VCs predict strong enterprise AI adoption next year — again

It’s been three years since OpenAI launched ChatGPT and kicked off a surge in innovation and a spotlight on AI. Since then, optimists have frequently claimed that AI will change into a crucial a part of the enterprise software program business, and so enterprise AI startups mushroomed on the again of immense quantities of funding.  

However enterprises are nonetheless struggling to see the good thing about adopting these new AI instruments. An MIT survey in August discovered that 95% of enterprises weren’t getting a significant return on their investments in AI.

So when will companies begin seeing actual advantages from utilizing and integrating AI? Trendster surveyed 24 enterprise-focused VCs, they usually overwhelmingly suppose 2026 would be the yr when enterprises begin to meaningfully undertake AI, see worth from it, and improve their budgets for the tech.  

Enterprise VCs have been saying that for 3 years now. Will 2026 really be totally different? 

Let’s hear what they must say:

Kirby Winfield, founding common associate, Ascend: Enterprises are realizing that LLMs aren’t a silver bullet for many issues. Simply because Starbucks can use Claude to jot down their very own CRM software program doesn’t imply they need to. We’ll give attention to customized fashions, high quality tuning, evals, observability, orchestration, and knowledge sovereignty.  

Molly Alter, associate, Northzone: A subset of enterprise AI firms will shift from product companies to AI consulting. These firms might begin with a particular product, similar to AI buyer help or AI coding brokers. However as soon as they’ve sufficient buyer workflows operating off their platform, they’ll replicate the forward-deployed engineer mannequin with their very own group to construct extra use instances for patrons. In different phrases, many specialised AI product firms will change into generalist AI implementers. 

Marcie Vu, associate, Greycroft: We’re very excited concerning the alternative in voice AI. Voice is a much more pure, environment friendly, and expressive means for individuals to speak with one another and with machines. We’ve spent a long time typing on computer systems and gazing screens, however speech is how we interact in the actual world. I’m desperate to see how builders reimagine merchandise, experiences, and interfaces with voice as the first mode of interplay with intelligence. 

Alexa von Tobel, founder and managing associate, Impressed Capital: 2026 would be the yr AI reshapes the bodily world — particularly in infrastructure, manufacturing, and local weather monitoring. We’re shifting from a reactive world to a predictive one, the place bodily methods can sense issues earlier than they change into failures.  

Lonne Jaffe, managing director, Perception Companions: We’re watching how frontier labs strategy the applying layer. Lots of people assumed labs would simply practice fashions and hand them off for others to construct on, however that doesn’t appear to be how they’re fascinated about it. We might even see frontier labs transport extra turnkey purposes straight into manufacturing in domains like finance, regulation, healthcare, and schooling than individuals anticipate. 

Tom Henriksson, common associate at OpenOcean: If I needed to decide one phrase for quantum in 2026, it’s momentum. Belief in quantum benefit is constructing quick, with firms publishing roadmaps to demystify the tech. However don’t anticipate main software program breakthroughs but; we nonetheless want extra {hardware} efficiency to cross that threshold.

Which areas are you trying to put money into? 

Emily Zhao, principal, Salesforce Ventures: We’re focusing on two distinct frontiers: AI coming into the bodily world and the following evolution of mannequin analysis.

Michael Stewart, managing associate, M12: Future datacenter expertise. For the final yr or so, we’ve been standing up a number of new investments that sign our curiosity in future “token manufacturing unit” expertise, with a watch in direction of what can actually advance how effectively and cleanly they run. That is going to proceed in 2026 and past, in classes that embody all the pieces inside the partitions of the information heart: cooling, compute, reminiscence, and networking inside and between websites. 

Jonathan Lehr, co-founder and common associate, Work-Bench: Vertical enterprise software program the place proprietary workflows and knowledge create defensibility, notably in regulated industries, provide chain, retail, and different complicated operational environments.

