WORK-SELF Expands Maya Enterprise Across the Enterprise Technology Stack

26 Sept

Maya Enterprise Brings Human Context to Agentic Workflows Across 12,000+ Applications

New vertical technology integration combines enterprise connectivity, distributed data infrastructure, AI engineering and platform implementation to help AI agents work with employees across the systems enterprises already run.

Enterprises are deploying AI agents faster than they can redesign the work around them. An agent can retrieve information, update an application, complete a task or trigger a workflow. What it does not inherently understand is which employee owns the outcome, what requires human judgement, when an issue should escalate or which operating rules apply.

That gap is where the expected return from agentic AI can begin to disappear. Work that appears automated can still create rereads, ownership corrections, review loops and escalations that land back on employees.

Today, WORK-SELF announced the vertical technology integration of Maya Enterprise across its technology ecosystem. The integration connects Maya from the data layer through to the applications and platforms where agents and employees do the work, combining enterprise connectivity, distributed data infrastructure, AI engineering and implementation capability around a single human context layer.

Working with Takara.ai, INT Global, UAIN Solutions and CockroachDB, WORK-SELF can now bring Maya into the existing enterprise technology stack rather than requiring organisations to replace it. An embedded enterprise integration and orchestration capability extends Maya across 12,000+ applications, from modern SaaS platforms to legacy mainframes, custom applications, databases, data lakes, LLMs and on-premise infrastructure.

Maya adds what those systems do not hold on their own: the human operating context around the work. It gives approved agents governed information about ownership, decision rights, review requirements, escalation boundaries and deadlines at the point where the agent is expected to act.

“Enterprises have already invested heavily in applications, data infrastructure, automation platforms and AI agents. The problem is not another shortage of technology,” said Wolf Magdelinic, CEO and Co-Founder of WORK-SELF. “The gap appears when an agent starts doing work that previously belonged to a person. It needs to understand who owns the outcome, what it is allowed to do, where human judgement is required and when it needs to stop and escalate.”

“Maya provides that human context. Our technology ecosystem now gives Maya the reach to carry that context across the systems enterprises already run.”

One Human Context Layer Across the Enterprise Stack

Most enterprise agent programmes are assembled from several separate layers. Data sits across HR, CRM, ERP, workflow and knowledge systems. Integration technology connects those systems. Automation and agent platforms execute tasks, while implementation partners configure those technologies around individual business processes.

Each layer performs an important function, but connecting the technology does not by itself define how an agent and an employee should share the work. An application can determine whether a user has technical access without understanding whether an agent should make a decision, whether a person needs to review it, who remains accountable for the outcome or when the workflow should escalate.

Maya Enterprise fills that gap. The Work Contract defines ownership between the employee, agent and manager for each task. The Context Capsule carries the minimum necessary task, role and policy context to an approved agent. The Review Map establishes what requires human judgement, the Autonomy Dial defines how far an agent may act independently, and the Digital Andon provides a mechanism for a person or agent to stop the workflow when risk appears.

The systems enterprises already use remain the systems of record and existing access controls stay in place. Maya operates across those systems, serving governed human and workflow context to approved agents at the point of execution.

The embedded integration and orchestration layer expands that model across more than 12,000 applications and supports multi-step workflows across connected enterprise systems. Governed MCP capabilities allow approved tools and actions to be exposed to compatible agents, while identity-aware execution allows configured actions to inherit the authenticated user’s identity and permissions. Sensitive actions can be routed to a designated human approver before execution, and governed actions can be recorded with the identity behind them to support traceability.

This creates a clear separation of responsibilities. The underlying enterprise technology connects systems and executes actions. Maya determines the human operating context around those actions: who owns the work, what the agent may do, what requires review and when it must escalate.

A Technology Ecosystem Built for Enterprise Deployment

The WORK-SELF technology ecosystem has been designed so that the layers surrounding Maya are supported by specialist infrastructure and implementation capability.

CockroachDB provides the distributed data infrastructure supporting Maya’s human context graph, context exchange records and audit trail. Its technology is designed as a system of record for agentic memory in production AI and supports capabilities including geo-partitioning for data residency. WORK-SELF has validated Maya’s data layer on CockroachDB without code changes, with the same schema running, the full context corpus loading and the audit ledger writing. This provides Maya with a distributed data foundation designed for enterprise deployments spanning teams, systems and geographies.

