Investment Thesis 2025

Reactive to Proactive AI

The world of artificial intelligence is on the cusp of a profound shift – from reactive systems that merely respond to commands, to proactive agents that understand real-world context and act autonomously to assist us. This transition represents not just a technological leap, but a fundamental reimagining of how AI integrates into our daily lives. For investors, it presents a rare opportunity to back the platforms that will underpin the next generation of ambient intelligence.

The Limits of Today’s Reactive AI

Despite the breathtaking advances in large language models and generative AI over the past year, today’s most sophisticated AI systems are still fundamentally reactive. They may excel at analyzing data or generating content, but they operate as passive tools, waiting for human input before springing into action.

Consider a real-world scenario: Imagine an AI assistant helping you navigate a busy airport. Current systems can capably answer questions about your flight status or recommend nearby restaurants – but only if you think to ask. They can’t proactively alert you that your gate has changed, or notice that you’re running late and offer to rebook your ride. These are the moments where AI assistance would be most impactful, but they remain out of reach for today’s command-driven paradigm.

The Promise of Proactive AI

The next frontier of AI is about closing this gap between human and machine initiative. Proactive AI systems will function as always-on, ambient intelligences that can understand context and take action to assist us in real-time.

What does this look like in practice? Let’s return to our airport example. A proactive AI assistant would be constantly ingesting real-world data such as your location, flight information, visual and auditory input from your smartphone or smart glasses. As you go about your travel, the system would be continuously monitoring context, ready to step in at crucial moments:

• As soon as you enter the parking lot, your AI alerts you that your flight is delayed by 30 minutes and offers to push back your rental car reservation.
• When you reach the security line, your assistant notices the line is unusually long and pings your airline to see if you’re eligible for an express lane.
• Once you reach your gate, your AI recognizes your upcoming 4-hour layover and begins to intelligently fill your calendar – suggesting a restaurant that fits your dietary preferences, rebooking a work meeting, and even setting a reminder to pickup a gift for your nephew’s birthday.

This fluid, proactive assistance isn’t just a convenience – it’s a glimpse of a future where AI evolves from a passive tool into an active collaborator, anticipating our needs and autonomously acting to support us. Enabling this experience requires major advances in how AI systems ingest real-world context, reason about our goals, and take real-time action.

Building the Infrastructure for Proactive AI

While the proactive AI future is exciting, the reality is that much of the infrastructure needed to support it is still in its infancy. The existing ecosystem of AI tools and platforms is heavily optimized for the reactive paradigm – ingesting clean, structured data, and producing narrow outputs in response to clear commands.

Making the leap to proactive AI will require substantial innovation and investment across the data, software and hardware stack. And this is where we see the most compelling investment opportunities in the coming years. Rather than betting on individual AI applications, we believe the winning strategy is to back the enabling technologies that will power the entire next wave of proactive experiences.

Here are three key layers of infrastructure I am excited abot:

Ingesting Real-World Context

Proactive AI begins with capturing real-time context – which means a massive scale-up in the devices, sensors and data pipelines that connect AI to the physical and digital world.

On the hardware front, we’ll need a new generation of low-power, always-on sensors and edge AI chips to continuously capture multi-modal context without compromising device size or battery life. Imagine earbuds that not only play audio, but use embedded microphones and inertial measurement units to capture 3D audio and head movements, providing rich context on the user’s physical state and environment.

For digital data, the key will be building unified context lakes where a user’s real-time information across communication, productivity and smart home applications can be aggregated, processed and made available to fuel proactive AI decision-making. Expect to see the emergence of dedicated context fusion platforms that solve challenges like schema management, real-time indexing and privacy-first processing.

Software Systems for Agentic Action

Data alone isn’t enough for proactive AI – we also need software systems that can translate these context signals into intelligent action. This is an area ripe for startup innovation (and acquisitions by larger AI platforms) in the coming years.

First, new abstractions and tools for context reasoning will be essential – including knowledge graphs, state tracking systems and causal reasoning engines that can identify meaningful patterns amidst noisy real-world data. An airport assistant can’t proactively rebook your ride if it doesn’t have a structured model of how flight delays connect to ground transport logistics.

Second, innovations in action planning and execution will be critical to deliver on the promise of AI that doesn’t just recommend, but autonomously acts on your behalf. These action engines will need to combine techniques from constraint programming, automated planning and reinforcement learning to find optimal strategies to achieve user goals while managing uncertainty and resource constraints.

Finally, all of these systems will need to be built with a privacy-first mindset from the ground up. Techniques like on-device processing, federated learning, differential privacy and secure enclaves will transition from research to commercial deployment to ensure proactive AI systems can leverage intimate context without compromising user trust.

Training and Model Improvements

The final piece of the proactive AI puzzle is the continued advancement of core model architectures and training regimes. While today’s language and vision models are marvels of statistical pattern matching, the next generation will need to master causal reasoning, goal-oriented planning and real-time adaptation.

A key challenge will be developing simulators and training environments that can teach AI systems to reason about the complex interactions between physical contexts and human goals. Instead of learning only from clean, well-labeled datasets, proactive AI models will need to train inside realistic virtual worlds complete with physics, human behavior models, and real-time feedback loops.

We also expect the rise of heterogeneous model architectures that combine the pattern recognition power of deep learning with the reasoning capabilities of classical symbolic systems. Just as the human brain deploys specialized neural circuits for different cognitive tasks, future AI systems will intelligently route context signals to domain-specific model components for optimal inference and decision-making.

Where to Place Your Bets

Based on these infrastructural requirements, here’s where I plan to focus my proactive AI investments in 2025 and beyond:

1. Wearables and Sensor Hardware

• Edge AI chips and sensors for capturing rich physical context
• Low-power, high-bandwidth communication for real-time data transfer

2. Unified Context Platforms

• Tools for aggregating multi-modal context data from fragmented digital sources
• Real-time knowledge graphs and state tracking systems for structured context representation
• Privacy-preserving pipelines for extracting context insights without raw data exposure

3. Automated Reasoning and Action Engines

• Goal-oriented reasoning systems for contextual decision making
• Domain-specific constraint solvers and planners for real-time optimization
• Learning-based frameworks for adaptive, personalized action policies

4. Simulation Environments and Tools

• Realistic virtual environments for training context-aware AI agents
• Instrumentation and monitoring for AI systems operating in real-world settings
• Platforms for secure evaluation and head-to-head benchmarking of proactive AI

5. Next-Generation Model Architectures

• Hybrid approaches combining deep learning with symbolic knowledge representation
• Efficient, sparse models optimized for real-time inference on resource-constrained devices
• Federated learning and adaptation techniques for privacy-preserving personalization

The Coming Age of Ambient Intelligence

The transition from reactive to proactive AI represents a rare tectonic shift in the technology landscape. As with the rise of mobile and cloud computing, this new era of ambient intelligence will reshape not just the applications we build, but the underlying infrastructure that powers the entire ecosystem.

For investors, the key is to look beyond the hype around any single AI demo or product, and instead focus on the picks and shovels that will enable the broader proactive paradigm. By investing in the core infrastructural layers – capturing context signals, building action planning engines, and advancing core model capabilities – we can equip the next generation of entrepreneurs to bring transformative proactive experiences to life.