The AI Investment Outlook // BRXND Dispatch vol. 73
Plus, Klarna-Salesforce Redux
You’re getting this email as a subscriber to the BRXND Dispatch, a newsletter at the intersection of marketing and AI.
Luke here. Hi everyone. In this edition, I’m going to share some thoughts about the ongoing drama about Klarna exiting Salesforce because I think it contains interesting lessons about how large companies get hooked on AI after a little experimentation. I’m also going to share a short recap of a popular conversation from BRXND LA, featuring Bloomberg Beta partner James Cham’s takes on AI investments, enterprise adoption, and more.
Speaking of last month’s event, we also heard from Airtable CEO Howie Liu about how AI is transforming document-heavy workflows—making complex processes faster, smarter, and more scalable. As a follow-up, Airtable is hosting a hands-on workshop in LA on 3/26, where you’ll be able to roll up your sleeves and build your own AI-powered app using their tools and guidance. At the last of these Airtable AI Accelerator workshops, Noah built a handy app for SOW-to-payment tracking which I use almost everyday.
Spots are limited—sign up here if you would like to join. If you need a refresher on Airtable, check out Howie’s presentation.
AI & The Enterprise
Last month at the BRXND LA conference, James Gross, co-founder of Alephic, had the chance to sit down with James Cham from Bloomberg Beta for a wide-ranging conversation on AI investment, enterprise adoption, and what’s actually working in the market today. (In this post, I’ll be referring to them as “JG” and “JC” for the sake of clarity).
JC brings a unique perspective as someone who’s been at Bloomberg Beta for 12 years and has both technical and management consulting experience.
The timing couldn’t have been more perfect, with major tech companies having just announced staggering AI infrastructure investments during earnings season: Microsoft ($80B), Google ($75B), and Meta ($65B) this year alone. Add to that the recently announced “Stargate” initiative with Donald Trump, Sam Altman, and Larry Ellison proposing $500B over four years, and we’re looking at unprecedented capital flowing into AI.
So what’s driving all this investment? And more importantly, what does it mean for marketers and brands?
We're in “Explore Mode”
JC made a critical observation about where we are in the AI adoption cycle. While there’s widespread recognition that AI is a transformative technology with broad applications, there’s still no dominant business model. “The terrible thing about AI right now,” JC noted, “is this is obviously a general purpose technology, but there’s not obviously a way to make a bunch of money.”
This uncertainty creates tension. On one hand, capital is flooding in because everyone sees the potential. On the other hand, as JC put it, “the moment capital comes in, they say ‘don't take too many risks, we really would like predictable revenue.’” But we simply haven't figured out the fundamental shifts in how business will operate with AI yet.
For JC, this means we’re firmly in “explore mode” rather than “exploit mode”—and that has major implications for how companies should approach AI projects. (Noah also spoke about the dichotomy between “explore and exploit” during his talk at the conference.)
The Sweet Spot for AI Projects
One of the most practical insights Cham shared was about AI project timeframes at large companies. He argued forcefully against big, multi-year initiatives, suggesting they’re almost certainly bound to fail given how rapidly the technology is evolving.
“We are at the point where it’s a terrible idea to do these one to three-year projects,” JC explained. “The real sweet spot is that sort of one-day to one-week project.”
This is a crucial perspective for marketers to hear. Rather than betting big on AI transformation projects, JC suggests hiring smart young talent to build targeted tools that deliver immediate value. It’s about creating something useful now rather than promising something revolutionary later.
Traditional ROI Thinking Is the “Worst Idea Right Now”
When JG asked about how companies should measure AI success, JC offered what might be his most contrarian take: traditional ROI thinking is entirely wrong for this moment.
“We are hard-nosed businessmen, we are measuring our ROI and being diligent about only working on the best projects—that is the worst idea right now,” he said, mimicking what he hears from CFOs.
Why? Because “the only thing you are certain of is that costs are going to go down” while the returns remain uncertain. By focusing solely on immediate ROI, companies risk missing the bigger opportunity as the technology matures.