Aaron Jacobson, associate, NEA: We’re on the restrict of humanity’s capability to generate sufficient vitality to feed power-hungry GPUs. As an investor, I’m searching for software program and {hardware} that may drive breakthroughs in efficiency per watt. This may very well be higher GPU administration, extra environment friendly AI chips, next-gen networking approaches like optical, or rethinking thermal load inside AI methods and datacenters. 

Relating to AI startups, how do you identify that an organization has a moat? 

Rob Biederman, managing associate, Uneven Capital Companions: A moat in AI is much less concerning the mannequin itself and extra about economics and integration. We search for firms which might be deeply embedded in enterprise workflows, have entry to proprietary or constantly bettering knowledge, and show defensibility by means of switching prices, value benefits, or outcomes which might be tough to duplicate.  

Jake Flomenberg, associate, Wing Enterprise Capital: I’m skeptical of moats constructed purely on mannequin efficiency or prompting — these benefits erode in months. The query I ask: if OpenAI or Anthropic launches a mannequin tomorrow and is 10x higher, does this firm nonetheless have a purpose to exist?  

Molly Alter, associate, Northzone: It’s a lot simpler in the present day to construct a moat in a vertical class slightly than a horizontal one. The very best moats are knowledge moats, the place every incremental buyer, knowledge level, or interplay makes the product higher. These are considerably simpler to construct in specialised classes like manufacturing, building, well being or authorized, the place knowledge is extra constant throughout prospects. However there are additionally fascinating “workflow moats,” the place defensibility comes from understanding how a job or challenge strikes from level A to level B in an business.  

Harsha Kapre, director, Snowflake Ventures: For AI startups, the strongest moat comes from how successfully they remodel an enterprise’s current knowledge into higher choices, workflows, and buyer experiences. Enterprises already sit on extremely wealthy knowledge; what they lack is the flexibility to purpose over it in a focused, reliable means. We search for startups that mix technical experience with deep business data and might deliver domain-specific options on to a buyer’s ruled knowledge, with out creating new silos, to ship insights or automation that weren’t beforehand attainable.   

Will 2026 be the yr when enterprises begin to achieve worth from AI investments? 

Kirby Winfield, founding common associate, Ascend: Enterprises are realizing that random experiments with dozens of options create chaos. They’ll give attention to fewer options with extra considerate engagement.  

Antonia Dean, associate, Black Operator Ventures: The complexity right here is that many enterprises, regardless of how prepared or not they’re to efficiently use AI options, will say that they’re rising their investments in AI to clarify why they’re reducing again spending in different areas or trimming workforces. In actuality, AI will change into the scapegoat for executives trying to cowl for previous errors.

Scott Beechuk, associate, Norwest Enterprise Companions: We’re undoubtedly getting nearer. If final yr was about laying the infrastructure for AI, 2026 is after we start to see whether or not the applying layer can flip that funding into actual worth. As specialised fashions mature and oversight improves, AI methods have gotten extra dependable in day by day workflows.  

Marell Evans, founder and managing associate, Distinctive Capital: Sure, however nonetheless incremental. There may be nonetheless a number of iteration, and AI remains to be bettering to the purpose of having the ability to showcase pain-point options for enterprises throughout a wide range of industries. I imagine fixing simulation to actuality coaching will seemingly open up many alternatives for a choice of industries, each current and nascent. 

Jennifer Li, common associate, Andreessen Horowitz: There have been sensational headlines this yr about enterprises not seeing returns on their AI investments. Ask any software program engineer if they’d ever need to return to the darkish ages earlier than they’d AI coding instruments. Unlikely. My level is, enterprises are already gaining worth this yr, and it’ll multiply throughout organizations subsequent yr. 

Do you suppose enterprises will improve their AI budgets in 2026? 

Rajeev Dham, managing director, Sapphire: Sure, I imagine they may, although it’s nuanced. Quite than merely rising AI budgets, organizations will shift parts of their labor spend towards AI applied sciences or generate such sturdy top-line ROI from AI capabilities that the funding successfully pays for itself three to 5 instances over. 

