At Dellecod Software, we spend a lot of time thinking about what’s next—not as speculation, but as preparation. When we look toward 2026 and the evolving shape of artificial intelligence, three themes stand out. They don’t just point to new possibilities, but to shifts already underway in how we approach science, connection, and business itself.
The first idea that resonates with us is the emergence of autonomous or semi-autonomous labs. Oliver Shu’s insights reveal more than a simple upgrade to traditional lab automation. What's happening goes deeper: a co-evolution between human reasoning and machine intelligence. AI is moving beyond rote analysis. Increasingly, it is learning to hypothesize, to design experiments, and to iterate in closed loops—not to replace scientists but to accelerate their work, with robots conducting tests and AI models steering inquiry.
We find that the most compelling applications are in the areas where experimentation is both expensive and slow: drug discovery, materials development, chemical synthesis. The prospect of “AI collaborators” in these sectors is deeply promising. But it’s not a solved problem. Any system operating in these domains will need to be not only accurate, but interpretable. Scientists don’t just want answers—they need to understand how and why those answers emerged. Trust in the system will come from transparency, not just performance.
Public-private partnerships are playing a key role here—and rightly so. Genesis Mission in the U.S. and DeepMind’s work in the UK suggest that early national coordination is likely to drive meaningful results, especially where the infrastructure costs of autonomous labs might still be out of reach for individual companies or smaller institutions.
The second shift, as Brian Kim puts it, is more emotional than technical—though it’s enabled by tech all the same. AI’s role in consumer life is transitioning from utility to intimacy. Until recently, most consumer-facing AI served practical ends: scheduling meetings, summarizing emails, offering recommendations. But the big swing now is toward connection. AI that understands us—our histories, our personalities, our inner worlds—is starting to shape how we relate to each other.
It’s easy to critique this trend, to worry about synthetic emotions or performative understanding. But there’s also an opportunity here to use passive signals—photos, location history, music preferences—not to sell more, but to surface shared memory and facilitate real conversation. Imagine a moment where two people's AI agents negotiate a context to meet or reminisce together—spontaneously surfacing a memory from a shared experience, or proposing a local event based on mutual interests. Done right, this could help people feel more seen, not less.
For startups building in this space, the challenge is differentiation. Large platforms have scale, but they often lack emotional texture. New entrants, meanwhile, have the freedom to experiment with form and tone. The most successful products in this next era might feel less like tools and more like companions or catalysts—encouraging introspection, empathy, or even reconnection.
Lastly, we come to what may be the most overlooked force: AI as a business model force multiplier. What David Haber notes—and what we’ve witnessed firsthand—is that the companies getting ahead with AI aren’t just saving money. They’re building new kinds of defensibility.
Take EVE, for example. It’s turning legal AI into a revenue engine by helping plaintiff-side lawyers scale up case volume and optimize outcomes. Because they only get paid if the client wins, performance matters—and AI plays a vital part in predicting which cases to take, what they’re worth, and how to frame them. The more outcomes they generate, the stronger the model becomes. That's a virtuous loop worth noting.
Or consider Salient. Its multilingual AI voice agents don’t just replace humans in call centers. They increase collection rates by appearing more accessible, less emotionally charged, and more scalable—especially in settings where language access was previously a barrier. Neither company thinks of AI as an add-on. For them, AI is the business, not just in the back-office, but as part of the economic engine. Over time, these kinds of models develop their own network effects. Proprietary data leads to better models. Better models bring more users. Scale improves accuracy. Accuracy creates trust—and that trust becomes the moat.
At Dellecod, our reflections on these three vectors—scientific collaboration, emotional AI, and business model synergy—help us think more clearly about where the field is going. They remind us that the most transformative AI applications are often not the most visible. The most meaningful shifts rarely happen in headlines. They emerge through quiet integrations, thoughtful design, and a steady focus on real outcomes.
