Blume 2026 Perspectives | From the Year of Agents to the Year of Agentic Workforce

This has been a year of intense learning and evolution for us at Blume in the context of investing in B2B (or, Enterprise) AI applications. 

The Relentless March of LLMs Continues

We’ll start with one of the year’s most striking but well represented analogies : 

(credit — X post)

An unsaid statement here — it may appear that AI progress in 2025 was simultaneously exciting and disappointing — depending on the way you slice it. No, we didn’t get AGI and infinite consumer surplus. However, we did make some solid progress on most benchmarks (though benchmaxxing is real) courtesy of inference scaling with reasoning models, and RLVR (reinforcement learning with verifiable rewards). 

Two charts succinctly capture the progress:

We think it is very hard to understate how real and massive the impact of agents can be, assuming the combination of both of these trajectories continues on a similar path into 2026. As agents gain the ability to take on longer, more complex tasks with deeper and richer context and the ability to do this at scale at lower costs than per hour/​per minute of white collar workforce, we will begin to see more and more displacement of human labor in workflows that are repetitive, complex but require highly paid professionals. As the latest OpenRouter report shows, this shift is no longer theoretical — the fastest growing usage patterns are ​“agentic inference” with AI application developers building workflows where models are not ​‘single-prompt answering monkeys’ but rather acting in more elaborate, connected sequences of steps.

The restatement of TAM into a much broader scope, like the below chart shows, we believe is very real and we expect to see more inroads into this in 2026 (cases in point being admins and receptionists at clinics, IT service staff, salesforce administrators etc).

What investment themes are most exciting for us?

Vertical AI applications continue to be amongst the fastest-growing subcategories that piqué our interest. 

From healthcare to home services, logistics to law, financial services to real estate — there is high market pull from industries desperate for an agentic workforce. While the evolution of these applications until now has been primarily around processing information and basic reasoning, as the data flywheel within each application/​industry continues to spin — we will increasingly see more collaborative use-cases across humans and other agents — the early substrates of an ​“agentic workforce”. Will we see multiplayer agents navigating across counterparties and across different functions internally within an enterprise? We postulate this isn’t too far away.

Systems of record within enterprises have been among the most successful horizontal software businesses over the last couple of decades. We’re beginning to see a very real shift away from legacy systems of record being where the buck stops — where workflows sit. As Jared Sleeper put it beautifully — the battle is to become the ​“System of Context” — encompassing reasoning, underlying data records, and human-in-the-loop. How much the existing systems of record will re-architect themselves to stay competitive in this fight is questionable — we think we’ll continue to see new-age AI-native companies eating away more significant chunks of their market. Making migration away from a legacy SoR easier needs to be a critical part of their right to win — coupled with the pitch that they are an order of magnitude less expensive wrt TCO.

Observations from working with founders

This year, we saw increasingly more founders consciously choose to start building x‑border very early on in their journeys. A decade ago, it used to be the norm that founders ​“building in India for the world” would be on a plane 30 – 40% of the time.. with that ratio increasing as they reach $100M in ARR. We’re now seeing founders choose to relocate lock, stock and barrel to their target markets (usually the US) sub-$1M ARR itself, while a majority of their tech/​engg orgs continue to be scaled out of India. 

This dramatic shift is being driven by two forces:

• Within every B2B software category, there is intense competition. From up-starts to legacy players who are trying to reinvent themselves, from a buyer (usually an enterprise CXO/VP) perspective — many pitches tend to look similar. If founders don’t show up to pitch meetings, industry conferences, round tables, and other engagements in person — it’s incredibly hard to gain mindshare (they will be”out of sight, out of mind”).
• Given the pace at which the underlying technology is transforming and shifting, enterprise buyers have gone back to relying significantly on softer aspects such as credibility, brand, referenceability, staying power etc as part of the buying process in addition to ensuring that the solution is best-in-class with the tech. This means the classic enterprise motions of partner-led sales, wine-and-dine, building ​“insider” networks etc — is back at the top of founder priorities. This necessitates them to be in-market on a continuous basis.

We are seeing a fundamental re-framing in many cases of the founders’ ambition — while $100M ARR used to be the ultimate aspirational goal last decade, now that has increasingly shifted to being a $1B ARR mark. We expect to see more and more world-class founders choose problems and industries that can support that kind of revenue and massive scale of ambition eventually. 

We fully expect the enterprise AI story in 2026 will encompass ever deeper customer/​buyer collaboration, deep workflow integrations, and endurance to cover the last mile of implementation and adoption. But this is happening already, and we can see that in front of our eyes with many special founders that we are privileged to be partners with. We need only to look at the green shoots — founder motivation, aggressiveness, hunger to win, customer obsession — to see that a fantastic 2026 is imminent. With that, we wish you a happy new year!