
Looking back on 2024: The AI playbook takes shape
By Tara Stokes
AI Founders’ Chicken-and-Egg Problem
Like many, we believe startups benefit from a moat. But creating one as a tech startup – especially an AI startup today – feels increasingly daunting. Investors and builders keen to find the next enduring businesses are increasingly looking to unique datasets as one potential moat. While some first-generation foundational model startups build upon public datasets (hence the name “wrappers”), the next wave of companies are seeking to amass and capitalize on their own novel datasets.
We believe the vertical application ideas leveraging proprietary datasets are the most likely to become long-term defensible businesses – but, how do founders acquire this data without an existing product, customer base, or go-to-market strategy? In this perspective, I’ll share my observations on how teams are navigating this paradox and offer illustrative examples as startups, founders and incumbents are collectively trying to write the next-generation’s playbook.
Propriety Datasets Can Help Startups Standout in a Crowd
We believe much of the low-hanging “software” fruit has been plucked. Either the software solution is increasingly commoditized and thus challenging the economics, and/or the incumbents possess an innovation advantage via resources, talent, historical data, existing customers, etc. Additionally, many pure-play software markets are becoming crowded markets – a Martech Map below for illustration:
Source: chiefmartec.com
Be Opinionated, But Openminded
However, this can encourage early-stage founders to reinvent the typical startup playbook and for investors to be more progressive when evaluating potential business strategies. In our framework below, we charted hypothetical paths to a data moat – and potentially big businesses – by analyzing both their target business model and end markets over time. The arrows below represent five examples of hypothetical startups building novel datasets and pivoting the business upstream – seeking higher margins and/or Annual Contract Values (ACVs):
For example, the two lower arrows 1 and 2 represent hypothetical hardware startups that transformed into software-with-a-hardware-component businesses. Arrow 1 initially sold hardware to enterprise customers before building a consumer product, and arrow 2 consistently sold to government – both chasing higher margins with a business model evolution.
- Rethink Hardware Startups: While hardware development is capital-intensive and often lower margin, startups creating novel data through their hardware are attracting renewed investor interest.
In a quest to graduate beyond a great demo to a great product, we’re seeing some software-as-a-service businesses position themselves to be the next software leaders.
- Trojan-horse Services Businesses: While sometimes difficult to scale, services businesses can leverage their initial experience to inform product development and achieve a “good enough” offering at scale.
Additionally, we’re optimistic about founders reconsidering non-traditional venture markets, such as midmarket or government customers.
- Energized SMB Startups: Traditionally associated with lower contract values, we believe SMB-focused startups are attracting venture capital by building comprehensive end-to-end solutions instead of point solutions.
- Superpowering Local and Regional Governments: Aided by Defense Tech’s tailwinds, we read a trend of states, cities, municipalities, etc. considering tech-forward solutions to improve efficiency and/or increase revenue.
Finally, these new datasets have the potential to be steadily improved to help that company stay ahead of competitors. In our view, the unique insights from this data can then open up more opportunities and use cases, making the data a valuable asset.
Enhancing “Better, Faster, Cheaper” for the AI Era
When building a data moat, we believe it’s crucial to determine whether you’re aiming for depth or breadth:
- Deep (Vertical Focus): Are you striving to become the default solution for a specific niche?
- Wide (Horizontal Focus): Are you aiming to solve a similar problem across various sectors and industries?
Next, you might consider some of the following questions:
- Quality: Are you improving the collection or utility of data?
- Collection: Are you building proprietary technology to enhance data quality? Quantune’s advancements in optics and photonics technology exemplify this approach, aiming to create an accessible instrument without compromising data quality.1
- Utility: Are you breaking down data silos? For example, Glyphic’s AI sales copilot aims to leverage customer conversations to empower data-driven decisions across departments. 1
- Time: Are you creating new markets, growing existing ones, or accelerating adoption?
- Ephemeral: Is the value of your data short-lived, similar to live sports? Blackshark’s efficient 3D mapping for change detection illustrates this concept. 1
- Perpetual: Are you generating valuable data byproducts? Netradyne’s fleet safety technology also enables fleet management and compliance solutions, demonstrates this principal. 1
- Pace: Are you accelerating your customers’ ability to act based on data insights? PolyAI’s customer-led voice assistants, deployable in weeks, showcase this rapid value delivery. 1
By identifying the desired data asset, we believe founding teams can more effectively shape their early product and choose initial partners or customers to help build that specific dataset.
Selling Budget Protection or Convenience Bliss?
One challenge many founders may face is acquiring those initial design partners or customers who will kickstart the data flywheel needed to attract future customers. To overcome this paradox, we believe you need a “wedge product” that delivers immediate value, establishes early relationships, and encourages customer retention.
Two common approaches to building a successful wedge product include:
- Sell on Budget: Can you offer a more affordable solution that provides immediate value?
- Sell on Convenience: Can you streamline a process to save customers time and hassle?
This initial offering may provide access to valuable data or generate proprietary data as a byproduct. From there, we advise founders to focus on creating a virtuous loop:
- More customers lead to…
- More insights, which enables…
- More value delivery, resulting in…
- More customers, continuing the cycle and reinforcing your data moat.
As your data asset grows, we believe you should begin introducing features that leverage the asset to push insights or actions. This old-school transition is exemplified by companies like Instagram, which evolved from a photo editing app to a data-driven social media platform.
NRR is the new ARR
Retaining your customers is paramount, as we believe many investors are increasingly prioritizing Net Revenue Retention (NRR) over net new Annual Recurring Revenue (ARR). This “stickiness” may reveal enterprises transitioning from their innovation budgets to core spend, and ultimately validating your product as “good enough” not just “good.” With a sharp focus on retention, startups can expand within existing accounts and leverage a repeatable onboarding process. At the same time, these startups can increase their margin profile by introducing higher-margin, software-like features powered by their growing data assets.
Navigating Common Pitfalls
Building a successful AI data moat is not without its challenges, and the below are some common pitfalls that we’ve observed:
- Over customization: Excessive customization for early customers can inadvertently lead to building a services business that lacks scalability.
- Targeting the Wrong Customers: Catering to the wrong customer profile can trap startups in local maxima, sacrificing the pursuit of a larger market opportunity.
- Rigid Ideal Customer Profile: Sticking too rigidly to an ideal customer profile can blindside startups to unforeseen opportunities.
Talk Data to Us!
Ultimately, we believe that regardless of your starting point, strategizing how you can turn your initial dataset into a novel asset will be time well spent. While you can’t predict every step amid this quickly changing landscape, you can think early and often about your data moat. By prioritizing data acquisition and taking a disciplined approach to product development and customer selection, we believe you can build a wedge product that sets your flywheel in motion. If you’re a founder building in the space, a researcher, or an operator, we’d love to hear from you.
- Selected portfolio companies. For a full list of our portfolio, please visit Portfolio – Point72 Ventures