It’s been three years since OpenAI released ChatGPT and kicked off a surge in innovation and attention on AI. Since then, optimists have regularly claimed that AI will become a critical part of the enterprise software industry, and so enterprise AI startups mushroomed on the back of immense amounts of investment.
But enterprises are still struggling to see the benefit of adopting these new AI tools. An MIT survey in August found that 95% of enterprises weren’t getting a meaningful return on their investments in AI.
So when will businesses start seeing real benefits from using and integrating AI? TechCrunch surveyed 24 enterprise-focused VCs, and they overwhelmingly think 2026 will be the year when enterprises start to meaningfully adopt AI, see value from it, and increase their budgets for the tech.
Enterprise VCs have been saying that for three years now. Will 2026 actually be different?
Let’s hear what they have to say:
Kirby Winfield, founding general partner, Ascend: Enterprises are realizing that LLMs are not a silver bullet for most problems. Just because Starbucks can use Claude to write their own CRM software doesn’t mean they should. We’ll focus on custom models, fine tuning, evals, observability, orchestration, and data sovereignty.
Molly Alter, partner, Northzone: A subset of enterprise AI companies will shift from product businesses to AI consulting. These companies may start with a specific product, such as AI customer support or AI coding agents. But once they have enough customer workflows running off their platform, they can replicate the forward-deployed engineer model with their own team to build additional use cases for customers. In other words, many specialized AI product companies will become generalist AI implementers.
Techcrunch event
San Francisco
|
October 13-15, 2026
Marcie Vu, partner, Greycroft: We’re very excited about the opportunity in voice AI. Voice is a far more natural, efficient, and expressive way for people to communicate with each other and with machines. We’ve spent decades typing on computers and staring at screens, but speech is how we engage in the real world. I am eager to see how builders reimagine products, experiences, and interfaces with voice as the primary mode of interaction with intelligence.
Alexa von Tobel, founder and managing partner, Inspired Capital: 2026 will be the year AI reshapes the physical world — especially in infrastructure, manufacturing, and climate monitoring. We are moving from a reactive world to a predictive one, where physical systems can sense problems before they become failures.
Lonne Jaffe, managing director, Insight Partners: We’re watching how frontier labs approach the application layer. A lot of people assumed labs would just train models and hand them off for others to build on, but that doesn’t seem to be how they are thinking about it. We may see frontier labs shipping more turnkey applications directly into production in domains like finance, law, healthcare, and education than people expect.
Tom Henriksson, general partner at OpenOcean: If I had to pick one word for quantum in 2026, it’s momentum. Trust in quantum advantage is building fast, with companies publishing roadmaps to demystify the tech. But don’t expect major software breakthroughs yet; we still need more hardware performance to cross that threshold.
Which areas are you looking to invest in?
Emily Zhao, principal, Salesforce Ventures: We are targeting two distinct frontiers: AI entering the physical world and the next evolution of model research.
Michael Stewart, managing partner, M12: Future datacenter technology. For the last year or so, we’ve been standing up a few new investments that signal our interest in future “token factory” technology, with an eye towards what can really advance how efficiently and cleanly they run. This is going to continue in 2026 and beyond, in categories that include everything within the walls of the data center: cooling, compute, memory, and networking within and between sites.
Jonathan Lehr, co-founder and general partner, Work-Bench: Vertical enterprise software where proprietary workflows and data create defensibility, particularly in regulated industries, supply chain, retail, and other complex operational environments.
Aaron Jacobson, partner, NEA: We are at the limit of humanity’s ability to generate enough energy to feed power-hungry GPUs. As an investor, I’m looking for software and hardware that can drive breakthroughs in performance per watt. This could be better GPU management, more efficient AI chips, next-gen networking approaches like optical, or rethinking thermal load within AI systems and datacenters.
When it comes to AI startups, how do you determine that a company has a moat?
Rob Biederman, managing partner, Asymmetric Capital Partners: A moat in AI is less about the model itself and more about economics and integration. We look for companies that are deeply embedded in enterprise workflows, have access to proprietary or continuously improving data, and demonstrate defensibility through switching costs, cost advantages, or outcomes that are difficult to replicate.
Jake Flomenberg, partner, Wing Venture Capital: I’m skeptical of moats built purely on model performance or prompting — those advantages erode in months. The question I ask: if OpenAI or Anthropic launches a model tomorrow and is 10x better, does this company still have a reason to exist?
Molly Alter, partner, Northzone: It’s much easier today to build a moat in a vertical category rather than a horizontal one. The best moats are data moats, where each incremental customer, data point, or interaction makes the product better. These are somewhat easier to build in specialized categories like manufacturing, construction, health or legal, where data is more consistent across customers. But there are also interesting “workflow moats,” where defensibility comes from understanding how a task or project moves from point A to point B in an industry.
