AI democratisation is eroding traditional data advantages: yesterday’s proprietary datasets can now be matched—or even surpassed—through APIs and general-purpose models

In 2012, Target could predict pregnancy before customers announced it to their families—a textbook example of first-party data creating competitive advantages. Today, those same capabilities are available through APIs.

For a while, proprietary datasets created obvious advantages—Target’s pregnancy predictions became a case study.

But a decade later, even highly specialised datasets can be leapfrogged by general-purpose AI. Bloomberg trained a 50-billion parameter model on 363 billion tokens from four decades of curated financial data, plus 345 billion tokens from general datasets. Despite this massive proprietary advantage, GPT-4 without specialised finance training beat BloombergGPT on almost all finance tasks. Bloomberg's 2023 launch of BloombergGPT reveals how quickly proprietary advantages erode.

The performance gap between leading AI models has compressed dramatically, as industry analysis documents. When sophisticated capabilities become this accessible this quickly, where we think of strategic value starts to shift fundamentally.

Until mere months ago, a company could effectively rely on a data moat to protect it from disruption. Now, with lookalike data and LLMs—and specialised companies generating synthetic data that closes the gap, or, in Bloomberg’s case, outperforms a data moat—we need to rethink data strategy.

Until LLMs it was reasonable to focus on building bigger and better data moats. Now, the question of strategy fundamentally shifts. Consider this alternative to shape your data strategy: “What execution capabilities will be impossible for competitors to replicate, even with similar data?"

Three forces are driving the shift that’s happening:

We saw something similar in the age of Big Data.

Big data taught us that more isn’t always better. Companies hoarded information, but without the right culture or strategy, data lakes quickly turned into swamps. We’re seeing the same pattern with AI: stockpiles of proprietary data aren’t enough when a competitor with an API key to a modern LLM can generate comparable—or even superior—insights.

Recommendation engines lose their edge once they’re everywhere. Originally confined to e-commerce, lookalike modelling now powers TV ads, financial services, and industries that never had recommender systems.

The market is projected to hit $7.3 billion by 2029, but ubiquity breeds sameness. As Sam Altman warns, we’ve entered a “fast fashion era of SaaS,” where accessible tools create more copycats than differentiation. If anyone can plug into a recommendation API, the real question becomes: what makes advantage last when technical superiority is temporary?

As tools become ubiquitous, the differentiator shifts from algorithmic sophistication to how quickly a company acts on insights.

Competitive edge now comes from how quickly you can execute, not how much data you own. Des Traynor of Intercom captures it well: “You have to move with the uncertainty. AI is evolving too fast…you need to be willing to move on intuition.” Intercom is betting on speed over certainty, going all-in on AI rather than waiting for perfect clarity.

The same logic is reshaping retail: smart players are winning with execution Amazon can’t copy—like in-store pickup, loyalty experiences, and services that turn data into tangible, real-time value. Those capabilities, rooted in speed and customer touchpoints, are far harder to imitate than another algorithm.

While organisations cite "limited AI skills" as barriers, experienced practitioners point to execution discipline as the real bottleneck.

Three patterns to watch for:

Technical teams optimise for model accuracy while business teams need customer impact. The disconnect kills execution speed and mires teams in misalignments. Teams that succeed treat AI implementation as change management, not just technology deployment. The issue isn't data quality—it's speed of turning insights into bottom line impacting improvements.

This isn't an argument to abandon data strategy. We can't make the same data bets we made five years ago and expect equivalent advantages. Smart companies are still building data moats—just differently.

Four contexts where first-party data genuinely creates lasting value:

Highly regulated domains: Healthcare, financial services, and industrial settings where proprietary data reflects years of compliance investment competitors can't easily replicate. The barrier isn't the data itself—it's proven processes for handling regulated information safely.

True network effects: When your data improves user experience in ways that attract more users, creating better data and experiences, you've built a reinforcing cycle. This only works when user experience improvements and data enhancement happen in near real-time. Brian Balfour, former VP of Growth at Hubspot, has a really great deep-dive on this subject.

Timing-critical decisions: Trading algorithms, real-time personalisation—contexts where split-second decisions and fresh data provide measurable advantage. These remain defensible because execution speed and data freshness compound together.

Proprietary process intelligence: The biggest data advantage lies not in what you know, but how you leverage insights. Companies with high-quality data pipelines and early predictive signals outweigh generic customer advantages. Supply chain optimisation and operational efficiency data prove harder to replicate than customer preference information.

Bloomberg's experience illustrates this perfectly. BloombergGPT becomes a retention moat, not a revenue line. Now, it reinforces the value of $30,000+ annual Terminal subscriptions through integrated workflow, not direct monetisation. The real moat isn't the AI model—it's the platform dependency around it. Bloomberg also has workflow intelligence, which as I suggested last week, is set to be more valuable than just data sets. The ability to have complex interdependent behavioural data will be more valuable than data itself.

