WORK-SELF, the London-based enterprise AI company, has expanded its Maya Enterprise platform across more than 12,000 enterprise applications, adding what it calls Human Context Infrastructure to agentic workflows. The expansion matters to the sector because it targets the human checking, correction and coordination that often follows an automated task.

WORK-SELF is a private, pre-seed company led by co-founder and CEO Wolf Magdelinic, formerly CEO of Overbond AI, and lists seven employees. The company says the platform supplies AI agents with organisational information such as who owns a task, who can approve decisions, when human review is needed and when work should be escalated.

The company reports up to 40% less employee rework in enterprise engagements, and says that for clients with more than 1,000 employees better coordination between people and agents can mean more than £5 million in annual savings and 500,000 skilled hours returned to the workforce. It also reports a further 20% improvement in workflow efficiency from semantic optimisation and a 60% improvement in cost per unit of delivered work, figures that are the company’s own measurements.

International Data Corporation projects that active AI agents will rise from 28.6 million in 2025 to 2.2 billion by 2030, the scale of adoption the company cites.

Partners named in the announcement include CockroachDB, which provides distributed data infrastructure for the knowledge graph, INT Global, which connects governed information through its OneSpace data platform, Takara.ai, which works on semantic optimisation, and UAIN Solutions, a UiPath implementation partner.

Maya works through a low-code interoperability layer without replacing existing systems, connecting to applications, databases, data lakes, large language models, legacy systems and on-premise infrastructure, and passes context to approved agents over the open Model Context Protocol.

Wolf Magdelinic, chief executive officer and founder of WORK-SELF, said: “If an employee still has to check, correct and coordinate the work around the agent, productivity is leaking back out of the workflow.”

Magdelinic added: “Connectivity is not coordination.”

The structural driver is the growth in deployed AI agents, which shifts the productivity question from agent speed and cost to the human coordination that surrounds each automated task.

For the sector, the expansion shows vendors competing on the human layer of agentic automation, with the productivity case resting on reduced rework rather than agent speed alone.

Source: itbrief.co.uk / crunchbase.com / bebee.com / work-self.com