New Opportunities in AI-Driven Startups

AI is finally making services businesses scalable—and—exciting to VCs. The global services market is in the trillions of💰s, far larger than today’s software market. Yet, services businesses haven’t been the darlings of venture capital, as they were perceived to lack rapid scaling potential. 𝗔𝗜 𝗶𝘀 𝗰𝗵𝗮𝗻𝗴𝗶𝗻𝗴 𝘁𝗵𝗮𝘁. By blending AI seamlessly with human expertise, there is an opportunity to get into much larger markets with models that have the potential to scale in ways services - or even SaaS businesses - can't. For example, instead of offering a marketing SaaS, an AI-powered Service-as-Software business can deliver what the customer really wants: high-quality leads or compelling content. We’ve seen this potential firsthand through Emergent Ventures’ investments in multiple AI-powered companies that leverage humans-in-the-loop. These models resonate with B2B customers because they offer faster, clearer paths to value—reliable outcomes delivered with greater efficiency. For many customers, it’s a significant upgrade over traditional agency or service-provider relationships. While the potential is huge, only a fraction of AI-powered services startups will scale. 𝗦𝘂𝗰𝗰𝗲𝘀𝘀 𝗱𝗲𝗽𝗲𝗻𝗱𝘀 𝗼𝗻 𝘁𝗵𝗼𝘂𝗴𝗵𝘁𝗳𝘂𝗹 𝗲𝗮𝗿𝗹𝘆 𝗰𝗵𝗼𝗶𝗰𝗲𝘀 𝗮𝗻𝗱 𝗲𝘅𝗰𝗲𝗽𝘁𝗶𝗼𝗻𝗮𝗹 𝗲𝘅𝗲𝗰𝘂𝘁𝗶𝗼𝗻. Here’s what we have learned works well: 𝟭. 𝗔𝗜-𝗛𝘂𝗺𝗮𝗻 𝗦𝘆𝗻𝗲𝗿𝗴𝘆: AI and software should do the heavy lifting, with humans involved strategically— e.g. for validating AI output, edge cases, enabling adoption, or acting on AI insights. Over time, reduce human input as the AI learns, and models improve. Target 60%+ initial gross margins, with a path to SaaS-like 75%+ margins over time. 𝟮. 𝗦𝗰𝗮𝗹𝗮𝗯𝗹𝗲 𝗛𝘂𝗺𝗮𝗻 𝗜𝗻𝘃𝗼𝗹𝘃𝗲𝗺𝗲𝗻𝘁: The dependency on hiring & training humans should not constrain scale and economics. Have a path to tapping into freelancers or agency partners. Leverage human experts in a high-talent location such as India. 𝟯. 𝗥𝗲𝗰𝘂𝗿𝗿𝗶𝗻𝗴 𝗥𝗲𝘃𝗲𝗻𝘂𝗲: Focus on high-value, recurring use-cases to ensure subscription-based revenue with strong net revenue retention (NRR). 𝟰. 𝗣𝗿𝗶𝗰𝗶𝗻𝗴 𝗣𝗼𝘄𝗲𝗿: Iterate to a solution that can command higher pricing, and a model that aligns incentives with customers, e.g. based on outcomes. 𝟱. 𝗗𝗮𝘁𝗮 𝗠𝗼𝗮𝘁𝘀: Build solutions that improve with use, creating compounding competitive advantages over time. 𝟲. 𝗠𝗼𝗱𝘂𝗹𝗮𝗿 𝗧𝗲𝗰𝗵: Architect a stack that can evolve with AI advancements. 𝟳. 𝗙𝘂𝗹𝗹-𝗦𝘁𝗮𝗰𝗸 𝗧𝗲𝗮𝗺: A founding team that has the technical expertise to build and rapidly improve complex AI-powered solutions, and deep operational acumen. A rare combination. These are complex businesses to build, and the right playbooks are yet to be perfected. But where this works, 𝗦𝗲𝗿𝘃𝗶𝗰𝗲𝘀-𝗮𝘀-𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗔𝗜 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀𝗲𝘀 𝘄𝗶𝗹𝗹 𝗿𝗲𝗱𝗲𝗳𝗶𝗻𝗲 𝗺𝗮𝗻𝘆 𝗕𝟮𝗕 𝗰𝗮𝘁𝗲𝗴𝗼𝗿𝗶𝗲𝘀 📈 #EnterpriseAI #startups #vc #SaaS

