- The Angle
- Posts
- The AI-native enterprise playbook
The AI-native enterprise playbook
Gil Dibner
The playbook for enterprise software in an AI-native world is still being written. Some days it feels like we’re trying to play soccer as the playing field itself is changing shape. While it's too early to draw conclusions as to what works short-term and what will be long-term sustainable, we are starting to identify patterns that could form the beginning of an AI-native enterprise playbook. Here are my thoughts as they stand today:
Domain expertise matters more than ever. This has always been true, but it is now crucial. True industry depth can both accelerate sales cycles and create barriers to entry that give a startup a competitive edge, often by allowing more time to implement complex AI features without immediate competitive pressure. A big part of our investment thesis in companies like CruxOCM (oil & gas), Fixefy (logistics), Portchain (ocean freight),and Wisery Labs (intelligence) is the deep domain expertise of the teams. If twenty CS graduations with $20M can’t build (or sell!) the product, we are interested!
Big painful problems in weird industries. It’s a big world out there with a lot of big, thorny problems – and most of them are problems few people have ever heard about. The weirder the industry, the less likely there is significant competition and the more valuable an entrepreneur’s domain expertise can be. We’ve increasingly been actively seeking out the weirder industries and problems.
Tight integration with human workflows. At the end of the day, humans are still the decision makers on software purchasing decisions, and that is likely to remain the case for some time. AI's ability to eliminate much of the pain of data integration is melting away old moats that protected incumbents. This impacts challengers as well. In an AI-first world, the strongest moats are in the minds of users. Deep integration into critical human workflows is therefore a powerful value driver for AI-native software companies. Software that is used and loved by people is stickier and more valuable than any technical advantage alone. Across our portfolio, we are excited by opportunities toi drive deep engagement with human workflows. Two great examples are Dataflint (data engineers) and Jux (frontend designers).
Objectivity. GenAI is great at creative non-deterministic tasks. It’s less good at solving edge cases with a zero percent error rate. When a customer can objectively assess whether or not an AI-native solution has delivered the correct (or incorrect) answer, it creates conditions under which the best product can win. For founders focused on building the best product, this is a necessary success condition. We’re consistently drawn to situations where there is a right answer in a high-stakes setting. Reco, which operates in the high-stakes world of SaaS security, is using this to its advantage.
Leverage the 80/0 rule. In the AI-native world, powerful demos are king until they are not – and a new 80/0 rule is emerging. It takes very little effort to get a powerful demo up and running that can ride on an LLM backend and illustrate how some complicated process can be instantly revolutionized by the magic of AI, or at least 80% of the way there. The problem is that 80% of a solution is worth – ultimately – $0 in the long run. For real high-stakes enterprise use cases, the demo is worthless. Instead, it’s the edge cases, the evals, the advanced RAG, the data pipelines, etc., that make all the difference. AI might be magic, but the real value is hard work. This means that it’s essential to focus on problems where the quality of a solution can be objectively evaluated. This is why we invested in companies like FalkorDB (advanced GraphRAG), RootSignals (roll-your-own evals), Tensorleap (DNN optimization and explainability) and other tooling designed to make AI infrastructure more valuable for real-world enterprise applications.
Winning the local AI race. In some cases, startup challengers have the opportunity to deploy AI faster than incumbents, and this forms the essence of the investment thesis. Jumping into a crowded space is daunting, but conviction that you can deploy AI faster than your rivals could provide a workable thesis. Harvey and Legola, two of the dominant AI-native legaltech companies, are good examples of this approach.
AI as penetration strategy. Sometimes, AI creates opportunities to quickly unlock value for customers by enabling new product approaches that were previously impossible. We’ve seen a number of companies using AI to extract data from text-heavy sources (contracts, invoices) which enables them to deliver some very simple penetration products that were previously impossible.
AI as accelerant. One powerful way to stand out is by marrying AI-powered software with another highly defensible innovation. This could be a unique distribution channel, a novel hardware component, or a proprietary data acquisition method that complements the AI. Groundcover, for example, is enhancing its world-beating observability technology with powerful AI-first root cause analysis that none of its competitors are well positioned to offer.
Novel data types. Some of the most interesting AI-native companies are able to leverage product innovation to unlock and access new data types that are not available to other software vendors or foundational model companies. Our portfolio company Sourcix, for example, is doing just that with their AI-powered mechanical parts procurement platform. Their unique AI-powered solution has granted them access to design document data that no one else has. Another portfolio company of ours, Motorica, is building on a proprietary (and hard-to-come-by) motion dataset.
Full-stack AI-native GTM. Any enterprise startup that wants to succeed in the AI-native world must adopt AI-first approaches throughout the GTM stack. Documentation, customer support, integration, onboarding, outreach, and many other capabilities and functions are all being reinvented by small, nimble teams doing much more with fewer people and more powerful AI tools. In a board meeting last week, I watched as a small (and recently default alive!) team of founders demoed part of the highly sophisticated sales automation workflows they had built using some standard agentic tooling. It was extremely impressive, and it’s rapidly becoming table stakes.