A Shift in Business Models
One fascinating trend JC highlighted is how AI is blurring traditional business models. SaaS companies are moving toward services, while service businesses are developing software. Pricing models are shifting from subscriptions to performance-based approaches.
“You are seeing people who are making that bet and making a lot of money on it,” JC noted. He mentioned call centers that have switched to per-unit pricing and share the savings, allowing them to capture more value as AI improves their efficiency.
The Coming of Agents
As our time wound down, JG asked JC about the much-hyped “year of agents” that many are predicting for 2025. He offered two thought-provoking perspectives:
First, agent technology could create a new kind of vendor lock-in: “It is possible that 10 years from now you’re locked into your agent vendor because you have 50,000 agents running all sorts of different kinds of business logic that you have no idea what’s going on.”
Second, pivoting to the idea of call center agents, he noted that “the EQ of the average LLM is better than the EQ of at least the median guy, maybe the 75th percentile guy”—suggesting that emotional intelligence might be an unexpected strength of AI agents.
The big question for marketers? JC suggested we may soon spend significant time optimizing for AI models alongside human consumers—essentially “kissing up to” the models that increasingly mediate consumer experiences.
If you enjoyed this presentation, the entire lineup of programming from BRXND LA is now available to rewatch on our YouTube channel.
What Else Caught My Eye This Week
Around the time I started writing about AI last fall, a viral story about Klarna abandoning Salesforce captured the tech industry’s attention. Headlines suggested the Swedish fintech giant was replacing traditional SaaS with homegrown LLM solutions, triggering a predictable wave of hot takes. To some, this was the canary in the coal mine, proof that AI would soon dismantle the entire SaaS industry. To others, it looked suspiciously like pre-IPO theater: a company eager to signal cost-cutting prowess and AI sophistication to potential investors without much substance behind the claims.
The real story, as shared by Klarna CEO Sebastian Siemiatkowski this week on X, is significantly more interesting because it offers a detailed account of how SaaS consolidation actually happens inside organizations.
The saga started after Klarna encouraged employees to play around with AI, and in the process, discovered that feeding fragmented corporate data into LLMs produced poor results (“shit in, shit out”). Siemiatkowski recalled that Klarna’s knowledge was splintered across multiple SaaS platforms, each with their own concepts and structures, creating an “unnavigable web of knowledge that required a tremendous amount of Klarna specific expertise to operate and utilize.” In response, the company started to develop an internal tech stack to unify their data and knowledge, which eventually led to significant productivity gains when combined with AI. “The side consequence of this was the liquidation of SaaS—not all of them, but a lot of them. And not for the license fees, even though those savings have been nice, but for the unification and standardisation of our knowledge and data,” explained Siemiatkowski. Ultimately, the company shut down around 1,200 SaaS applications, including Salesforce.
With this story, I’m less interested in the strategic soundness of Klarna’s AI push or the age-old question of build vs buy than I am in the specific set of circumstances that led the company to reevaluate its tech stack.
Noah has written before about how turning unstructured data into structured data is the single problem AI is best at solving:
While we’ve spent years focusing on making better user interfaces for data entry and manipulation, AI is shifting the paradigm entirely. The future isn’t about building better forms—it’s about seamlessly converting unstructured information into structured data, automating workflows, and connecting systems.
As more companies begin tinkering with AI, they’re inevitably going to hit the same wall Klarna did: each SaaS product exists as its own island of data with proprietary structures that actively resist integration. And AI is at its most powerful when it is acting as a bridge within a workflow, transforming data in one format into data in another. To Siemiatkowski’s point: The license fee savings of less SaaS are good, but they’re almost nothing compared to the massive productivity unlocked when you can create new connections between previously siloed datasets, workflows, and tools.
If any of these ideas resonate with you, we’ll explore them further at BRXND NYC on September 18th. The waitlist for tickets is now open. After two successful events in NYC, we’re excited to be bringing it back to where it all started—with some new twists to be announced.
That’s it for now. Thanks for reading, subscribing, and supporting. As always, if you have questions or want to chat, please be in touch.
Thanks,
Luke