🚨 Bessemer Venture Partners’ Janelle Teng has carved out a niche backing early-stage startups at the intersection of AI and data, and was just promoted to partner. She shares her insights on AI infrastructure’s second wave, how she evaluates startups and what DeepSeek AI’s rise means for the sector.🚨 ✒️ What part of AI is the most poised for growth and why? Every major technology shift requires infrastructure or the foundational ‘picks and shovels’ before applications can scale on top. But innovation doesn’t stop once the apps are built, it evolves in response to app-layer needs. We’ve already seen the first wave of AI infrastructure companies. As more applications emerge, we’ll see a second wave, focused on data ops and handling, orchestration, observability and new training techniques. AI builders on the application layer have a front-row seat to real-world challenges, and we’re just starting to see new companies pop up to address them. ✒️ What are some your investments that speak to the above? Historically, searching through massive amounts of video or images meant relying on human-tagged metadata. With Coactive AI, you can label a small dataset and generalize that labeling across your entire corpus — allowing for much more precise search and discovery, particularly relevant given unstructured data is growing exponentially. In a completely different sector, TurbineOne is focused on ML Ops for the military. ✒️ How do you evaluate early-stage AI startups that don’t yet have a product? First, conviction in the founding team’s ability to approach a problem in a novel way — often rooted in their research and deep expertise. Second, I rely on feedback from developers and AI engineers experimenting with these technologies. Even if they aren’t paying customers yet, their enthusiasm and early adoption signals are powerful indicators of potential success. ✒️ What’s one thing everyone is missing about the current DeepSeek frenzy? From an AI infrastructure standpoint, I look at it in terms of hyperscaler CapEx. DeepSeek has shown that you can build a cheaper, more efficient model. So the question now becomes: What does that mean for efficiency across the landscape? Everyone needs to start thinking about efficiency, because maybe it's not access to the largest and most expensive data centers that gives you the edge. ✒️ Could it lead to investors shifting away from LLMs? Because LLMs are based on the transformer architecture, they fall short of reasoning and have reliability issues and hallucinations, making them unsuitable for industries like financial services and healthcare. That's why we’re already seeing new model architectures like sub-quadratic architectures and categorical deep learning emerge that don’t have the same limitations, and allow for true reasoning rather than just predictive pattern-matching. But there’s still space for both for unique use cases. #VCWednesdays #vc #venturecapital #startups #TechonLinkedIn

Why Investors Should Bet on Pittsburgh’s AI Boom Before It’s Too Late 🔹 In my last post, I talked about Skild AI, a Pittsburgh-born startup that raised $300M and is now valued at $4B. It’s proof that billion-dollar AI companies can be built here. But here’s the real question: Why aren’t more investors paying attention to Pittsburgh? 📢 According to Carnegie Mellon University’s recent report on Pittsburgh’s AI Renaissance, the city is at the forefront of human-centric AI innovation. Pittsburgh’s AI & Robotics Edge: ✅ World-Class Research → CMU is ranked among the top AI & robotics research institutions in the world. ✅ Cost Efficiency → Startups can stretch capital 3X further than in Silicon Valley. ✅ AI for Real-World Impact → Unlike the hype-driven tech in SF, AI in Pittsburgh is focused on healthcare, manufacturing, energy, and defense—high-value, deeply impactful sectors. ✅ Access to Non-Dilutive Funding → AI startups here can tap into federal grants (DARPA, NSF, DoD), corporate partnerships, and local innovation funds. YCombinator’s founder Paul Graham recently said: 💬 “I would raise maybe $500K, keep the company small for the first year, work closely with users to make something amazing, and otherwise stay off SV’s radar.” Guess what? APittsburgh is the perfect place for that approach. It offers everything needed to build capital-efficient AI startups without the inflated burn rates of the Bay Area. 🔹 The Investor Blind Spot ❌ VCs continue pouring billions into overhyped AI startups in SF and NY—many of which have high burn rates, talent wars, and unsustainable models. ✅ Meanwhile, Pittsburgh’s AI ecosystem is producing deep-tech companies with real commercial applications. The Opportunity? The next wave of AI unicorns will come from cities like Pittsburgh, where world-class talent meets real-world applications. Investors who recognize this NOW will have a massive edge. If you had $10M to invest in an AI startup outside of SF/NY, would you bet on Pittsburgh? #AI #Startups #Pittsburgh #VentureCapital #DeepTech #Robotics #Entrepreneurship #PaulGraham #CMU

Innovation strives under constraints. Whether its cost, access to data, or access to compute, constraints have always fueled some of the most disruptive innovations. The AI stack is no exception—it’s ripe for disruption, starting at the chip level with NVIDIA challengers (like Groq, Etched and others) to alternative foundational models like Liquid AI.ai or... DeepSeek AI. This Chinese-built open-source AI model is 95% cheaper than U.S.-based competitors, presumably built for under $6 million, and performs almost on par with models from giants like OpenAI, Google, and Meta. It’s also more energy-efficient—lowering not just costs but also the environmental impact of #AI. This is a powerful signal: AI innovation is no longer the exclusive domain of big players. Smaller, nimble companies now have the tools and blueprints to build competitive models. Startups will disrupt industries with purpose-built models and novel vertical and applied AI. The pace of innovation will continue to accelerate. And its exactly why I am SO excited to be backing early-stage human centric AI startups. The lesson? You’ve got to keep innovating. In tech, standing still is not an option. And as we’ve seen time and again, constraints aren’t obstacles—they’re opportunities. The entire AI stack is up for grabs. The question is: who will seize it?