Rob Biederman, managing associate, Uneven Capital Companions: Budgets will improve for a slim set of AI merchandise that clearly ship outcomes, and can decline sharply for all the pieces else. Total spend might develop, however it will likely be considerably extra concentrated. We anticipate a bifurcation, the place a small variety of distributors seize a disproportionate share of enterprise AI budgets whereas many others see income flatten or contract. 

Gordon Ritter, founder and common associate, Emergence Capital: Sure, however spend will focus. Enterprises will improve budgets the place AI expands on institutional benefits, and pull again from instruments that merely automate workflows with out capturing (and securing!) proprietary intelligence. 

Andrew Ferguson, vp, Databricks Ventures: 2026 would be the yr that CIOs push again on AI vendor sprawl. At the moment, enterprises are testing out a number of instruments for a single use case — month-to-month spend and switching prices are low in lots of instances, so the inducement to experiment is there — and there’s an explosion of startups centered on sure shopping for facilities like [go-to-market], the place it’s extraordinarily laborious to discern differentiation even throughout [proof of concepts]. As enterprises see actual proof factors from AI, they’ll minimize out among the experimentation price range, rationalize overlapping instruments, and deploy these financial savings into the AI applied sciences which have delivered.  

Ryan Isono, managing director, Maverick Ventures: In combination, sure, and there might be some shifting from pilots/experimental budgets to budgeted line gadgets. A boon for AI startups in 2026 would be the transition of enterprises who tried to construct in-house options and have now realized the problem and complexity required in manufacturing at scale.  

What does it take to lift a Collection A as an enterprise-focused AI startup in 2026? 

Jake Flomenberg, associate, Wing Enterprise Capital: The very best firms proper now mix two issues: a compelling “why now” narrative — often tied to GenAI creating new assault surfaces, infrastructure wants, or workflow alternatives — and concrete proof of enterprise adoption. $1 million to $2 million [annual recurring revenue] is the baseline, however what issues greater than that’s whether or not prospects view you and your product as mission-critical to their enterprise versus simply being a nice-to-have. Income with out narrative is a function; narrative with out traction is vaporware. You want each. 

Lonne Jaffe, managing director, Perception Companions: You must goal to indicate you’re constructing in an area the place the [total addressable market] expands slightly than evaporates as AI drives down prices. Some markets have excessive elasticity of demand – a 90% value decline results in a 10x improve in market dimension. Others have low elasticity, the place dropping the value can vaporize the market, so the purchasers hold all the worth being created. 

Jonathan Lehr, co-founder and common associate, Work-Bench: Prospects are utilizing the product in actual, day-to-day operations, and are keen to take reference calls and speak truthfully about affect, reliability, and shopping for course of, and so on. Firms ought to be capable to clearly present how the product saves time, reduces value or will increase output in a means that holds up by means of safety, authorized, and procurement evaluations. 

Michael Stewart, managing associate, M12: We (buyers) had been casting a uncertain eye in direction of [estimated annual recurring revenue] or pilot income till just lately. Now, it’s not seen as a lot of an asterisk as a lot because the buyer’s curiosity and willingness to guage an answer within the face of so many choices pushed their means. Getting these engagements and buyer buy-in when it comes to operating an analysis isn’t only a matter of forward-deployed engineers making it simpler for the client. It takes high quality and a profitable advertising message to do it in 2026. Buyers predict to see conversions change into the main a part of the story after 6 months of pilot use. 

Marell Evans, founder and managing associate, Distinctive Capital: Execution and traction. The very best sign is customers genuinely delighted to make use of the product, and the technical sophistication of the enterprise. We take a look at an enormous North Star of actual contractual agreements, 12+ months. Along with that, was this founder in a position to appeal to top-tier expertise to hitch their startup over opponents or the normal hyper-scalers?

What position will AI brokers play at enterprises by the top of 2026? 

Nnamdi Okike, managing associate and co-founder, 645 Ventures: Brokers will nonetheless be of their preliminary adoption section by the top of 2026. There are numerous technical and compliance hurdles that have to be overcome for enterprises to actually profit from AI brokers. There additionally have to be requirements created for agent-to-agent communication. 

