One question kept coming back this quarter: how much of its AI does a company want to operate on its own terms? It showed up in a funding round, in a benchmark result, and in deployment conditions we had not anticipated. Here is Q3.
Series C: Customer Groups Became Shareholders
On September 17, Enhans announced a $38 million Series C round, bringing total funding to about $60 million. Timefolio Asset Management and Stonebridge Ventures co-led the round.
Just as notable as the size of the round was who invested. Affiliates of the POSCO, LG, and Lotte groups joined as strategic investors. All three groups already run Enhans software in production. This is the first time affiliates of customer groups have made a strategic investment in Enhans.
The product came first, and the investment decision came after. The same pattern appeared across three groups in three different industries at roughly the same time. This quarter showed us that trust built in live operations can translate into investment.
A Benchmark for Execution
In August, ACT-2, Enhans's Computer-Use Agent, ranked No. 2 globally in the automated evaluation of Online-Mind2Web, based on results published in August 2026.
Online-Mind2Web measures whether an AI agent can understand a user's request on a live website, navigate to the information it needs, and carry a task through multiple steps.
What mattered to us was less the ranking itself and more what the benchmark actually tests. Answering a question is one thing. Navigating a changing web environment, planning the steps, and carrying the task through to completion is another.
The ACT-2 result gave us an external measure of that ability. Execution is where real work starts.
In Enterprise Environments, Execution Was Not Enough
We knew the agent could do the work. Once it entered a real enterprise environment, the next problem came into view.
Early in the quarter, we treated this as a question of where the data sits. If a customer cannot send its data out, bring the AI in.
Conditions in the field turned out to be far more complex. Who can access what, which environment the software can be installed in, how it gets deployed, and how a configuration validated in development moves into production all had to be defined. What each customer could support also varied. In some environments, the deployment approach we considered standard could not be used at all.
The more AI can do, the more a company has to decide for itself.
This is why we call AgentOS an operating system. Where the model runs is only one question among many. The scope of what companies wanted to control directly was broader than we expected, and that demand was not limited to Korea.
What Q3 Points To
Can an agent finish real work? Can a company control that execution inside its own environment? The questions we encountered in Q3 came in that order.
Answering well and finishing the job are different things. Enhans is built for the second.
Finishing the job goes beyond the product. The hard part is rarely the idea. It is carrying it through until it becomes a result. The Builder, which we published this quarter, is a record of the people who do exactly that: the ones who find problems nobody has named, connect scattered pieces, and stay with the work until the technology actually works. Season 2 arrives in Q4.
You Decide. Our Agents Act. AgentOS gives companies control over how their decisions become execution.
Q3 showed us, in the field, that the scope of what we call an operating system is much broader than we had thought.
in solving your problems with Enhans!
We'll contact you shortly!