As in-house AI efforts struggle, AI startups are speeding to commercial maturity. MIT’s latest study found that 95% of internal AI pilots fail to deliver meaningful results. But, while enterprises stumble with proprietary programs, specialized AI startups are racing toward commercial maturity and profitability at unprecedented speeds. CB Insights’ latest data reveals that 79% of AI unicorns created in 2025 are already in the "scaling" or “established” stages – up from 36% just last year and outpacing the 61% rate for non-AI unicorns. What does this commercial maturity shift highlight about the (immediate) future of enterprise AI initiatives? → Integration over innovation. While enterprises struggle with "flawed integration," AI startups are built from the ground up around their AI capabilities. There's no legacy system to integrate – the AI is the system. → Targeted problems win: Successful AI deployments use AI to solve targeted problems – exactly what we see in the 2025 new AI unicorn data. These aren't general-purpose AI experiments but laser-focused solutions that leverage AI to solve specific problems. → Buy beats build: External vendors succeed twice as often as internal pilots. Why? They can amortize development costs across multiple clients and attract specialized talent. → Build the AI-native business infrastructure: While enterprises mistakenly pour AI budgets into sales and marketing, smart startups are automating the unsexy stuff – workflow automation (Tines), medical documentation (Abridge), and sales operations (Clay). Are we reaching a tipping point where enterprises will stop trying to be an AI company when they can partner with one? The 5% of internal pilots that succeed are likely solving narrow, well-defined problems – exactly what these maturing AI startups do at scale. Which AI programs are you choosing to buy vs build? Why?

Influencers hyping an AI bubble burst are cooked. It’s earnings season, and AI’s impact on revenue growth is being quantified for all to see. Atlassian’s AI platform grew 25X in just a year, helping the company increase subscription revenue by 30%. IBM reported $5 billion in generative AI bookings so far. AI demand helped increase its software segment growth by 10%. Palantir’s AI platform drove a 64% increase in US commercial revenue, and its stock is at an all-time high today. Spotify Wrapped (a SQL query and a dashboard) was one of its biggest user engagement drivers of 2024, helping it post its first annual profit and proving that business leaders can’t overlook innovations built on simple data products. ✅ Here are the biggest takeaways ✅ AI is much more powerful as a revenue driver than a cost saver. Business leaders must shift their focus to customer-facing AI products. Don’t go straight to AI because data and #analytics are significant revenue drivers. Build for the future, deliver incrementally, and get paid today. An aligned data and AI product roadmap is more critical than ever. Mid-tier tech companies have massive opportunities. AI isn’t just for the Magnificent 7 and Big Tech. #Data and AI teams should present opportunities that align with and amplify the current business model. Startups can leverage low-cost AI features and even data products to accelerate their path to profitability. AI isn’t just for large corporations. Business leaders at SMEs don’t need to wait on the sidelines. Satya Nadella said that #AI is a new input for growth, and the evidence supporting his thesis keeps growing.

Stanford University researchers released a new AI report, partnering with the likes of Accenture, McKinsey & Company, OpenAI, and others, highlighting technical breakthroughs, trends, and market opportunities with large language models (LLMs).  Since the report is 500+ pages!!! (link in comments), sharing a handful of the insights below: 1. Rise of Multimodal AI: We're moving beyond text-only models. AI systems are becoming increasingly adept at handling diverse data types, including images, audio, and video, alongside text. This opens up possibilities for apps in areas like robotics, healthcare, and creative industries. Imagine AI systems that can understand and generate realistic 3D environments or diagnose diseases from medical scans. 2. AI for Scientific Discovery: AI is transforming scientific research. Models like GNoME are accelerating materials discovery, while others are tackling complex challenges in drug development. Expect AI to play a growing role in scientific breakthroughs, leading to new materials and more effective medicines. 3. AI and Robotics Synergy: The combination of AI and robotics is giving rise to a new generation of intelligent robots. Models like PaLM-E are enabling robots to understand and respond to complex commands, learn from their environment, and perform tasks with greater dexterity. Expect to see AI-powered robots playing a larger role in manufacturing, logistics, healthcare, and our homes. 4. AI for Personalized Experiences: AI is enabling hyper-personalization in areas like education, healthcare, and entertainment. Imagine educational platforms that adapt to your learning style, healthcare systems that provide personalized treatment plans, and entertainment experiences that cater to your unique preferences. 5. Democratization of AI: Open-source models (e.g., Llama 3 just released) and platforms like Hugging Face are empowering a wider range of developers and researchers to build and experiment with AI. This democratization of AI will foster greater innovation and lead to a more diverse range of applications.