Rajeev Dham, managing director, Sapphire: One common agent will emerge. At the moment, every agent is siloed in its position – for instance, inbound [sales development representative], outbound SDR, buyer help, product discovery, and so on. However by late subsequent yr, we’ll begin to see these roles converge right into a single agent with shared context and reminiscence, breaking down long-standing organizational silos, and enabling a extra unified, contextual dialog between firms and their customers. 

Antonia Dean, associate, Black Operator Ventures: The winners might be organizations that determine the fitting steadiness of autonomy and oversight rapidly, and that acknowledge agent deployment as collaborative augmentation slightly than a clear division of labor. Quite than brokers dealing with all routine work whereas people do all of the pondering, we’ll see extra refined collaboration between people and brokers on complicated duties, with the boundary between their roles constantly evolving. 

Aaron Jacobson, associate, NEA: Nearly all of data staff could have at the least one agentic co-worker they know by identify! 

Eric Bahn, co-founder, common associate, Hustle Fund: I feel that AI brokers will most likely be the larger a part of the workforce than any people in enterprises. Proliferating AI brokers is actually free and 0 marginal value. So why not develop by means of bots? 

What sorts of firms in your portfolio are seeing the strongest development? 

Jake Flomenberg, associate, Wing Enterprise Capital: The businesses rising quickest are those that recognized a workflow or safety hole created by GenAI adoption, then executed relentlessly on product-market match. In cybersecurity, it’s instruments addressing knowledge safety so LLMs can work together with delicate knowledge safely, and agent governance making certain autonomous methods have applicable controls. In advertising, it’s new areas like Reply Engine Optimization (AEO) — getting found in AI responses, not simply search outcomes. The widespread thread: these weren’t classes two years in the past, however are actually must-haves for enterprises deploying AI at scale. 

Andrew Ferguson, vp, Databricks Ventures: We’re seeing development tied to some widespread themes. One is firms that land with centered use instances — firms that begin with a narrower wedge (may very well be a centered goal persona or use case), actually nail it, change into sticky and earn the fitting to develop from the preliminary wedge. 

Jennifer Li, common associate, Andreessen Horowitz: Firms that assist enterprises put AI into manufacturing are doing nicely. Areas like knowledge extraction and structuring, developer productiveness for AI methods, infrastructure for generative media, voice and audio for media and apps like help or name facilities. 

What sorts of firms are seeing the strongest retention? 

Jake Flomenberg, associate, Wing Enterprise Capital: Firms with retention and enlargement share a sample: they clear up issues that intensify as prospects deploy extra AI. Sturdy retention comes from three issues: being mission-critical (elimination breaks manufacturing workflows), accumulating proprietary context that’s laborious to recreate, and fixing issues that develop with AI adoption slightly than being one-and-done. 

Tom Henriksson, common associate at OpenOcean: Retention is trickier to measure for youthful firms, however the highest retention we’re seeing is within the severe enterprise software program suppliers, particularly these enhanced with AI. instance is Operations1, which digitizes employee-led manufacturing processes end-to-end. These firms go deep into the client’s group, remodel how they function, and construct up proprietary knowledge and data that makes them very laborious to do with out. 

Michael Stewart, managing associate, M12: Startups serving the enterprise in knowledge tooling and vertical AI apps, with forward-deployed groups helping in buyer satisfaction, high quality, and product enchancment. This appears to be the profitable components that has been adopted by all main startups in these markets. Long term, the embedded groups would possibly recede as the purchasers begin to internalize the usage of AI of their organizations and workday practices. 

Jonathan Lehr, co-founder and common associate, Work-Bench: Retention is highest the place software program turns into foundational infrastructure slightly than some extent resolution. Authzed has sturdy retention as a result of authorization and coverage sit on the core of recent methods, and are extraordinarily expensive to tear out as soon as embedded. Courier Well being and GovWell act as methods of document and orchestration layers for end-to-end workflows, affected person journeys in healthcare, and allowing in authorities, which makes them deeply embedded as soon as dwell.