INT Global provides enterprise systems integration and governed data capability around Maya deployments. Its capabilities span technology, data, cloud and security, supported by a team of more than 1,000 professionals across the United Kingdom, United States, Canada, India, Poland and Singapore. WORK-SELF and INT Global are also pairing INT’s enterprise capabilities with Maya as the human context layer and developing Maya Work Contracts for operational service workflows, beginning with property repair operations.

Takara.ai provides AI engineering, evaluation and delivery capability around Maya. The company specialises in production-grade enterprise AI systems and in moving AI from proof of concept into production. Takara.ai has also worked directly with WORK-SELF to design and run structured Maya evaluations, allowing the effect of structured human context on agent behaviour to be tested rather than assumed.

UAIN Solutions, an official UiPath implementation partner, provides implementation capability across the platforms where enterprises are already deploying agents. Through UAIN and its specialist implementation partners, WORK-SELF can build and implement agentic workflows across Microsoft, ServiceNow, Salesforce and Agentforce, SAP and UiPath environments, with Microsoft delivery capability led by Microsoft AI MVPs. UAIN supports the technical implementation of the agentic workflow, while WORK-SELF provides the human context and coordination layer between the agent and the employee.

Together, these capabilities give WORK-SELF a vertically integrated route from the data that holds Maya’s context through to the applications and agent platforms where work is executed. The objective is not to replace the enterprise stack, but to make the existing stack work more effectively once agents begin taking on meaningful parts of employee work.

Tested With and Without Human Context

WORK-SELF and Takara.ai have also tested whether providing structured human and task context changes the way an agent performs the work.

Across 42 customer service cases covering seven task categories, the same underlying workflows were evaluated with Maya context available and without it. The success criteria were fixed before the evaluation dataset was built. With Maya context supplied, adherence to the agreed operating rules covering ownership, review route, escalation and deadline increased from 32% to 74%, with improvement recorded across all seven task categories.

The evaluation also established an important boundary. When generic human feedback data containing no structured employee context was used, no Maya advantage was measured. The difference was not simply giving the agent more information. It was giving the agent the specific human and operating context required for the work it was performing.

The finding reflects the central premise behind Maya Enterprise. An agent may have access to the application, data and tools required to complete a task and still lack the organisational context required to complete that task in the way the enterprise expects.

From Agent Deployment to Operating Value

As enterprises move from AI experimentation toward scaled agent deployment, the question is shifting from whether an agent can perform a task to whether the resulting workflow actually removes work.

An agent can produce a technically correct output quickly while still creating additional review, correction or coordination for the employee responsible for the outcome. In that situation, automation has created capacity inside one step of the workflow without necessarily creating operating value across the workflow as a whole.

WORK-SELF is focused on that gap between the agent completing an action and the organisation accepting the resulting work. Maya makes ownership, human judgement, review requirements and escalation boundaries explicit so that the agent can operate within the same organisational context that would otherwise need to be reconstructed manually by an employee.

“Connectivity is essential, but connectivity is not the same as coordination,” Magdelinic said. “You can connect an agent to thousands of applications and still leave an employee manually working out whether its output is right, who owns the next decision or what happens when the workflow moves outside the expected path.”

“The combination is what matters. Our technology ecosystem gives Maya reach across enterprise data, systems and agent platforms. Maya adds the human context that determines how the work should move between the agent and the employee. That is how enterprises move from deploying agents to redesigning work around them.”

Available for Enterprise Deployment

Maya Enterprise is available for pilot deployment with organisations running agentic workflows across Microsoft, ServiceNow, UiPath, Salesforce, SAP and other connected enterprise systems.

Enterprise pilots begin with a defined workflow and an agreed success measure. WORK-SELF then evaluates how the workflow performs with and without Maya’s human context and coordination layer, including measures around review, correction, escalation and the capacity released back to employees.

Pilot programmes can be operational within 30 days.

To request an executive briefing or discuss an enterprise pilot, visit work-self.com or contact the WORK-SELF enterprise team directly.