Leah Launches Harness Builder to Turn Business Goals Into Governed AI Applications

Leah has launched Harness Builder, a new capability in its Maestro platform that lets business specialists describe an objective in natural language and work with AI to build a governed application around it. The launch targets legal, contracting, procurement, and finance teams that need to coordinate work across people, policies, and existing enterprise systems.

The proposed starting point is a business outcome: manage outside counsel spending, for example, rather than write a detailed technical specification for each agent and workflow. Maestro works with the user to define the scope, design the approach, and assemble the operational architecture and application employees will use.

“With Harness Builder, customers can now create and evolve these systems themselves, built around the needs of their business,” said Sarvarth Misra, Leah’s co-founder and CEO, in the announcement.

What Harness Builder Is Designed to Build

In this context, a harness is the system surrounding AI agents: the specialist capabilities, business knowledge, tools, connected applications, processes, and controls needed to carry out a defined piece of work.

Leah says Harness Builder assembles and tests those components alongside a purpose-built application. Once deployed, the resulting harness runs within Maestro, coordinating activities across connected systems, applying business rules, and escalating decisions that require human judgment.

The application is an important part of that design. Employees need a way to interact with the work, review exceptions, and manage execution. Building the operational process and its user interface together gives teams a shared place to oversee what their agents are doing.

How Maestro Connects Agents and Business Systems

The Maestro product page describes a hierarchy of orchestration. Maestro coordinates the enterprise-level objective; function-specific orchestrators handle areas such as contracts, legal, procurement, and finance; and sub-orchestrators perform business processes within those domains.

For contracting, the company lists capabilities spanning creation, review and negotiation, approval routing, signature, repositories, and obligation tracking. Its wider integration examples include Salesforce, NetSuite, Workday, SAP Ariba, DocuSign, Coupa, and ServiceNow. The website notes that capabilities and integrations are configurable to the organization.

The surrounding Leah Agentic OS combines orchestration, domain agents, workflows, and application building. It describes agents sharing context across connected records, with permissions and human-review requirements configured per agent and workflow. Records of actions and reviewer decisions provide a basis for tracing execution.

For a workflow crossing departments, that shared context is consequential. A contract term can affect a supplier relationship, an invoice, and an approval requirement. The system needs to carry those dependencies through the process so that each subsequent action uses the relevant business rules.

A Legal-Spend Example Shows the Intended Workflow

Leah illustrates Harness Builder with an outside-counsel spending system. A legal operations team would describe the objective, and the builder would identify and connect the capabilities needed to review invoices against billing guidelines, monitor budgets, identify discrepancies, and route exceptions for approval.

That is an illustrative use case in the announcement, rather than a named customer deployment. It shows the kind of coordination the product is intended to support: reviewing an invoice is one step, but completing the process also requires knowing which policy applies, whether the relevant budget is being exceeded, and who can approve an exception.

“The challenge isn’t building individual AI agents, it’s building complete systems that can reliably execute complex business processes,” said Anurag Malik, Leah’s president and CTO.

The company’s domain knowledge is intended to supply a starting structure for that work. An organization’s own policies, playbooks, and business context then shape the decisions and controls specific to its operations.

Governance Remains Part of the Application

The release describes people entering the process where expertise, judgment, or approval is required. Customers retain visibility and control over execution and can adapt their harnesses as requirements evolve.

Maestro’s website also describes ordered execution, traceable handoffs, and approval before irreversible actions. It offers a sandbox for rehearsing work before connecting it to live operations and describes single-tenant cloud deployment or operation within a customer’s network perimeter.

Those controls address a different requirement from generating an initial application. Business users need to understand what an agent is authorized to do after the application is built, how exceptions will be handled, and what evidence remains when someone reviews a decision.

From Contracting Software to Enterprise Execution

Leah’s company history traces the business to its founding in 2015 and its earlier identity as ContractPodAi. The launch announcement says the London-founded company serves more than 400 global enterprises and is backed by SoftBank and Insight Partners.

Harness Builder extends that legal and contracting background into a broader application-building proposition. Subject matter experts supply the objective and operational context; Maestro assembles the coordinated system intended to deliver it.

The practical test is whether those generated systems can preserve the policies, dependencies, approvals, and visibility that make a commercial process usable in production. Leah is positioning Harness Builder as a way for the people closest to the work to define and evolve that system themselves.