Atlassian's strategic acquisition of The Browser Company reveals how execution-focused companies are thinking about data moats in the Age of AI. Hard to copy, interconnected, consolidating across a layer to create sustainable advantages.

Their bet, which I walked through on Friday last week shows how workflow intelligence and data flows are the bet over traditional data collection.

Given there are still places first-party data will make a difference, take time to identify those areas where data will be hard to copy. When you've identified where first-party data genuinely creates advantage, focus becomes critical.

Amplify that value while building execution capabilities that compound over time. Teresa Torres emphasises focusing on specific moments where insights directly influence customer outcomes rather than broad data collection strategies. Not all moats are equal.

How are you re-engineering products for better learning at each experience stage? The question shifts from "What can we learn?" to "How does learning improve what customers experience immediately?"

Companies mastering execution excellence while others optimise data complexity will build the most defensible positions.

Competitive advantages aren't permanent. The companies that recognise this pattern first tend to win sustianably and decisively, always pivoting to the next advantage.

History is full of late movers who won through execution, not first-mover data advantages. Google wasn’t the first search engine, Facebook wasn’t the first social network, Amazon wasn’t the first bookstore, and Southwest wasn’t the first airline.

What they shared was sharper execution: faster algorithms, cleaner interfaces, better customer experience, and operational focus where competitors lagged. Even in cloud computing, it wasn’t early adopters who defined the market, but the companies that scaled automation and reliability into everyday workflows. The pattern is clear: execution consistently outlasts incumbency.

Organisations that grasp this shift today will build tomorrow's defensible positions. While competitors defend yesterday's advantages—execution speed is becoming the foundation all other advantages are built on.

Phase 1: Audit and Baseline
Map your current decision-making workflows. Prioritise actionable data over comprehensive analytics, and focus on well-defined tasks with reliable processes.

Phase 2: Embed and Experiment
Create cross-functional insight squads rather than centralised data teams. Embedding data scientists and data engineers into the same team/area means that there are squads who can query data, interpret patterns, and implement changes that customers experience as immediate improvements. This cuts cycle times and increases team autonomy to solve impact challenges internally and externally.

Phase 3: Scale What Works
Build measurement frameworks that work. I’ve got a post coming about quality and productivity. But for today, tracking speed of learning, not just accuracy of prediction or success rates. Consider DORA, SPACE, and DX Core 4 as frameworks for measuring success.

Reward teams that take risks, document failures and help others learn. Scale the things that work for your company culture.

This week's selection explores the fundamental shift from data hoarding to execution excellence—examining how smart companies build sustainable advantages when AI capabilities become universally accessible.

The Browser Wars Are Over. The Platform Wars Have Begun | Casey Newton

Newton's prescient analysis of how browsers are evolving from navigation tools to productivity platforms. His examination of Arc's design philosophy—"the browser as your computer's new operating system"—directly supports this week's thesis about execution advantages over data collection. Essential reading for understanding how platform thinking trumps feature accumulation.

Platform Engineering is Failing | Lukas Gentele

Platform engineering often fails when companies prioritise developer tools over a solid infrastructure foundation. Focusing first on standardised, automated, and governed infrastructure ensures platforms are stable, secure, and actually accelerate developer productivity. Lukas talks through what it takes to succeed at platform engineering.

Winners consistently prioritise workflow integration and execution speed over proprietary features or exclusive data.

The companies building tomorrow's defensible positions understand that in an AI-democratised world, how quickly you can turn insights into customer value matters more than what insights you have exclusive access to.

I've been thinking about execution versus flashy features this week, which led me down an unexpected rabbit hole that connects surprisingly well to this article's themes.

The speaking circuit calls again. I’m excited to be sharing thoughts and ideas at events again. If you're around Oxford or interested in enterprise search and data fundamentals, here's where you'll find me:

FromSoftware's approach to game development mirrors everything I've been writing about execution advantages. While competitors chase microtransaction revenue and live-service models, FromSoftware focuses relentlessly on craft—world design, gameplay mechanics, and player experience. The result? Elden Ring has sold over 25 million copies without a single microtransaction, proving that execution excellence beats monetisation gimmicks.

Critical Role is a group of voice actors who play Dungeons & Dragons together—and have managed turned it into a media empire worth millions.

While other actual-play shows focus on production value and celebrity guests, Critical Role succeeds through consistent storytelling execution and genuine character development. Their success is about showing up every week and executing collaborative storytelling at an exceptionally high level while also crafting and co-creating with the communities their work happens in. There's something deeply satisfying about watching a team build something meaningful through sustained, excellent execution.

I’m thrilled about Brennan Lee Mulligan taking over as the GM for campaign four, and I could talk about that for hours, but it’s time for you to get back to other things.

This week’s edition gives you a view of exactly how nerdy I am. 😅

Thanks for making it all the way to the end. I'd love to know what's sparking your curiosity this week—hit reply or share your own "Outside the Terminal" pick. Until next time, keep learning fast and building well.

– Saielle