Orchestra raises $3.3m to put data teams in charge of AI agents

On 1 September, Orchestra, a London-based data orchestration company, announced a $3.3m (£2.4m) seed round led by Differential Ventures, taking its total funding to $4.6m (£3.4m). Moonfire, which led a $1.3m (£1m) pre-seed round, returned for the seed alongside Breakers, Tokyo Black and Contour Ventures. The company used the same announcement, which it called its formal launch, to set out Orchestra Runtime, an environment for running AI agents next to the data pipelines it already schedules, and to describe itself as an “agentic control plane” for enterprise data and AI.

Hugo Lu, co-founder and chief executive, wrote on the company blog that he was “super excited to finally formally announce our fundraise”. As reported by Dealroom, the round closed about a year ago, and the company says usage has grown more than tenfold since. The announcement puts the same figure as platform usage growing “more than 10x” in a year.

Lu incorporated Orchestra Technologies Limited on 30 January 2023 and is its only person with significant control, holding more than half of the shares, according to Companies House. He started in banking at Rothschild in London, then joined the strategy team at JUUL, where a lack of data pushed him to learn data engineering, he told the OxIgnite podcast in August 2023. He went on to lead the data team at Codat, a London fintech. The idea for Orchestra came from reviewing a code repository far more elaborate than its job of triggering scheduled tasks required, he told the podcast, where he also said he had started the company as a solo founder. William Davies joined the board as a director in August 2023; Dealroom names him as co-founder and says he previously led product engineering at a video AI infrastructure startup.

Orchestra spent its first years selling data teams a scheduler and monitor for their pipelines. The seed announcement stretches that pitch to cover the AI agents those teams are now being asked to build and run. That makes the round less a bet on pipeline scheduling than a bet on which team inside a company ends up answerable for its AI agents.

From pipeline scheduler to agent supervisor

Orchestra Runtime runs each agent in isolation, with access limited to the resources its task needs, according to the company. It lists three uses: creating pipelines, investigating production incidents and automating the triage of pipeline failures. Runtime draws on a Context Layer that organizes information about a company’s systems and workflows, so an agent gets what it needs without access to the wider environment. Orchestra says the design works with any model and uses fewer tokens.

In his blog post Lu argued that teams are building workflows in ungoverned tools such as n8n and Cursor, creating what he called an “AI Swamp”, with agents running “without supervision”. In the company’s announcement he said: “AI is creating enormous demand across the enterprise, but data and now AI teams are still managing infrastructure that was never designed to support it.” Orchestra, he said, “gives them the control they need to move beyond being a service desk”.

The customer moving its whole estate

“We’re moving our entire dbt estate from Airflow to Orchestra,” said Mike Holmes, global head of data engineering at Octopus Energy. “We’re excited about running agents alongside our data, giving them direct & secure access to the context they need to take action.”

When Orchestra joined the Databricks partner program on 11 August, it said Octopus Energy uses Orchestra with Databricks to run a multi-tenant analytics platform of more than 5,000 dbt models across more than 10 business units, in place of a custom fork of Airflow. Experian is the other customer the company names in its announcement. In July 2024, when it disclosed the earlier £1m round led by Moonfire, Orchestra listed Graniterock Construction, Powell Industries and the San Francisco Giants among its users.

Orchestra says customers have cut pipeline runtime costs by up to 80% and pipeline development time by up to 95%, and that the platform automates hundreds of failure investigations a day. It does not say which customers achieved those figures.

A seed round sized for pipelines

Orchestra says it will spend the money on engineering and go-to-market hiring in London, on product development and on growth across the United States and Europe. Nick Adams, co-founder and managing partner at Differential Ventures, said that “one of the most significant barriers to powerful AI systems deployed in production is reliable data pipeline management”.

The money is going on hiring in London and growth in two regions, and the agent product it sets out has no named customer running agents on it yet. Holmes describes running agents at Octopus Energy as something the company is excited about, while the dbt migration is the work under way.

Orchestra has a founder-controlled business, a pipeline product that a large energy supplier is moving its whole dbt estate onto, and usage that it says has grown tenfold in a year. The agent layer is the new claim, and for now it rests on the data teams that already trust Orchestra with their pipelines deciding to trust it with their agents too.