Before joining Dalarna Digital, I spent over a decade setting strategies for some of Sweden’s largest media buyers and advertisers. Although every media strategy differed from brand to brand, one discussion was almost always present: why spend money on brand building when sales-driving advertising is what actually converts and drives business? What I often returned to in my effort to explain the importance of brand building was the theory of “the CAC valley of death” – one of my absolute favorite theories, both in name and content.
Essentially, it’s about companies that heavily focus on conversion-driven marketing and its optimization, which naturally often succeed in lowering their Customer Acquisition Cost (CAC) over time. Simply put, the more you spend on converting communication, the lower the cost for a new customer becomes. And once you’ve reached a truly low cost per new customer, the temptation grows to focus even more on this part of the communication pie. You reduce your brand-building efforts—because why would you need them when the cost for a new customer is so low anyway?
The catch here is that the moment you choke off the brand budget, you also begin to stifle your growth. It’s a slow process, but everyone who embarks on this seductive path meets the same fate—the bucket isn’t being refilled with new customers, the cost of acquiring a new customer increases as the bucket’s contents dwindle, and eventually, you have very effectively killed your own business. The CAC valley of death, scary stuff.
Now that I’ve stepped into Dalarna Digital and, in doing so, partly shifted my professional focus towards more tech and digital transformation, I see a similar phenomenon, albeit in a new guise. Unsurprisingly, AI is involved (as it is in most things connected to our industry right now), and just as with the CAC issue, it is short-sighted decisions without long-term consequence thinking that once again form the hub of a new potential ‘valley of death’.
Here’s how: The absolute majority of all Swedish (and international) organizations are built upon an organizational theory nearly a hundred years old (a truly hefty work by Gulick and Urwick, for the curious…). The basic principle is that organizations thrive best when they are clearly hierarchical. Someone decides something, which is then delegated to someone else, who in turn delegates to someone who performs the actual task. The basic idea is that high specialization makes us efficient. However, the result is often inefficient organizations, an extreme number of middle managers, and coordination ad absurdum. Just a normal day at work for many, simply put.
When agentic AI now enters the scene, all of this is turned upside down. Something that previously required large coordination efforts, long lead times, and often entire departments to decide and execute, can now be performed by a single individual with the right agent support, in no time. Or even by the agent itself, completely without human involvement. In other words, actually building functions, testing them in practice, and adjusting what doesn’t work is often far cheaper than holding meetings about whether the function should be built. The logic is reversed—previously, heavy investment went into preparatory work, coordination, and decision-making. Now, we can get straight to the point—true “test and learn,” correctly executed with low financial risk and a potentially enormous reward.
The catch here is that the vast majority of organizations still haven’t grasped this logic. They continue to hold meetings about AI’s existence or non-existence. They layer AI on top of existing working methods (chatting and getting direct answers is fun and significantly reduces the time it takes for me to write a blog post…!). And most importantly—when they conclude in one of their many meetings that the value creation generated by the somewhat routine AI chatting isn’t high enough, they shut down the AI initiative in question and move on to the next item on the agenda. The result: most AI investments fizzle out, cost money, and are judged not to provide any direct value.1
This is where the logic surrounding “the CAC valley of death” comes back into the picture. My absolute conviction is that companies that prematurely extinguish their AI initiatives (believing they are trimming unnecessary overhead costs) will meet the same fate as the conversion crowd who extinguish their brand investments. It might take a few years, perhaps even a bit longer, but sooner than most realize, AI-mature organizations will have an unassailable head start compared to those who are late to the party.
What is required, then, to be early to the party? First and foremost, obviously, the willingness to dare to change, to truly invest in AI, to embrace rather than dismiss the technology. But perhaps even more: operational support. Many want to dare, want to invest, but don’t know how and lack the resources or expertise to delve into the technology. Add to that complicated legislation and moral considerations, and the whole thing starts to get quite thorny.1
Good thing then that there are those who take these issues seriously, who dare to think big, invest, and commit time and resources to building and developing. Who sort out the tangled legal aspects, navigate the question of responsibility, and build an infrastructure that enables the secure use of AI in practice.
We are working on something behind the scenes, something big that we believe will genuinely help many companies—not through grand visions and vague tech talk with dubious practical utility, but through operational support that delivers actual business value. Something that solves bottlenecks, relieves heavy administrative processes, and, not least, frees up time. Our vision is to future-proof Dalarna, and we will soon have the tools that make it possible. Stay tuned..
