Question 1

What is Agent/works™ by Thoughtworks, and how does it fit with Thoughtworks services?

Accepted Answer

Agent/works™ by Thoughtworks is a control plane that governs, observes, and cost-manages AI agents across any models, frameworks, clouds, or other environments. It is a governed agentic runtime that Thoughtworks uses within enterprise AI transformation engagements to help organizations run AI agents safely, manage AI costs and operate across models, frameworks and clouds. Our teams configure, integrate and optimize it around your environment, workflows and business outcomes.

Question 2

What enterprise problems does Agent/works™ help solve?

Accepted Answer

Agent/works™ helps address the gap between a promising AI pilot and dependable enterprise use. Thoughtworks uses it to help organizations govern agent access and behavior, improve AI economics, bring agents into production and measure whether workflows are delivering business value.

Question 3

How can I see what my AI agents are doing and what they can access

Accepted Answer

Today, Business and IT leaders are unable to confidently answer questions like - how many agents are running in my estate today, what stage of scaling they are at, what systems and data they can access, etc. Agent/works™ registers agents, models, tools and data sources in a governed inventory; checks policies before agents act; and applies scoped identities, time-limited permissions and usage controls. It also captures execution records to support monitoring, investigation and continuous improvement. Thoughtworks configures and tests these controls within your operating model.

Question 4

How does Agent/works™ help reduce AI and token costs?

Accepted Answer

Agent/works™ helps route work to an appropriate model and manage context, caching, token budgets and usage limits. Thoughtworks establishes a cost baseline, tunes the configuration and measures savings against quality, performance and business requirements, so cost reduction is tied to the value the workflow must deliver.

Question 5

Can Agent/works™ help move from pilot to production-grade agents?

Accepted Answer

Yes. Anyone can create an agent in hours. Almost nothing built that way survives contact with real users, real data, and real audits. The gap between a working pilot and a governed production system is where most enterprise AI programs stall. Agent/works™ is designed to help move agents from experimentation into dependable enterprise use through defined agent identity, time-limited credentials, policy enforcement, scoped access (that can only narrow, never expand), execution traceability records, isolation and usage controls. Isolation is configurable per agent, so high-risk work runs in a tighter sandbox without slowing down the rest of the estate. Thoughtworks can also support production readiness through integration, testing, human review, recovery design and operator enablement.

Question 6

Where does Agent/works™ run, and does it lock us into a model or cloud provider?

Accepted Answer

Agent/works™ is deployed in your environment, including cloud, bare-metal or air-gapped environments. It works with any model, cloud or framework, so your governance layer does not have to move when your providers or workloads change. Your data does not need to leave your environment or connect to Thoughtworks systems.

Question 7

How does Agent/works™ relate to AI/works™?

Accepted Answer

AI/works™ is Thoughtworks' agentic development platform for building and modernizing software. Agent/works™ is governed agentic runtime that Thoughtworks uses within enterprise AI transformation engagements to govern and run AI agents safely and cost effectively. We select platforms and integrations according to the workflow, environment and controls in scope.

Question 8

Does Agent/works™replace data governance or business logic systems?

Accepted Answer

Agent/works™ is a governed runtime, not a replacement for data governance, business logic or semantic systems. It works alongside the tools that own your business context. Thoughtworks can integrate the data products, rules and interfaces an AI-enabled workflow needs.

Question 9

What models, clouds and frameworks does Agent/works™ support?

Accepted Answer

Agent/works™ supports any model, any cloud, including air-gapped and bare-metal environments, and any framework. It can support frontier and open-weight models, as well as vendor-hosted and self-hosted infrastructure, through a consistent governance layer. Our teams configure, integrate and optimize it around your environment, workflows and business outcomes..

Question 10

How are token cost savings measured?

Accepted Answer

Agent/works™ calculated savings are verifiable. We agree on the baseline, comparable work, quality and business requirements while accounting for usage changes. For suitable engagements, the agreed measurement approach can support an outcome-based or gainshare model.

Question 11

How can I tell whether we are getting value from our AI token spend?

Accepted Answer

Token costs are rising, but many organizations cannot see which agents, teams or use cases are driving that spend, or whether it is delivering value. Agent/works™ routes requests to the right model for the task and uses live testing to verify savings rather than relying on projections. Cost, behavior and safety can be measured by agent, team and use case.

