The Agile Brand with Greg Kihlström®

The episode contains several solid, practitioner-level insights about data strategy preceding AI deployment and the importance of pre-building assets rather than expecting reps to self-serve. However, the core thesis - that foundational data gaps matter more than chasing better AI - is repeated multiple times without substantial new evidence or frameworks introduced after the midpoint. The conversation lacks granular specificity about *how* to solve data silos or what constitutes a 'complete' dataset.

“if you're asking them to do very general type, you know, Q and A, they're gonna go out and search, you know, everything that's available on the Internet. And that's probably okay. I think it's more when you start to try to get into, you know, what we call, um, expert type of requests, that you're only going to get responses based upon what data they have access to”

“if the CRM data is not very complete and it's pointing to that, or your er, system doesn't talk to the sales systems, um, if you've got information that's living in silos and it doesn't really have that connective tissue, that's really where I think once companies sort of get a handle on their own data estate, making that available to, you know, an enterprise version of these AI solutions to use internally is going to be extremely valuable”

The core argument - data quality precedes AI effectiveness - is not new and has been circulating in enterprise tech circles for 2+ years. The counterintuitive angle promised in the intro (that chasing better AI is a distraction) is somewhat fresh framing but the underlying insight is conventional. Ann's personal anecdote about the Slack/Looker time-savings is concrete but illustrative rather than revelatory. The episode largely reiterates standard data governance talking points without introducing novel frameworks or first-principles challenges.

“Whatever you're using, whether It's Claude or ChatGPT or Gemini, I, um, think they're only as good as the data set that they're built upon”

“what is the return on this investment going to be?”

Ann Davis is highly credible: 30+ years in tech sales, 7 startups scaled to $100M+, Looker founder (acquired by Google for $2.6B at 26x run rate), 5 years at Google Cloud, now CRO at Crunchbase. She is a practitioner who has actually executed at scale and sits in a role directly managing the ROI pressures she discusses. However, she is more of a serial founder/operator than a current deep expert in AI/ML infrastructure, which limits her ability to speak authoritatively on some technical data strategy nuances she mentions.

“I've been in tech sales for about 30 plus years. Um, I'm a serial startup person. I've done seven startups, um, taking them from everything from zero to a hundred million dollars”

“I was an IC, which I was one for 16 years, I would spend an exorbitant amount of time getting ready um, for a meeting”

The episode contains some specific numbers (Looker acquisition price $2.6B, 26x multiple; 31,000+ AI companies in Crunchbase; one rep reducing territory planning from 5 weeks to 15 minutes; Slack/Looker partnership saving 1000 hours/month) but these are scattered and rarely tied to clear causality or timelines. Most claims about data silos, FOMO-driven AI adoption, and the need for internal data integration remain high-level. The conversation lacks detailed case studies, before/after metrics, or named enterprise examples beyond brief mentions.

“My last startup was Looker that I got acquired by Google in 2020, uh, for 2.6 billion, which was a 26x multiple of our run rate”

“we have a company that's buying our data right now Simply for targeting AI companies, we have 31,000, over 31,000 um, AI companies in our database”

Greg asks reasonable opening questions and allows Ann space to develop ideas, but rarely pushes back, challenges claims, or digs deeper into contradictions. For example, when Ann states 'we haven't seen a lot of end to end' ROI yet, Greg doesn't probe what that means or why a CRO at a data company wouldn't have stronger conviction. The host doesn't challenge the repetitive framing of 'data silos' or ask for concrete diagnostic tools. Follow-ups are often surface-level (e.g., 'and so you know...') rather than sharp probes. The conversation reads as a friendly, uncontested narrative rather than substantive examination.

“Yeah, I definitely, that conversation is starting now, but yeah, definitely”

“Well and I think it also sounds like going back to your, you know, getting, keeping people focused on revenue generating activities”