Within Reach: 2026’s Breakthrough Ideas in Work AI

Even as AI delivers unprecedented gains in productivity, its real promise isn’t just about doing things faster and cheaper. It’s about reshaping work to be more human, purposeful and rewarding.

In this installment of our Breakthrough Ideas series, we spotlight AI’s potential to radically improve how we work. We see it expanding the boundaries of who technology can serve, opening doors to people, professions, and markets that have long been considered too small, too fragmented, or too “unscalable” to matter. In 2026, the most important shift in AI won’t be about efficiency alone, but about who gets access to capability — and what becomes possible when it does.

Last week, our healthcare team explored the opportunities around a new operating system for longevity care, the first AI doctor, and a predictive model of care. Here, our work investors share their excitement around how AI will empower SMBs, unlock big market opportunities in microvertical industries, and set a new standard for what’s possible with human-agent coordination.

See here for our other Breakthrough Ideas in Learning and Health.

1. Niche No More: The Rise of Microvertical AI

James Kim, Partner

For decades, vertical software followed a blunt economic rule of thumb: only markets with a “big enough” total addressable market — often a hand-wavy billion dollars or more — were worth the cost of building and selling into. Hence we ended up with vertical software for logistics, hospitality, and healthcare, but almost never for libraries, churches, or Chinese takeout restaurants.

In 2026, we will start to see this constraint erode with the emergence of “microvertical AI”: software built for specific industries that were once considered too small or niche to matter. Three AI-driven unlocks are making this possible:

1. dramatically lower cost and time to build custom software;
2. automation across go-to-market, implementation, and user support; and
3. immediate, outcome-driven ROI, as AI moves beyond digitizing workflows to actually doing the work.
   
   (Think: a coding agent that spins up custom landing pages, a marketing agent that autonomously runs local ad campaigns, a phone agent that takes orders, or an email agent that reaches out to best-fit job candidates.)

Importantly, the traditional VC objection (”TAM is only $100M!”) misses how the winners in this new era will operate. They won’t be single-vertical point solutions. They’ll be lean teams with a reusable AI agent stack that can adapt to different niches, each of which can be a small, defensible monopoly, built through deep workflow integration and proprietary data, and reinforced by state-of-the-art memory, context, and retrieval.

2. Giving Fortune 500 Powers to Main Street

Enzo Cavalie, Principal

In 2026, the capability gap between a Fortune 500 company and a local small business will finally collapse. For years, we have built AI to help large corporations save 10% on operational costs, but we’ve ignored the much larger opportunity: giving “big company superpowers” to the 90% of businesses that still run on Excel, email and gut instinct.

Historically, access to expertise required some upfront expense and commitment — a “minimum order quantity” of sorts. You either paid a $10,000-per-month retainer for a CFO or marketing agency, or you went without. Many SMBs chose the latter, operating with limited visibility and stalled growth, not because they didn’t value expertise, but because the cost made it inaccessible.

AI solves this lumpiness by turning expertise from a fixed salary or retainer into an on-demand, per-use token cost. A bakery can buy 15 minutes of CFO-level guidance to evaluate a purchase. A local shop can launch a sophisticated ad campaign without committing to an agency retainer fee. This dramatically lowers the entry price to building world-class business operations. 

We are investing in founders who package complex professional services into accessible, high-trust AI agents, turning historically “non-consuming” SMBs into the largest market in the world. The most valuable AI companies won’t just be selling efficiency to the Fortune 500. They will become the fractional C-Suite for Main Street.

3. The Future of AI Work is Coordination, Not Autonomy

Esteban Sosnik, General Partner

As labor shortages persist and AI shifts from generating content to executing work, every company will deploy digital workers. The winning systems, however, will not be fully autonomous agents. They will be coordinated ones.

The most valuable AI companies won’t build agents that can do everything passably. They will build agents that do specific work dependably and know how to work with others. Capability is rapidly commoditizing. Dependability — judgment, timing, and trust — is not. That will be the moat.

For mission-critical tasks, agent autonomy will be deliberately constrained. In regulated and high-stakes environments, people need to stay in control. The winning interface won’t be “full automation,” but high-signal escalation: agents that know when to act independently and when to involve humans, bringing the right context, tradeoffs, and next steps. Human-agent coordination is not a fallback, but a core feature.

At the same time, digital workers will become company-specific. They will learn “how we do things here” from unstructured operational data — emails, tickets, calls — and codify best practices. As agents specialize and own discrete steps, coordination becomes the central problem — and opportunity. Value will shift to the infrastructure that enables them to work together reliably: managing handoffs, verification, shared context, and escalation. The next generation of AI platforms won’t be defined by smarter individual agents, but by the coordination layer that governs how agents and humans work together.

I’m excited to back founders building this future of digital labor: specialized software workers and the coordination systems around them that earn trust not by being invisible, but by knowing when to act, when to ask, and how to collaborate — at a token’s cost.

If you’re building in any of these areas, we’d love to hear from you!