Thoughtworks has launched Agent/works, a platform for governing and operating enterprise AI brokers throughout any cloud. It supplies a single management airplane and a ruled runtime for AI brokers.
The launch comes as corporations transfer from experimenting with AI instruments to deploying autonomous techniques that may entry knowledge, use software program instruments, and perform workflows with much less direct human enter. That shift is creating new stress on safety, compliance, and finance groups as AI techniques unfold throughout organisations.
Agent/works is designed to present corporations a single registry for brokers deployed throughout their know-how property, with oversight of governance, efficiency, and AI spending. It’s meant to sit down beneath customized agent-based functions constructed by enterprise product groups, reasonably than function as a standalone software.
Thoughtworks framed the product round what it calls agent sprawl: organisations shedding visibility into what number of AI brokers are in use, what knowledge they’ll entry, and what sources they eat. It additionally linked the difficulty to the unfold of AI-assisted software program improvement, which is growing the amount of code and automatic workflows getting into manufacturing techniques.
Trade analysis cited by Thoughtworks factors to each adoption and threat. In response to Sonar’s 2026 State of Code Developer Survey, builders reported that 42% of dedicated code is now AI-generated or AI-assisted. Analysis from AppSec Santa discovered that 25% of AI-generated code samples contained essential safety vulnerabilities.
Governance focus
Agent/works consists of controls aimed toward managing how AI brokers behave earlier than and through execution. The system checks workflow paths earlier than an agent runs to substantiate that not less than one compliant route exists from finish to finish.
It additionally applies permissions tailor-made to brokers reasonably than human customers. These permissions are capability-based, scope-bound, and time-limited, and may slim robotically when an agent interacts with extra delicate techniques or knowledge.
The runtime is constructed to assist various kinds of brokers, together with autonomous workflow brokers and interactive coding brokers. Insurance policies can adapt throughout execution, whereas the platform can join with completely different fashions, instruments, cloud providers, and third-party brokers by means of normal interfaces.
A central registry sits on the core of the platform, offering visibility into brokers, fashions, instruments, and insurance policies, alongside evaluations, utilization analytics, and price controls.
Shayan Mohanty, Chief Knowledge and AI Officer at Thoughtworks, linked the launch to the operational calls for of large-scale autonomous techniques. “When anybody can generate software program with a textual content immediate, AI governance is not a guidelines you bolt on after the very fact. It is a foundational requirement for working autonomous techniques at scale,” Mohanty stated.
He additionally tied the difficulty to price administration as AI brokers develop into extra frequent inside enterprises. “Each AI-powered workflow now carries an working price. The problem for enterprises is now not simply the way to construct brokers, however the way to govern the sources they eat. With out runtime controls, prices can scale as rapidly because the brokers themselves. Agent/works offers organisations visibility and governance throughout their agent ecosystem, serving to them management spend, handle threat, and scale AI with confidence. Governance accomplished this manner stops being a brake and turns into the engine that lets groups transfer quick, safely, and at scale,” he stated.
Databricks tie-up
Thoughtworks additionally highlighted its work with Databricks on enterprise AI governance and agentic techniques. The connection centres on serving to corporations apply governance fashions used for enterprise knowledge to agent workflows.
David Nasi, Director of Product Administration, AI and Agentic Platform at Databricks, stated that method is gaining traction amongst early adopters. “The organisations transferring quickest with brokers are extending the identical governance fashions already used for enterprise knowledge throughout agent workflows themselves,” Nasi stated.
He added: “This shift issues as a result of it turns governance from a bottleneck into an enabler, making it a lot simpler to scale autonomous techniques safely throughout all the enterprise.”
Thoughtworks stated its agentic improvement providing, AI/works, already runs on Agent/works, which it introduced as proof that the platform is being utilized in manufacturing. It positioned Agent/works as a foundational software program layer for enterprises making an attempt to handle the rising complexity of autonomous AI techniques whereas sustaining oversight of threat, compliance, and spending.