For many companies, there are barriers to fully adopting and benefiting Agent AI.
IBM Blockers bet on governing them in production rather than building AI agents.
At today’s TechXChange 2025 conference, IBM unveiled a set of features designed to bridge the gap. Project Bob is an AI-first IDE that coordinates multiple LLMs to automate application modernization. Agents for real-time agent governance. First Open Source Integration langflow Inside Watson’s OrchestrateIBM’s platform for deploying and managing AI agents. IBM’s announcement represents a three-fold strategy to address the challenges of interconnected enterprise AI.
The company claims that 6,000 internal developers within IBM use Project BOBs, with an average productivity increase of 45% and a code commit increase of 22-43%.
Project Bob is not a separate vibe coder, it’s an enterprise modernization tool
There is no shortage of AI-powered coding tools in the market today, such as tools such as Github Copilot and vibe coding tools such as replica, cursor, bolt, and Lovable.
"Project Bob takes a fundamentally different approach to tools like Github Copilot and Cursor." Bruno Aziza, IBM’s vice president of data, AI and Analytics Strategy, told VentureBeat.
Aziza said Project Bob is focused on enterprises and maintains full repository context throughout the editing session. Automate complex tasks like Java 8 into more modern Java, and framework upgrades from Struts or JSF from React, Angular, or Liberty.
The architecture is coordinated between Claude of Mankind, Mistral, Meta Llama, and recently released IBM Granite 4 model Through a data-driven model selection approach. The system routes tasks to the optimal LLM, balancing real-time accuracy, latency and cost.
"Understand the entire repository, development intent, and security standards, allowing developers to design, debug, refactor and modernize their code without breaking the flow,” he said.
Of the 6,000 early recruits within IBM, 95% used Bob to complete tasks rather than code generation. The tool integrates DevSecops practices, such as vulnerability detection and compliance checks, directly into your IDE.
"Bob goes beyond code assistance – organizing intelligence throughout the software development lifecycle, helping teams ship safer and modern software faster." He said.
Project Bob benefits from a new human partnership
Part of Project Bob is a new partnership between IBM and humanity
The two vendors have announced a partnership that will integrate the Claude model, starting with Project Bob, directly into the Watsonx portfolio. This collaboration includes what IBM describes as its first guide to deploying enterprise AI agents, beyond model integration.
Co-creation between IBM and humanity "A guide to arranging Secure Enterprise AI Agents on an MCP server," Focused on the Agent Development Lifecycle (ADLC). The ADLC framework provides a structured approach to the design, deployment and management of enterprise AI systems. MCP refers to a model context protocol, a widely adopted open standard for humankind for connecting to systems and data that requires AI assistants to use.
Easily build enterprise-grade AI agents
In addition to Project Bob, IBM has announced that it will extend Watsonx’s orchestration technology to integrate the open source Langflow Visual Agent Builder. Langflow is an open source technology led by DataStax and was acquired by IBM in May this year. Langflow integration is intended to address what Aziza calls "A prototype for production cracks."
"Today there is no seamless path from open source prototyping to reliable, compliant, and scalable enterprise-grade systems." Aziza said. "Watsonx Orchestrate converts the composition of agents like Langflow into an enterprise-grade orchestration platform Adding governance, security, scalability, compliance and operational robustness makes it production-ready for mission-critical use. ”
Aziza explained that the integration of Langflow and Watsonx orchestration brings important features on top of open source tools, including:
Agent Lifecycle Framework: Provisioning, versioning, deployment and monitoring using multi-agent coordination and role-based access.
Integrated AI Governance: Embedded Watsonx.Governance provides audit trails, accountability for agent decisions, bias and drift monitoring and policy enforcement. Langflow does not have native governance control.
Enterprise Infrastructure: SaaS or on-premises with data separation, SSO/LDAP integration, and granularity permissions. Langflow users need to manage their own infrastructure and security.
No Code and Pro Code Options:langflow is "Low code." IBM has added a visual, no-code agent builder and a pro-code agent development kit for seamless promotions from prototype to production.
Pre-built domain agents: A catalog of HR, IT and financial agents integrated with Workday, SAP, and ServiceNow.
Production Observability: Built-in dashboards, analytics and enterprises support SLAs with continuous performance monitoring.
Agent Op and Agent Workflow: From Building to Governance
IBM is introducing two new features in Watsonx Orchestrate that work in conjunction with the Langflow integration. Agent workflows for standardized agent coordination and Agentops for production governance.
Agent workflow deals with what Aziza calls "Fragile Script" problem. Today, developers are building agents using custom scripts that break when scaled across an enterprise environment. Agent workflows provide standardized, reusable flows that consistently sequence multiple agents and tools. This connects directly to the Langflow integration. LangFlow provides a visual interface for building individual agents, but the agent workflow handles orchestration layers and coordinates multiple agents and tools into a reproducible enterprise process.
Agentops then provide governance and observability to those who are running the workflow. The new built-in observability layer provides real-time monitoring and policy-based control throughout the agent’s lifecycle.
Governance gaps become concrete in enterprise scenarios. Without Agentops, HR Onboarding agents could set profits and payroll, but the team lacks visibility into whether they are properly enforcing policies until they encounter issues. Agentops allows you to monitor all your actions in real time and immediately flag and fix anomalies.
What does this mean for a company?
Technical debt is something many organizations struggle with and can represent a non-trivial barrier for organizations looking to participate in the deployment of agent AI. Project Bob’s value proposition is most obvious in organizations with important legacy Java codebases. The 45% productivity gain measured by IBM suggests meaningful acceleration in Java 8 to upgrade the framework from more modern versions of Java and Struts or JSF to Modern Architectures. However, these metrics come from IBM developers working on IBM systems. A significant unknown is whether multi-model orchestration will produce the same outcomes for customer codebases with various architectural patterns, technical debt profiles, and team skill levels.
Langflow integration addresses the true gaps in teams already using the open source agent framework. The challenge is not about building an agent with tools like Langchain, Langgraph, N8N. It adds the governance layer, lifecycle management, enterprise security controls, and observability needed for production deployment.
For businesses looking to lead AI adoption, IBM’s announcement will help strengthen the fact that governance infrastructure is currently in table stakes. Quickly build agents with existing tools. Safely scales them requires lifecycle management, observability, and policy control.
Project Bob is now available in Private Technology Preview. A wider availability is expected in the future. IBM accepts access requests through the developer portal. The agentop and agent workflow integration is now available in Watsonx Orchestrate, but its Langflow integration is expected to become generally available at the end of this month.
