2025 Investment Themes: Human + Machine Superpowers

The power of the symbiotic relationship of Humans and AI, and my top 3 startup investment themes for 2025

As we enter 2025, the venture market presents a striking dichotomy: soaring valuations for hot AI companies alongside a highly challenged environment for most other sectors. This has ignited a debate within the VC / startup ecosystem as to whether AI is fueling yet another bubble, and if so, to what extent. In parallel, there is a much broader societal debate happening both inside and outside of the tech community on the rise of AI being good vs. bad for us all as humans.

While AI is often portrayed as a threat, I believe this moment instead presents an opportunity for us to gain control back from the technology we’ve created and harness its power for a positive impact on humans and our society. For startups and other participants in the tech ecosystem, I see a generational opportunity to combine the collective strengths of what AI models are capable of alongside the uniquely human traits we possess to create new categories of businesses and reinvent age-old industries.

The "software-eating-the-world" phenomenon has, in many ways, made us subservient to technology. We spend much of our time entering data into systems, reading / responding to emails, searching the web for information, coding software, and so on. The current wave of AI has the potential to liberate us from many of these mundane tasks we’ve become accustomed to and instead allow us to work alongside technology in a more symbiotic relationship. With the widespread availability of cutting-edge AI models and powerful computing infrastructure, the stage is now set for an explosion of new applications that will completely reinvent technology user experiences as we know them. So while the venture market is forward investing and overvaluing most of the AI-native ecosystem that is still in its infancy, I believe we’re in the early innings of one of the largest value creation events the startup / tech ecosystem has ever seen.

Technology is at its best when we combine our human superpowers with machine superpowers.

Human Superpowers:

• Creativity & Ingenuity

• Imagination

• Empathy & Emotional Intelligence

• Trust & Connection

• Consciousness & Self-Awareness

• Adaptability

AI / Machine Superpowers:

• Accuracy & Precision

• Speed & Efficiency

• Predictability & Reliability

• Endurance

• Data Processing & Analysis

For sci-fi enthusiasts, I am surely setting myself up for debate on several of these. I recently rewatched one of my favorite films, Ex Machina, and we can debate the philosophical merits of whether machines can truly possess consciousness… but we’ll save that for another day.

There is a common belief that AI’s primary function is to replace humans altogether. I believe this perspective is flawed because it misses the distinct advantages that humans have and machines do not possess. A fun example of this in the real world is in art and music. While AI-powered tools can assist artists in their creative processes, true artistic brilliance stems from personal experiences, imagination, and unique human perspectives. The masterpieces of Picasso and Van Gogh, and the iconic music of Bob Dylan and The Beatles, are revolutionary not because of their precision or efficiency, but because they reflect the unique human experiences and creative vision of their creators. Their out-of-the-box thinking was a step function different from anything that came before it, making it impossible for a machine to model.

The true power of AI lies in its ability to augment human capabilities. By leveraging the strengths of both humans and machines, we can create a symbiotic relationship where technology enhances our lives rather than diminishes them. I am particularly drawn to startups that explore this human-machine symbiosis, developing solutions that empower humans to leverage our unique strengths while harnessing the power of AI to overcome our limitations. I believe the companies that seek to combine the inherent advantages of humans + machines will be at the forefront of innovation and create enduring value.

Theme #1: AI Assistants for All

I envision a future where every individual and professional role is augmented by a personalized AI assistant. These assistants will not be generic; they will be tailored to each user's unique needs and the specific context of their work.

For employees within an organization, these AI assistants will be pre-configured with relevant integrations, datasets, and AI models, all secured by robust role-based access controls. Each employee will have the tools they need to maximize productivity in their role, while ensuring that sensitive data is being used responsibly within the organization. Similarly, for specific industries such as healthcare, finance, or manufacturing, AI assistants will be optimized with specialized models, data sources, and integrations that address the unique challenges and requirements of that sector. Depending on the use case, the assistants could be best offered as a simple copilot or a full-fledged multimodal agent to complete tasks end-to-end working with nearly any data format.

In the consumer realm, we can expect to see highly personalized AI assistants that seamlessly integrate into our daily lives. These assistants will not just automate individual tasks but will learn our preferences and habits to proactively anticipate our needs and streamline our routines.

In healthcare, there are several compelling examples of this potential. Doctors are currently burdened by administrative tasks such as note-taking, data entry, and reviewing patient records. AI assistants can alleviate this burden by automating these processes, allowing doctors to dedicate more time to patient care. AI assistants are already underway that can transcribe doctor / patient interactions, analyze medical images, and synthesize complex patient histories, enabling doctors to make more informed and timely decisions at the point of care.

This transformative potential extends far beyond healthcare. We can envision AI assistants tailored for every role in the enterprise, from sales and marketing to engineering and finance. Imagine sales reps automatically receiving key technical product information at the point of sale while customers are asking the hardest questions. Or a CFO querying in natural language the status of key financial metrics in real-time without asking the data team to compile a custom dashboard. We can envision similar AI assistants tailored for manufacturing line operators, construction workers, home services reps, biotech researchers, and on and on. These assistants will leverage relevant internal and external data sources to provide real-time contextual insights, automate routine tasks, and empower professionals to perform at their highest level.