Harsha Kapre, director, Snowflake Ventures: For AI startups, the strongest moat comes from how effectively they transform an enterprise’s existing data into better decisions, workflows, and customer experiences. Enterprises already sit on incredibly rich data; what they lack is the ability to reason over it in a targeted, trustworthy way. We look for startups that blend technical expertise with deep industry knowledge and can bring domain-specific solutions directly to a customer’s governed data, without creating new silos, to deliver insights or automation that weren’t previously possible.
Will 2026 be the year when enterprises start to gain value from AI investments?
Kirby Winfield, founding general partner, Ascend: Enterprises are realizing that random experiments with dozens of solutions create chaos. They will focus on fewer solutions with more thoughtful engagement.
Antonia Dean, partner, Black Operator Ventures: The complexity here is that many enterprises, despite how ready or not they are to successfully use AI solutions, will say that they are increasing their investments in AI to explain why they are cutting back spending in other areas or trimming workforces. In reality, AI will become the scapegoat for executives looking to cover for past mistakes.
Scott Beechuk, partner, Norwest Venture Partners: We’re definitely getting closer. If last year was about laying the infrastructure for AI, 2026 is when we begin to see whether the application layer can turn that investment into real value. As specialized models mature and oversight improves, AI systems are becoming more reliable in daily workflows.
Marell Evans, founder and managing partner, Exceptional Capital: Yes, but still incremental. There is still a lot of iteration, and AI is still improving to the point of being able to showcase pain-point solutions for enterprises across a variety of industries. I believe solving simulation to reality training will likely open up many opportunities for a selection of industries, both existing and nascent.
Jennifer Li, general partner, Andreessen Horowitz: There have been sensational headlines this year about enterprises not seeing returns on their AI investments. Ask any software engineer if they would ever want to go back to the dark ages before they had AI coding tools. Unlikely. My point is, enterprises are already gaining value this year, and it will multiply across organizations next year.
Do you think enterprises will increase their AI budgets in 2026?
Rajeev Dham, managing director, Sapphire: Yes, I believe they will, though it’s nuanced. Rather than simply increasing AI budgets, organizations will shift portions of their labor spend toward AI technologies or generate such strong top-line ROI from AI capabilities that the investment effectively pays for itself three to five times over.
Rob Biederman, managing partner, Asymmetric Capital Partners: Budgets will increase for a narrow set of AI products that clearly deliver results, and will decline sharply for everything else. Overall spend may grow, but it will be significantly more concentrated. We expect a bifurcation, where a small number of vendors capture a disproportionate share of enterprise AI budgets while many others see revenue flatten or contract.
Gordon Ritter, founder and general partner, Emergence Capital: Yes, but spend will concentrate. Enterprises will increase budgets where AI expands on institutional advantages, and pull back from tools that simply automate workflows without capturing (and securing!) proprietary intelligence.
Andrew Ferguson, vice president, Databricks Ventures: 2026 will be the year that CIOs push back on AI vendor sprawl. Today, enterprises are testing out multiple tools for a single use case — monthly spend and switching costs are low in many cases, so the incentive to experiment is there — and there’s an explosion of startups focused on certain buying centers like [go-to-market], where it’s extremely hard to discern differentiation even during [proof of concepts]. As enterprises see real proof points from AI, they’ll cut out some of the experimentation budget, rationalize overlapping tools, and deploy those savings into the AI technologies that have delivered.
Ryan Isono, managing director, Maverick Ventures: In aggregate, yes, and there will be some shifting from pilots/experimental budgets to budgeted line items. A boon for AI startups in 2026 will be the transition of enterprises who tried to build in-house solutions and have now realized the difficulty and complexity required in production at scale.
What does it take to raise a Series A as an enterprise-focused AI startup in 2026?
Jake Flomenberg, partner, Wing Venture Capital: The best companies right now combine two things: a compelling “why now” narrative — usually tied to GenAI creating new attack surfaces, infrastructure needs, or workflow opportunities — and concrete proof of enterprise adoption. $1 million to $2 million [annual recurring revenue] is the baseline, but what matters more than that is whether customers view you and your product as mission-critical to their business versus just being a nice-to-have. Revenue without narrative is a feature; narrative without traction is vaporware. You need both.
Lonne Jaffe, managing director, Insight Partners: You should aim to show you’re building in a space where the [total addressable market] expands rather than evaporates as AI drives down costs. Some markets have high elasticity of demand – a 90% price decline leads to a 10x increase in market size. Others have low elasticity, where dropping the price can vaporize the market, so the customers keep all of the value being created.
Jonathan Lehr, co-founder and general partner, Work-Bench: Customers are using the product in real, day-to-day operations, and are willing to take reference calls and talk honestly about impact, reliability, and buying process, etc. Companies should be able to clearly show how the product saves time, reduces cost or increases output in a way that holds up through security, legal, and procurement reviews.