If you are building an AI-native enterprise software business, I’d love to hear from you. What are the elements of your strategy to penetrate and dominate a market? How are you thinking about value capture, stickiness, and defensibility over the long run? What am I missing from the framework above?
FROM THE BLOG
No More Painting by Numbers
It’s the end of the “SaaS playbook.
The Age of Artisanal Software May Finally be Over
Every wave of technological innovation has been catalyzed by the cost of something expensive trending to zero. Now that’s happening to software.
Founders as Experiment Designers
David on why founders should run everything as an experiment.
When Growth Stalls
Or why to kickstart growth you should narrow your ICP.
WORTH READING
ENTERPRISE/TECH NEWS
Meta’s superintelligence super team. In a wide-ranging interview with The Information, Mark Zuckerberg laid out Meta's AI playbook, which is guided by a new vision for "personal super intelligence" and a strategy to deploy massive amounts of capital. He revealed that Meta is building multiple multi-gigawatt data centers, funded by its own cash flow, to create what he believes will be the largest compute fleet of any company. For the rest of the industry, this signals a dramatic escalation in the AI arms race, where the ability to self-fund infrastructure at the cost of hundreds of billions of dollars is being positioned as a primary competitive moat.
Introducing Kiro. Andy Jassy announced Kiro, an all-new agentic IDE from Amazon. Goal is to bring structure and process to the often chaotic "vibe coding" development style. The platform introduces "specs" to help developers plan and document their work and "hooks" to automate repetitive tasks, creating a more robust development lifecycle from prototype to production. For founders and investors, Kiro represents a bet on a new category of developer tools that move beyond simple code completion to offer an opinionated, end-to-end system for building maintainable software.
HOW TO STARTUP
Sweden’s latest moonshot. In a stunning display of velocity, Swedish AI startup Lovable has achieved unicorn status just eight months after its launch, raising a $200M Series A led by Accel at a $1.8B valuation. The company, whose platform allows users to create websites and applications with natural language, reached an astounding $75 million in annual recurring revenue within its first seven months with a lean team of only 45 employees. Lovable's trajectory demonstrates the new fundraising benchmark for AI-native companies, showcasing a massive market opportunity in tools that empower non-technical users to build and launch their own software.
Claude’s growth engine. In a recent podcast, Reforge's Brian Balfour and Fareed Mosavat break down Anthropic's unique growth strategy for Claude, centered around a new feature called "artifacts." These "artifacts" are essentially mini-apps that leverage a user's own API quota, creating a powerful, self-sustaining growth loop for the platform. For founders and investors, this is a compelling example of product-led growth in the competitive AI landscape, demonstrating how to build distribution and defensibility beyond simply having a powerful model.
OpenAI reflections. In a detailed post-mortem, former OpenAI employee Calvin French-Owen shares his reflections on the company's internal culture during its hypergrowth phase from 1,000 to over 3,000 employees. He describes an "incredibly bottoms-up" and meritocratic environment that runs entirely on Slack, where good ideas from anywhere quickly gain traction and small teams can launch major products without asking for permission. For founders, the post is a fascinating look at how one of the world's most important companies maintains a startup-like "bias to action" and agility, offering lessons on navigating the controlled chaos of scaling.
Cognition <3 Windsurf. In a major consolidation within the AI coding market, Cognition, the company behind the AI software engineer Devin, announced it is acquiring the popular AI-powered IDE, Windsurf. In a post on Twitter/X, CEO Scott Wu revealed the deal brings over Windsurf's product, its enterprise business with $82M in ARR, and its world-class go-to-market team. The move pairs a futuristic, fully autonomous agent with a widely adopted IDE and a scaled sales motion, signaling a strategy to accelerate enterprise adoption and own the end-to-end developer workflow.
HOW TO VENTURE
The dominance of small VC? New data from Peter Walker at Carta analyzing VC funds from 2017-2018 shows that after eight years, smaller funds are outperforming larger ones on a TVPI basis. Funds in the $1M-$10M bracket lead the pack, with the 95th percentile hitting a 7.01x multiple, while funds over $100M reached just 2.63x at the same tier. While this provides fresh evidence for LPs to bet on emerging managers, the analysis also notes that top-quartile (75th percentile) performance remains below a 3x multiple across all fund sizes, a figure described as "not great" for the industry overall.
PORTFOLIO NEWS
SteadyBit claims industry first MCP server for chaos engineering.
Groundcover listed as honorable mention in the Gartner Magic Quadrant for observability platforms for the first time.
PORTFOLIO JOBS
Firebolt
SRE, Engines (Munich)
Reply