What is AI agent governance? AI agent governance is the practice of tracking, securing, and cost controlling the autonomous AI agents a company runs, and on June 16, 2026, Thoughtworks made it concrete by launching Agent/works, a single control plane and governed runtime for enterprise AI agents that deploys on any cloud. For commercial real estate firms now piloting agents for underwriting, lease abstraction, and lead generation, AI agent governance is fast becoming the difference between a useful tool and an unmanaged liability. For the broader toolkit, see our guide to the best AI tools for commercial real estate investors.
Key Takeaways
- Thoughtworks launched Agent/works on June 16, 2026, a control plane that gives enterprises one registry and a governed runtime for AI agents across any cloud.
- The product targets agent sprawl, the loss of visibility into how many AI agents exist, what data they can access, and what they cost to run.
- For CRE firms running multiple deal agents, agent sprawl creates security, compliance, and budget risk that a registry and spend controls are designed to contain.
- JLL research finds organizations now pursue an average of five AI use cases at once, so sprawl is a near term reality, not a distant concern.
- Agent/works sits beneath custom agent applications as a foundational layer, and was demonstrated at the Databricks Data and AI Summit in San Francisco.
AI Agent Governance Explained
For most of 2025, enterprise AI was about experimentation. In 2026 the bill, and the risk, came due. As teams stood up autonomous agents that can read data, call tools, and execute workflows, many organizations lost track of how many agents they were running and what those agents could touch. Thoughtworks calls this agent sprawl, and Agent/works is its answer. According to the company's launch announcement, the platform gives enterprises a single registry for agents deployed across their technology estate, with oversight of governance, performance, and AI spending.
Crucially, Agent/works is built as a foundational layer rather than a standalone app. It sits beneath the custom agent applications a company's own product teams build, so governance is baked in from the start instead of bolted on later. Thoughtworks demonstrated the platform at the Databricks Data and AI Summit in San Francisco, where David Nasi, a director of product management for AI and agentic platforms at Databricks, noted that the organizations moving fastest with agents are extending the same governance models they already use for enterprise data to the agent workflows themselves. Thoughtworks says its own agentic development product, AI/works, already runs on Agent/works in production.
What Agent Sprawl Looks Like in CRE
Picture a mid sized investment shop in late 2026. The acquisitions team has an underwriting agent. Asset management built a lease abstraction agent. Marketing runs a lead generation agent. Investor relations has a reporting agent. Each was useful on its own, and each was deployed by a different team with different data access. Now ask three questions a lender or compliance officer would ask: How many agents do we run? What confidential data can each one reach? What are they costing us per month? At most firms, nobody can answer all three. That is agent sprawl, and it is the governance gap Agent/works is built to close.
The data risk is not theoretical. We have covered cases where AI agents reached unauthorized data, and national security agencies have issued agentic AI security guidance precisely because autonomous agents expand the attack surface. In CRE, the data at stake includes rent rolls, tenant PII, loan documents, and nondisclosure bound deal terms. A single ungoverned agent with broad access is a breach waiting to happen.
The Cost Angle CRE Investors Cannot Ignore
Agents are not free, and their costs are unpredictable because every task may trigger many model calls and tool invocations. Without a central registry, AI spend hides inside a dozen separate team budgets. Agent/works puts spend oversight in one place, which matters as adoption scales. JLL research finds that 92% of corporate occupiers have initiated AI programs and that firms now pursue an average of five AI use cases simultaneously, yet only 5% report achieving most of their goals. Spreading spend across many half governed agents is one reason that gap persists. For more on controlling these bills, see our analysis of AI cost management and usage based billing. CRE investors looking for hands on AI implementation support can reach out to Avi Hacker, J.D. at The AI Consulting Network.
Key Benefits of Agent Governance for CRE
- One registry: A single inventory of every agent, so leadership can finally answer how many agents run and what they do.
- Access control: Oversight of what data each agent can reach, protecting rent rolls, LP data, and deal documents.
- Spend visibility: Centralized AI cost tracking replaces budget surprises spread across teams.
- Cloud flexibility: A cloud agnostic runtime avoids locking a firm into one provider's agent stack.
- Built in, not bolted on: Governance lives in the foundation layer, so new agents inherit controls by default.
How CRE Firms Should Respond
You do not need to buy an enterprise platform tomorrow to act on the lesson behind Agent/works. The discipline is what matters. Start by inventorying every AI agent or assistant already in use across your firm, including the informal ones running inside ChatGPT, Claude, or Gemini accounts. For each, document what data it can access, who owns it, and roughly what it costs. That inventory alone often surfaces surprises. Then set a simple rule: no new agent goes live without a named owner, a defined data scope, and a cost expectation. This is the same registry logic Agent/works automates, applied manually. For personalized guidance on building this kind of governance, connect with The AI Consulting Network.
Firms standardizing on Microsoft can pair this discipline with platform native tooling; our companion piece on the Microsoft Work IQ knowledge layer covers how permission inheritance and admin center controls address the same sprawl and security concerns inside Microsoft 365. The throughline across both releases is clear: 2026 is the year enterprise AI grew up, and governance is now table stakes. If you are ready to bring order to a growing fleet of AI agents, The AI Consulting Network specializes in exactly this.
Frequently Asked Questions
Q: What is agent sprawl?
A: Agent sprawl is the loss of visibility and control that happens when an organization deploys many AI agents across different teams without a central inventory. Leadership cannot easily say how many agents exist, what data each can access, or what they cost, which creates security, compliance, and budget risk.
Q: What does Thoughtworks Agent/works do?
A: Agent/works, launched June 16, 2026, is a control plane and governed runtime that gives enterprises a single registry for their AI agents, with oversight of governance, performance, and AI spending across any cloud. It is designed as a foundational layer that custom agent applications are built on top of.
Q: Why does AI agent governance matter for commercial real estate?
A: CRE agents routinely touch confidential rent rolls, tenant PII, loan documents, and deal terms. Without governance, an ungoverned agent with broad access is a data breach and compliance risk, and uncontrolled agent spend can quietly inflate technology budgets across multiple teams.
Q: How can a small CRE firm start governing AI agents?
A: Begin with a manual inventory of every AI agent and assistant in use, noting each one's data access, owner, and cost. Then require that no new agent launches without a named owner, a defined data scope, and a cost expectation. This applies the core logic of a platform like Agent/works without buying one.