2026 Investment Themes: Entering the Era of Embedded Intelligence

Why the next phase of AI is about systems, not prompts

Over the past few years, AI has transitioned from what was once a cutting-edge technology reserved primarily for researchers and engineers into something that even our grandparents now use and discuss around the holiday dinner table. As with any foundational technology shift, adoption has happened faster than expected, and familiarity has set in quickly.

Despite this rapid normalization, we are still early in understanding how AI will ultimately reshape work, organizations, and entire industries.

The pace of innovation at the model layer, or intelligence layer, is unlike anything we have seen in technology history. But model capabilities have skated far ahead of our collective ability to harness them into a broad set of real-world applications that are deeply embedded into how we work and live.

In 2026, we are entering a phase where AI is no longer just a new tool, but a co-creator, collaborator, and partner embedded directly into our systems, workflows, and environments. This is a transition toward embedded intelligence; systems where AI agents are built directly into products, workflows, and physical environments, capable of executing actions on behalf of humans rather than merely responding to prompts.

Most AI products in use today still live inside standalone applications with primarily chat or Q&A-based interfaces such as ChatGPT. While these products have led to a paradigm shift, they represent an intermediate step rather than the end state. A deeper transition is now underway:

• From AI assistants to autonomous collaborators embedded directly into how work gets done

• From data infrastructure to governed, reliable AI systems that can be trusted in production

• From single-model, narrow use cases to multi-modal embedded intelligence capable of tackling complex problems

I tend to agree with Jensen Huang’s framing that “AI won’t take your job, but someone using AI will.” And contrary to what some have suggested, venture capital is not immune to this transformation of work, especially given how competitive and information-dense our market is. I’m actively applying this mindset in my own work and in how we are evolving systems and processes at Crosslink. The opportunity is not to react to AI adoption, but to build, adapt, and upskill ahead of the curve. Organizations that do so early have a meaningful advantage.

Theme 1: Collaborative AI + Human Systems Are the Winning Combination

Models continue to improve with each release, adding more firepower to an already powerful foundation. In 2026, progress is no longer just about better models, it’s about better collaboration between humans and machines.

We are moving from AI assistants and copilots that sit out of band from the work toward autonomous collaborators that:

• Are context-aware

• Persist across time with short-term and long-term memory

• Understand goals and constraints

• Learn preferences and workflows continuously

These systems operate as ongoing partners rather than transactional tools.

AI + human collaboration will become universal once it is properly embedded into where work actually happens. This is not limited to developers or customer support teams. Writers, designers, researchers, analysts, investors, and operators will all work alongside AI collaborators that help ideate, draft, simulate outcomes, pressure-test decisions, and improve output quality. The result is not replacement, but an expansion of creative surface area and more productive work.

The best products will feel less like tools and more like thinking partners: deeply contextual, personalized, and continuously improving.

Companies that take this system-level approach also build natural defensibility. As AI collaborators become embedded in workflows, they generate data flywheels that compound product quality and increase switching costs over time.

Theme 2: Vertical AI Begins to Take Flight

While horizontal AI platforms from Google, Microsoft, OpenAI, Anthropic, Glean, and others will continue to improve, some of the most compelling opportunities lie in vertical markets. Vertical AI enables startups to build deeply integrated, multi-modal systems that are tailored to the unique workflows, incentives, and edge cases of a specific industry. Horizontal tools will feel generic by comparison.

Founders with deep domain expertise are best positioned to build step-function improvements in user experience. In these systems:

• AI handles orchestration, repetition, and execution

• Humans focus on judgment, creativity, and high-leverage decisions

Vertical software companies have historically been underestimated in how large they can become. They often begin with a focused wedge product that solves a painful, underserved problem, then expand as they build trust with customers. With modern LLMs, voice, vision, and multimodal capabilities now broadly accessible, we are likely to see a surge of vertical AI companies that land quickly and expand aggressively within large industry verticals.

Theme 3: AI-Native Organizations Take Over

This theme is less about a specific sector and more about how companies are structured and operated. AI is transitioning from surface-level productivity tooling to core operational infrastructure.

AI-native organizations share several defining characteristics:

• Humans and agents collaborate by default

• Knowledge compounds through continuous feedback loops

• Decision-making is iterative rather than episodic

• Coordination costs fall dramatically

This is not merely a tooling shift; it is an organizational redesign. Teams that fully embrace this model will consistently outperform competitors that treat AI as an add-on rather than a foundation.

Theme 4: Responsible AI Infrastructure Becomes Mandatory

As AI systems become more autonomous and move into production, trust becomes the bottleneck. No CIO or executive will green-light AI deployments without confidence in reliability, safety, security, and cost control.

Traditional software infrastructure was designed for deterministic systems. AI introduces probabilistic behavior, requiring a new generation of infrastructure that is:

• Auditable

• Explainable

• Governable

• Secure by design

• Efficient

Key pillars of responsible AI include auditable model behavior, system-level risk management, and embedded governance controls. Trust must be designed into AI systems from the start, not bolted on after deployment. The companies that lead the way will enable autonomous AI agents without sacrificing control, a prerequisite for enterprise adoption at scale.

Theme 5: Embodied AI Arrives in the Physical World

Robotics has long felt like it is perpetually “ten years away,” but that timeline now feels meaningfully compressed. Advances in embodied AI allow machines to:

• Understand physical environments

• Learn new tasks dynamically

• Execute complex actions before they are defined

The convergence of large pre-trained models, reinforcement learning, and compact models deployable at the edge has the potential to unlock new industrial and commercial applications. While robotics remains capital-intensive and difficult to scale, technology readiness is finally aligning with real-world opportunity. Capital sources at the growth and later stages also appear more willing than they have been historically to support companies that demonstrate progress along their journey.

Closing Thoughts

One personal challenge I continue to grapple with is how crowded AI investing has become. From 2010-2020, AI still felt like frontier technology. Today, mainstream adoption has made AI the most competitive investing theme I’ve encountered.

As a result:

• Many startup pitches operate in crowded markets

• Competition comes from both well-funded startups and incumbents

• Traditional automation categories feel increasingly saturated

Because of this, I’m prioritizing founders with deep domain insight; builders who understand workflows and incentives at a level that enables them to reimagine products from the ground up and see the future before it has arrived. While the differentiated opportunities are harder to find, they are more valuable than ever when discovered.

For the first time in over a decade, there is a credible threat to the existing software ecosystem and a meaningful opportunity to expand into the much larger services market. It’s an exciting and important moment to be building and investing in the next generation of companies shaping our future.

Cheers to 2026!