When it comes to VC, AI is making the investment world go 'round. The Q2 2024 Venture Capital from CB Insights report offers some key insights into how startup investing is emerging - and evolving - from the turbulence of the past two years. Overall, the numbers remain below the peak of 2021 and 2022. But it's clear that investing momentum is returning - in large measure thanks to the race to back #ai startups. Key takeaways: 📈 Global Venture Funding: Rose to $65.7bn, an 8% QoQ increase. 💸 Europe + US: Startups in Europe raised $14bn in Q2 compared to $39bn in the US. The rebound is not as strong in Asia where startups only raised $9.7bn. 🤖 AI Dominance in VC: AI startups raised 28% of all VC dollars, reaching $18.3bn, the highest quarterly share on record. 💰 Large Deals Drove AI Funding: Those included $6.0bn for xAI, $1.1bn for CoreWeave, and $502m for Mistral AI. 🇫🇷 France + AI: France's AI potential has been getting a lot of attention over the past year. That may be more than hype. Many of the most notable AI deals in Q2 were in France. Besides Mistral, that includes: $100m for H (Seed); $32m for Adcytherix (Seed); and $30m for FlexAI (Seed). 🥖 France is Up: Overall, French startups have raised more in the first half of 2024 than in 2023 (again, still below 2022 levels). Beyond AI, France also had one of the largest late-stage deals with $145m for Pigment! Bottom line: AI continues to drive VC investments. The challenge now amid this new frenzy is whether investors will have the discipline to deeply analyze these deals - and ensure the economics match the technological promise. Link to the report: https://lnkd.in/eE7QCUgK #tech #ai #genai #frenchtech

How AI/ML startups today are different from the startups of yesterday ⤵️ PitchBook just released new data that quantifies the AI boom: 2021–2023: $330B into ~26,000 AI/ML startups 2018–2020: ~$200B into 20,350 AI/ML startups That’s a 66% increase in funding and a 27% increase in the number of AI/ML startups over the past three years. The rise in startups is in part due to the emergence of capital-light businesses that no longer require a large engineering team and resources to scale. AI products like #ChatGPT, #Gemini and #Midjourney create labor efficiency by helping teams complete tasks faster, and have shown to be a structural cost reduction in knowledge work. It’s the latest in a line of paradigm shifts: The internet cut the cost of distribution. ↓ The cloud cut the cost of storage and computing. ↓ Now, AI is cutting the cost of the entire venture creation process, from ideation to production, increasing capital efficiency at the same time. The proliferation of open-source (#MetaLlama) and AI-powered software development (#GitHubCopilot) tools help collapse the distance between language and code to zero. Technology barriers to entry no longer exist as they used to and as a result, #GenerativeAI is ushering in a new era of entrepreneurship – one which is rapidly moving towards a seamless, natural language based “idea-to-product workflow.” It's sparking a new generation of entrepreneurs who are able to harness industry expertise and relationships to solve meaningful problems in their fields. Our investments in Paxton AI, a legal tech start-up using LLMs to tackle regulatory compliance and legal drafting, and Fintary, an AI InsurTech platform automating account reconciliation, are just two examples of Kyber Knight Capital backing founders with deep industry expertise leveraging new technology to capture legacy market opportunities. It’s clear that AI is proving to be a force of economic empowerment across the economy, and minting many of these new breeds of startups. #Funding #VC #Startups #AI #MachineLearning

One of the realities in AI is that most of the benchmarks or best practices from the past don’t apply to this world 😅 The playbook is being written right now. 💪 🔑 This means… founders should take any / all advice with a huge pinch of salt 🛑 There is a truly unique opportunity for AI founders.. in rethinking how they build companies — sequence strategic bets, fundraise, think about team building etc. ✅ This market opportunity is truly unique with high stakes and will require leaning into first principles. 🙌 For example: Given the insane pace of technical innovation, your distribution engine can be the difference maker between an ok outcome vs a truly large one (vs a traditional plg saas biz) 🚀 1️⃣ Invest in your technical capabilities and a best in class GTM muscle earlier than you would typically 💪 2️⃣ Build a team that can run fast to capture existing demand while deploying new bets 🏃♀️ 3️⃣ Ignore the noise from your “advisors”. Focus on your customers and market to make key decisions 🔥 Stoked to see how this plays out!! #artificialintelligence #artificialintelligenceforbusiness #ai #aichallenges