Crosslink Portfolio examples: Overjet, Inscribe, Argon AI, UserEvidence, Leapfin

Theme #2: Voice as an Interface

The ability of AI to both understand and generate human language with remarkable accuracy has potential to disrupt current incumbent software and unlock entirely new categories of applications. In many cases, I believe we are moving towards a future where typing becomes less prevalent, replaced by voice command and response for a more intuitive user experience.

This shift will revolutionize how we interact with technology. We are already seeing call centers powered by AI that can engage in more natural, error-free conversations with customers, minimizing human intervention. We may even see the rise of a new "voice-first" economy, leading to the development of novel devices, infrastructure, and coding frameworks specifically designed for voice interactions.

This paradigm shift will redefine our user interfaces, moving beyond traditional point-and-click interactions towards more conversational and intuitive experiences. The impact may be the most transformative in our physical world industries. Consider a construction worker using voice commands to access and update information on the go, hands-free. Or a field service technician seamlessly interacting with their systems of record while on site without having to hunch over their smartphone or tablet, entering structured information into fields. For knowledge workers, voice interfaces offer significant advantages as well. I personally have started to use voice dictation to draft emails or internal documents, and leverage AI-powered tools to refine and edit the text where necessary. I imagine there will be more complete applications with UI / UX flows that will incorporate individual needs for a role or industry vertical.

While existing software companies are gradually integrating voice capabilities, startups have a unique opportunity to swiftly capitalize on this trend. By prioritizing voice as the primary interface from the ground up, they can develop innovative solutions tailored to specific industries and roles. This approach can unlock entirely new possibilities and address unmet needs that traditional software providers may have overlooked or will not execute on because of the innovator’s dilemma that stands in their way. I am eager to see how founders leverage their domain expertise to build groundbreaking voice-driven applications that redefine how we interact with technology.

Theme #3: “Collapsing the stack” of Infrastructure <> Applications

The emergence of the new AI stack presents exciting opportunities for startups at the infrastructure layer as well. One observation I have is the potential for AI to “collapse the stack”, blurring the lines between applications and underlying infrastructure.

Historically, a significant portion of software development has focused on integrating disparate data sources and building complex data pipelines. However, with advanced AI models capable of processing diverse data formats (documents, voice, images, databases) with minimal pre-processing, the need for intricate integrations may diminish. Will we even need to build custom integrations anymore? We may be looking at a future of integration-less unstructured data access where applications and models can directly utilize data from all sources through a simple interface for developers or even business users. AI bots with the appropriate security clearances could even authenticate as users would, query data from various systems of record or databases, and transform data types on the fly without need for an API or integration between systems.

This collapsing of the stack would have profound implications. Infrastructure providers may expand their offerings into the application layer, while application providers may offer their customers more customization and flexibility over their data, code, and infrastructure. It isn’t yet clear how this will all shake out, but I do believe the shift demands a re-evaluation of the software development landscape. Satya Nadella shared some interesting insights on this dynamic in a recent interview here.

The rise of AI-powered tools like Code Generation and automated QA further accelerates this trend. Developers can now utilize natural language and intuitive interfaces to build and test applications more rapidly. This empowers individuals like designers to play a more active role in the development process, translating their creative visions directly into functional code. Consequently, the focus of infrastructure and development teams may shift towards optimizing and scaling these AI-generated prototypes, ensuring they are production-ready, secure, and compliant. We are already seeing a new landscape of observability, security, and data privacy tools emerge, and I believe this will be a hotbed area for startups who capitalize on the opportunity set that is opening up.

Another trend I’m seeing in the evolving AI landscape is the growing importance of professional services as an offering by software infrastructure providers. Enterprises grappling with the complexities of AI adoption often seek more comprehensive solutions that combine infrastructure and application components instead of purchasing individual point products. By leveraging AI to rapidly prototype and develop custom applications, infrastructure providers can offer integrated solutions that address specific enterprise needs more holistically. This clashes with the traditional model of selling infrastructure and application components separately, furthering the potential to collapse the software stack. While the AI ecosystem will continue to mature with specialized best-in-class solutions emerging for each layer, the current trend suggests a preference for fewer, more integrated vendors among enterprises.

Crosslink Portfolio examples: Griptape, Knapsack, Qualiti, Abstract Security

Conclusion

I’m excited to be investing in a future of work that is defined by a more symbiotic relationship between humans and AI, where technology empowers individuals to reach their full potential while enhancing efficiency and productivity for their organizations. By offloading mundane and repetitive tasks to AI, humans will be liberated to focus on unique superpowers such as creativity, empathy, and critical thinking. This shift will not only enhance individual productivity but also foster a more fulfilling and meaningful work experience for us all.

As a VC, it’s easy to be skeptical anytime you see this level of widespread excitement around one category. And it is likely that we’re nearing the peak of the AI hype cycle. However, history has shown that dismissing transformative technologies (Internet in the 90s, mobile in the 2000s, cloud in the 2010s) can be a costly mistake. The most important thing as an early-stage investor is to stay focused on where the value accrues and find the best founding teams out there building in that arena. My current view is that startups that truly understand the power of human-machine symbiosis and are reimagining user experiences will build businesses that sustain through the trough of disillusionment to become the next great companies of our time.

Cheers to 2025!