Box has announced new capabilities to transform how organizations put agents to work across their most critical business processes. Box unveiled Box Architect for autonomously building, managing, and optimizing agentic workflows, redefining how organizations deploy AI on their most critical processes.

With Architect, teams will be able to leverage frontier AI to create new workflows and apps, configure agents, connect to Box tools, and manage workflows at scale – all built on their trusted content and existing permissions in Box.

In tandem, Box is also building new threat protection capabilities to help detect and contain agent attacks such as bulk exfiltration and account takeover. These increased security measures will complement Box’s agent security controls, which are now available and help govern what agents can access and do with enterprise content.

“The real opportunity with AI is not just helping individuals work faster, but rethinking how entire business processes get done,” said Aaron Levie, CEO of Box.

“Until now, putting AI to work across those processes has often required specialized expertise and significant effort, but breakthroughs like OpenAI’s Astra and Anthropic’s Fable are changing what’s possible. That’s why we’re building Box Architect – to help teams turn these advances in frontier AI into workflows that bring together agents, people, and content. At the same time, as agents become more capable and the threat landscape grows more complex, organizations need to advance security alongside innovation. Box is helping them do both by protecting enterprise content and giving organizations the confidence to explore what’s possible with AI.”

Introducing Box Architect

Organizations of every size and industry are seeing the potential for AI to transform their business processes, but deploying agentic workflows at scale can be complex, costly, and requires careful oversight. With Box Architect, teams will be able to deploy agents to transform complex business processes into collaborative and repeatable workflows, while keeping people involved at key decision points.

Box Architect will give teams the ability to deploy autonomous agents across the files and tools they already use in Box to create a unified process, drawing on capabilities such as Box Extract, Box Automate, Box Apps, Box Forms, and more.

Customers will also have the flexibility to choose from leading models, including Anthropic, OpenAI, Google, NVIDIA, IBM, and more, selecting the right model for each workflow step – from lower-cost models for simpler tasks to more advanced models for complex reasoning. Throughout these workflows, the content will remain in Box, governed by existing permissions and policies, with security controls that help organizations manage agent access and activity.

With Box Architect, users will be able to:

  • Design workflows by describing a process or goal in plain language. Box Architect uses content and procedures already in Box to define stages, decisions, and handoffs, with every step governed by existing permissions, classifications, and retention policies;
  • Test workflows before they go live using real content and edge cases. Teams can preview agent outputs and routing, add human review and approvals at key decisions, and choose the right model for each step to move from pilot to production with confidence;
  • Deploy complete workflows with the Box Apps, agents, automations, forms, and document templates each process needs. Box Architect monitors results and recommends improvements, while administrators publish and standardize workflows across the enterprise.

Box Architect will bring the context held in Box into enterprise critical workflows across industries:

  • Law firms will be able to ask Box Architect to build a contract review workflow that learns the company’s approved playbook from past agreements, compares incoming contracts against it, flags exceptions, and routes high-risk clauses to attorneys for final review;
  • Banking organizations will be able to turn client onboarding into a single connected workflow that collects applications through Forms, verifies documents with Extract, generates and signs agreements with Doc Gen and Sign, and starts account setup with Automate;
  • Government agencies will be able to use Box Architect to build a permitting workflow that checks applications for completeness, extracts key details from plans and supporting documents, requests missing items, and routes applications for reviewer approval, all under existing records retention policies;
  • Insurance companies will be able to orchestrate claims processing by collecting information through Forms, extracting details from supporting documents, generating claim summaries, and routing settlement documents for signature in a connected workflow;
  • HR agencies will be able to build an agentic workflow for employee onboarding that reviews packets, extracts key information, requests missing details, routes exceptions for review, and moves approved files into the next step.

“Enterprise AI maturity is moving beyond isolated conversational interfaces into autonomous process orchestration grounded in enterprise content,” said Alan Pelz-Sharpe, Founder, Deep Analysis. “While legacy approaches require brittle point-to-point configurations for extraction and execution, an integrated agentic design layer dynamically orchestrates content understanding, process steps, and analytical visualization within a single governed ecosystem. This enables organizations to drive measurable business transformation rather than fragmented tactical automation.”

Protecting enterprises in the age of agentic threats

As employees, external collaborators, applications, and agents access and act on enterprise content, organizations face an expanding attack surface that demands stronger security and governance. Box extends protection at the content layer across these interactions, helping organizations defend against threats, enforce content access controls, and apply guardrails to agents wherever they work.

As agentic threats evolve, Box is building new innovations in threat protection to help organizations detect and respond to attacks targeting enterprise content, including bulk exfiltration, session hijacking, account takeover, and mass deletion. These capabilities are designed to empower enterprises to:

  • Detect anomalous downloads faster and automatically restrict them, helping stop bulk exfiltration sooner;
  • Strengthen protection against session hijacking and account takeover with session binding and device-bound session credentials, helping prevent stolen tokens or sessions from being replayed on another device;
  • Identify suspicious sharing and mass deletion, which can precede data exposure, exfiltration, or destruction;
  • Apply automatic, severity-based responses, from restricting downloads to requiring step-up authentication or terminating a session.

Alongside these new threat detection capabilities, Box is also delivering enhanced security controls to help protect enterprise content from unauthorized agent action. As agents gain the ability to access and act on sensitive data, these controls add protections against activity that could exfiltrate, expose, or delete content. Now available, these controls allow users to:

  • Control what agents can access with classification-based policies that govern agents and the classified content available to them, helping prevent agents from reaching sensitive information they don’t need;
  • Constrain agent actions with MCP Guardrails that let Box admins set boundaries on the tools agents can use, including creating, moving, and sharing content externally, helping prevent unauthorized or risky changes;
  • Detect potentially harmful activity with threshold-based policies that alert teams when agent actions exceed customer-defined limits, helping teams intervene before malicious activity causes widespread damage.