us to control all channels from one platform in a comparable and efficient way." Separate media plans for Disney+, YouTube, or Netflix? A thing of the past. Success proves Beresford's strategy right: In June 2025 alone, Taxfix was able to increase its video reach by 56 percent without spending more budget. "This increase in efficiency is made possible by the consolidated media purchasing via DV360 and the cross-inventory optimization of ad frequency," emphasizes Beresford.
In June 2025, Taxfix increased its video reach in DV360 by 56 percent – without a budget increase.
In addition to platform consolidation, Taxfix also relies on AI-supported modeling for success measurement to close measurement gaps. For this, the scale-up has activated Consent Mode and Enhanced Conversion. The Consent Mode tool models conversions based on the consent status and provided Taxfix with a 20-percent plus in measured conversions.
Although artificial intelligence at Taxfix is optimizing more and more processes with increasing success, it is important for the marketing managers to stress: It does not take over strategic thinking and creative responsibility. "AI is not an autopilot," says Creative Director John. "It provides suggestions – but the decision about what fits the brand lies with us." For Beresford, it is also clear: Those who work with AI do not give up control – they win it back. "I remain in the driver seat," says the CMO. AI enables more variants, more speed, more testability – but notably without strategic judgement. "It's not about scaling blindly, but about steering in a targeted way. That requires people who know where they want to go."
