Melius Wants Creative Teams to Stop Operating Tools

--- title: Melius Wants Creative Teams to Stop Operating Tools slug: melius-wants-creative-teams-to-stop-operating-tools meta_title: Melius Bets the Creative Bottleneck Is Tools meta_description: Melius raised $25 million for a shared creative operating system where agent teams handle image, video and audio production work. category: ai tags: Melius,creative AI,agent workflows,funding,General Catalyst,CRV,Ampersand,OneByZero,creative operations,agents lab ---

Melius raised $25 million this week, split between a $5 million seed led by General Catalyst and a $20 million Series A led by CRV. The New York company describes itself as the first agents lab for creative work, and its product is a shared creative operating system where teams of AI agents handle production.

The cofounders built the company on a specific complaint. Creative teams, in CEO Joowon Kim's framing, do not have an idea-quantity problem. They have a resource problem. Generative AI made content cheap to produce, but using it still means prompt-engineering your way through a dozen disconnected tools.

What the bet actually is

Melius makes no claim about a better model. It claims that the bottleneck sits between the idea and the finished asset, and that agents can own that space.

The company came out of stealth in July 2026 and says it crossed a seven-figure run rate within three months, with thousands of teams building assets on it daily. General Catalyst's Yuri Sagalov framed the thesis as consolidation: most generative tools for creative teams are single-purpose, a prompt box here and an editor there, and the advantage compounds when agents can take a workflow from first draft to finished asset.

The founding team's background is worth noting. CEO Joowon Kim and CTO Young Kim worked together at Ramp, where Kim was director of engineering and a co-founder of Venue, which Ramp acquired in 2023. COO Arnav Ramu came from software investing at Hg and investment banking at Guggenheim. That is a team built for operational discipline rather than model research, which fits a product whose claim is workflow ownership.

The name of the category Melius is claiming, an agents lab for creative work, points at what it thinks will be scarce. Not images, which are now abundant and cheap. The judgment about which images are right.

The same week told a consistent story

Three other raises landed within days and point the same direction.

Ampersand closed a $15 million Series A led by Bessemer to build read-and-write plumbing between AI applications and systems of record like Salesforce and NetSuite. Its pitch against traditional integration platforms is depth: enterprise systems are heavily customized, and a shallow connector that ignores those customizations fails in practice. The company also launched Andi, an AI integration agent that handles per-customer implementation work.

The framing is precise. Ampersand leaves the systems of record where they are. It gives agents the ability to read and write inside them, which is a smaller and more defensible claim. Its total funding now sits around $20 million, and Yelp and Mana Ventures joined as new investors.

OneByZero raised a $20 million Series A led by Jungle Ventures, its first external funding since founding in 2023. The Singapore company builds enterprise AI deployment and governance infrastructure and sells it the opposite way from self-serve agent builders. Its forward-deployed engineers embed agents inside customer systems and own accountability for what goes live. The company says some customers have automated more than 90 percent of customer-facing interactions and cut certain data migration delivery times by half, operating across nine markets.

The detail that matters in that pitch is governance. NEO lets enterprises define what AI is allowed to do, where human judgment stays in the loop, and every agent action leaves an audit trail. That is what gets approval to operate in regulated industries rather than stalling at proof of concept.

Zeroset closed a $5.2 million pre-seed co-led by Gradient Ventures and 2048 Ventures for Nebula, a state layer that lets long-horizon agents query the current state of a workflow instead of rebuilding context every turn.

What connects them

None of these companies is selling a model. Each is selling the layer that makes models usable inside an existing process, and each is explicit that capability is no longer the constraint.

That is a shift in where founders expect value to accumulate. During the model release cycle of the past two years, every launch reset the capability ceiling. The raises this week suggest investors think the ceiling is high enough already, and the money is moving to whoever removes the friction between a capable model and a finished piece of work.

The vocabulary has shifted with it. Ampersand sells plumbing. OneByZero sells accountability. Zeroset sells state. Melius sells an operating system. Not one of those words describes intelligence, and all four describe what is missing when a capable model meets a real organization.

The risk in the pattern

There is a version of this bet that fails without anyone noticing. If agents become good enough at using tools, some of these layers get absorbed into the model or the platform. An integration layer that exists because agents could not write reliably to Salesforce becomes unnecessary once they can.

The counter-argument is that customization is the durable part. No general model will know how a specific company routes approvals, and no platform vendor will implement every customer's exceptions. That is the gap these companies are betting stays open.

The customer profile differs across the four, which spreads the risk. Melius sells to creative teams, Ampersand to developers building on enterprise systems, OneByZero to large regulated enterprises, Zeroset to platform and ML engineers. Different buyers, different adoption cycles, same underlying wager about where the friction lives.

For Melius specifically, the test is whether a creative operating system can hold up across teams with different tastes and different brands. Agents that handle production work well are valuable. Agents that handle it consistently, in a way a creative director can sign off on, are a different product. Less than three months of revenue does not answer that question yet.

There is also a quality ceiling that no amount of workflow automation raises. A creative operating system that produces on-brand assets efficiently still depends on someone deciding what the brand should be. The founding claim, that creative teams have a resource problem rather than an idea problem, is true for production volume and less obviously true for the work that actually lands.

What the batch of raises does establish is a consensus about where the work is. After two years of measuring progress by model capability, the industry is funding the layers that decide whether that capability reaches a finished deliverable.