Michael Stewart, managing partner, M12: We (investors) were casting a doubtful eye towards [estimated annual recurring revenue] or pilot revenue until recently. Now, it’s not seen as much of an asterisk as much as the customer’s interest and willingness to evaluate a solution in the face of so many options pushed their way. Getting those engagements and customer buy-in in terms of running an evaluation isn’t just a matter of forward-deployed engineers making it easier for the customer. It takes quality and a winning marketing message to do it in 2026. Investors are expecting to see conversions become the leading part of the story after 6 months of pilot use.
Marell Evans, founder and managing partner, Exceptional Capital: Execution and traction. The best signal is users genuinely delighted to use the product, and the technical sophistication of the business. We look at a huge North Star of real contractual agreements, 12+ months. In addition to that, was this founder able to attract top-tier talent to join their startup over competitors or the traditional hyper-scalers?
What role will AI agents play at enterprises by the end of 2026?
Nnamdi Okike, managing partner and co-founder, 645 Ventures: Agents will still be in their initial adoption phase by the end of 2026. There are many technical and compliance hurdles that need to be overcome for enterprises to truly benefit from AI agents. There also need to be standards created for agent-to-agent communication.
Rajeev Dham, managing director, Sapphire: One universal agent will emerge. Today, each agent is siloed in its role – for example, inbound [sales development representative], outbound SDR, customer support, product discovery, etc. But by late next year, we’ll start to see these roles converge into a single agent with shared context and memory, breaking down long-standing organizational silos, and enabling a more unified, contextual conversation between companies and their users.
Antonia Dean, partner, Black Operator Ventures: The winners will be organizations that figure out the right balance of autonomy and oversight quickly, and that recognize agent deployment as collaborative augmentation rather than a clean division of labor. Rather than agents handling all routine work while humans do all the thinking, we’ll see more sophisticated collaboration between humans and agents on complex tasks, with the boundary between their roles continuously evolving.
Aaron Jacobson, partner, NEA: The majority of knowledge workers will have at least one agentic co-worker they know by name!
Eric Bahn, co-founder, general partner, Hustle Fund: I think that AI agents will probably be the bigger part of the workforce than any humans in enterprises. Proliferating AI agents is essentially free and zero marginal cost. So why not grow through bots?
What kinds of companies in your portfolio are seeing the strongest growth?
Jake Flomenberg, partner, Wing Venture Capital: The companies growing fastest are the ones that identified a workflow or security gap created by GenAI adoption, then executed relentlessly on product-market fit. In cybersecurity, it’s tools addressing data security so LLMs can interact with sensitive data safely, and agent governance ensuring autonomous systems have appropriate controls. In marketing, it’s new areas like Answer Engine Optimization (AEO) — getting discovered in AI responses, not just search results. The common thread: these weren’t categories two years ago, but are now must-haves for enterprises deploying AI at scale.
Andrew Ferguson, vice president, Databricks Ventures: We’re seeing growth tied to a few common themes. One is companies that land with focused use cases — companies that start with a narrower wedge (could be a focused target persona or use case), really nail it, become sticky and earn the right to expand from the initial wedge.
Jennifer Li, general partner, Andreessen Horowitz: Companies that help enterprises put AI into production are doing well. Areas like data extraction and structuring, developer productivity for AI systems, infrastructure for generative media, voice and audio for media and apps like support or call centers.
What kinds of companies are seeing the strongest retention?
Jake Flomenberg, partner, Wing Venture Capital: Companies with retention and expansion share a pattern: they solve problems that intensify as customers deploy more AI. Strong retention comes from three things: being mission-critical (removal breaks production workflows), accumulating proprietary context that’s hard to recreate, and solving problems that grow with AI adoption rather than being one-and-done.
Tom Henriksson, general partner at OpenOcean: Retention is trickier to measure for younger companies, but the highest retention we’re seeing is in the serious enterprise software providers, especially those enhanced with AI. A good example is Operations1, which digitizes employee-led production processes end-to-end. These companies go deep into the customer’s organization, transform how they operate, and build up proprietary data and knowledge that makes them very hard to do without.
Michael Stewart, managing partner, M12: Startups serving the enterprise in data tooling and vertical AI apps, with forward-deployed teams assisting in customer satisfaction, quality, and product improvement. This seems to be the winning formula that has been adopted by all leading startups in those markets. Longer term, the embedded teams might recede as the customers start to internalize the use of AI in their organizations and workday practices.
Jonathan Lehr, co-founder and general partner, Work-Bench: Retention is highest where software becomes foundational infrastructure rather than a point solution. Authzed has strong retention because authorization and policy sit at the core of modern systems, and are extremely costly to rip out once embedded. Courier Health and GovWell act as systems of record and orchestration layers for end-to-end workflows, patient journeys in healthcare, and permitting in government, which makes them deeply embedded once live.