If 2023 was the year AI broke into the public’s imagination, 2026 may be the moment it finds its tone. Less spectacle, more substance. And in that space, there’s room—still—for invention.
The first idea that resonates with us is the emergence of autonomous or semi-autonomous labs. Oliver Shu’s insights reveal more than a simple upgrade to traditional lab automation. What's happening goes deeper: a co-evolution between human reasoning and machine intelligence. AI is moving beyond rote analysis. Increasingly, it is learning to hypothesize, to design experiments, and to iterate in closed loops—not to replace scientists but to accelerate their work, with robots conducting tests and AI models steering inquiry.
We find that the most compelling applications are in the areas where experimentation is both expensive and slow: drug discovery, materials development, chemical synthesis. The prospect of “AI collaborators” in these sectors is deeply promising. But it’s not a solved problem. Any system operating in these domains will need to be not only accurate, but interpretable. Scientists don’t just want answers—they need to understand how and why those answers emerged. Trust in the system will come from transparency, not just performance.
Public-private partnerships are playing a key role here—and rightly so. Genesis Mission in the U.S. and DeepMind’s work in the UK suggest that early national coordination is likely to drive meaningful results, especially where the infrastructure costs of autonomous labs might still be out of reach for individual companies or smaller institutions.
The second shift, as Brian Kim puts it, is more emotional than technical—though it’s enabled by tech all the same. AI’s role in consumer life is transitioning from utility to intimacy. Until recently, most consumer-facing AI served practical ends: scheduling meetings, summarizing emails, offering recommendations. But the big swing now is toward connection. AI that understands us—our histories, our personalities, our inner worlds—is starting to shape how we relate to each other.
It’s easy to critique this trend, to worry about synthetic emotions or performative understanding. But there’s also an opportunity here to use passive signals—photos, location history, music preferences—not to sell more, but to surface shared memory and facilitate real conversation. Imagine a moment where two people's AI agents negotiate a context to meet or reminisce together—spontaneously surfacing a memory from a shared experience, or proposing a local event based on mutual interests. Done right, this could help people feel more seen, not less.
For startups building in this space, the challenge is differentiation. Large platforms have scale, but they often lack emotional texture. New entrants, meanwhile, have the freedom to experiment with form and tone. The most successful products in this next era might feel less like tools and more like companions or catalysts—encouraging introspection, empathy, or even reconnection.
Lastly, we come to what may be the most overlooked force: AI as a business model force multiplier. What David Haber notes—and what we’ve witnessed firsthand—is that the companies getting ahead with AI aren’t just saving money. They’re building new kinds of defensibility.
Take EVE, for example. It’s turning legal AI into a revenue engine by helping plaintiff-side lawyers scale up case volume and optimize outcomes. Because they only get paid if the client wins, performance matters—and AI plays a vital part in predicting which cases to take, what they’re worth, and how to frame them. The more outcomes they generate, the stronger the model becomes. That's a virtuous loop worth noting.
Or consider Salient. Its multilingual AI voice agents don’t just replace humans in call centers. They increase collection rates by appearing more accessible, less emotionally charged, and more scalable—especially in settings where language access was previously a barrier. Neither company thinks of AI as an add-on. For them, AI is the business, not just in the back-office, but as part of the economic engine. Over time, these kinds of models develop their own network effects. Proprietary data leads to better models. Better models bring more users. Scale improves accuracy. Accuracy creates trust—and that trust becomes the moat.
At Dellecod, our reflections on these three vectors—scientific collaboration, emotional AI, and business model synergy—help us think more clearly about where the field is going. They remind us that the most transformative AI applications are often not the most visible. The most meaningful shifts rarely happen in headlines. They emerge through quiet integrations, thoughtful design, and a steady focus on real outcomes.
If 2023 was the year AI broke into the public’s imagination, 2026 may be the moment it finds its tone. Less spectacle, more substance. And in that space, there’s room—still—for invention.