Question 12

Can Agent/works™ help me manage my token supply chain?

Accepted Answer

Yes. Agent/works™ registers frontier and open-weight models, along with vendor-hosted and self-hosted infrastructure, through the same interface. This gives teams a consistent way to govern a mixed model estate, while changing providers becomes a catalog update rather than a governance rebuild.

Question 13

What happens to an agent's access when its owner leaves the business?

Accepted Answer

Each agent has its own identity, separate from the person who created it. During offboarding, ownership is reviewed, reassigned to an accountable owner or retired, and access is revisited against current policy. Scoped and time-limited permissions reduce the risk of standing credentials outliving their purpose.

Question 14

Can Thoughtworks improve agents we already run?

Accepted Answer

Yes. We can integrate supported workloads with Agent/works™ in your environment, then configure policies, cost controls and tracing. Depending on the outcome, the engagement may also include workflow redesign, data and business-context integration, quality testing or operator enablement as part of larger transformation delivery.

Question 15

How do I make my agents handle business context and semantic understanding?

Accepted Answer

Access to a system is not the same as understanding what its data and processes mean. Agent/works™ is not a data governance or business logic platform. It is designed to work alongside the systems that own your semantic layer, while Thoughtworks can help integrate the business rules, context and data products your AI workflow needs.

Question 16

Does the engagement include the data and business context AI needs?

Accepted Answer

Yes. We integrate the business rules, context and data products the workflow needs within your environment, so agents can use relevant information and take the right actions. Wider data modernization is scoped when the outcome requires it.

Question 17

What does our team take forward if we use Agent/works™ on one program to start with and decide to manage it on our own?

Accepted Answer

Your team takes forward the agreed capability, configurations, evaluations, operating playbooks and knowledge needed to run it. We enable your team and make code rights and support arrangements explicit, so you can continue independently or work with Thoughtworks on ongoing optimization and operation.

Question 18

Who should be involved in our organization?

Accepted Answer

Your workload owner and a technical counterpart help us integrate systems and check results. Security and operations stakeholders join decisions about access, controls and day-to-day running.

Question 19

How do I get started?

Accepted Answer

For supporting existing workloads, we can focus on deployment, integration, configuration and checking results. New workflows may also require design, data or engineering work. We match the services to your starting point and desired outcome.

Question 20

What is the commercial model?

Accepted Answer

Commercial terms depend on deployment model and scale. For suitable engagements, the agreed measurement approach can support an outcome-based or gainshare model. Connect with our team to discuss further.

Run AI agents.
Control what they do.
Keep costs in check.

Agent sprawl. Hard to govern and expensive to scale

Behind those pressures are three questions:

Agent/works™ is how you solve all three

One governance layer across your stack

An audit trail from the start

Verified lower token spend

Model choice without starting over

Consistent governance as your stack evolves

Agent/works™ runs wherever you do

One governance layer across your stack

An audit trail from the start

Verified lower token spend

Model choice without starting over

Consistent governance as your stack evolves

Agent/works™ runs wherever you do

Under the hood

The Agentic Enterprise you are building towards

Agent/works™ plays well with others

Agent/works™ resources and more

Inside Agent/works™: The technical guide

Speak to Catherine, your guide to Agent/works™

Explore the Thoughtworks AI Labs research behind Agent/works™

Views on agentic tokenomics, value and governance

Rewire your data for agents

Invest in your AI Factory Infrastructure

Build and scale your AI agent platform

Three things you can do today:

Frequently asked questions

Run AI agents.Control what they do.Keep costs in check.

Agent sprawl. Hard to govern and expensive to scale

Behind those pressures are three questions:

One governance layer across your stack

An audit trail from the start

Verified lower token spend

Model choice without starting over

Consistent governance as your stack evolves

Agent/works™ runs wherever you do

One governance layer across your stack

Under the hood

The Agentic Enterprise you are building towards

Agent/works™ plays well with others

Inside Agent/works™: The technical guide

Speak to Catherine, your guide to Agent/works™

Explore the Thoughtworks AI Labs research behind Agent/works™

Views on agentic tokenomics, value and governance

Rewire your data for agents

Invest in your AI Factory Infrastructure

Build and scale your AI agent platform

Three things you can do today:

Frequently asked questions

Run AI agents.
Control what they do.
Keep costs in check.