Bringing AI agents into the enterprise software development lifecycle is increasingly becoming the norm. As developers experiment with new platforms, organizations are exposed to potential security and orchestration failures. The systems that worked in the pilot may fail once agents start working with real-time data.
Legacy tech giant IBM is one of many companies trying to address that gap by introducing more structure into the way these workflows are run. Yesterday, it announced the global launch of Bob, its AI-powered software development platform designed to help write and test code throughout the development cycle, with more than 80,000 of its employees already using it after launching with just 100 internal users in the summer of 2025.
BOB has introduced a structured layer that continually pauses for human-led checkpoints, yet by using AI models to perform agentic tasks, IBM says it has saved some teams. up to 70% time "On selected tasks…equivalent to an average time savings of 10 hours per week."
Typical models supported include IBM’s own Granite series, Anthropix Cloud, some from French AI firm Mistral, and other smaller distilled models – no Alibaba Quon or other completely open source ones.
This approach reflects a shift in how enterprises want to approach AI-led development: building systems that not only build applications but also execute complex, multi-step workflows that do not rely on a single model or a single orchestration framework. It provides a structured, protected approach to automation that attempts to center humans more in the process and bridge audit gaps.
Neil Sundaresan, general manager of automation and AI at IBM, told VentureBeat in an exclusive interview that a large part of using AI for software development is being organized.
“Model capability alone is not enough,” Sundaresan said. “How you deploy it, how you structure the context, and how you keep humans in the loop determines whether AI actually delivers.”
This divide is shaping how enterprises choose AI tools, whether they prioritize flexibility and experimentation or reliability and auditability.
Different approaches to AI-led development
The growing class of open or autonomous agent systems has pushed the boundaries of what developers can do. They can now run extended or stateful workflows without any human intervention.
The rise of OpenCL showed enterprises how far experiments can go, especially when trained on local data and run in a sandbox. But it also meant choosing between easy agent and workflow creation and security.
Some companies have adopted this spirit of experimentation.
Enterprise providers like Nvidia decided to adopt systems like OpenClave by adding a fence around sandbox environments running autonomous agents using NemoClave. Kilo launches Kilo Claw, which aims to provide security to autonomous agents. OpenAI, in its updated Agent SDK, has added support for sandbox agent implementations that reflect the usage patterns of a lot of systems like OpenClaw.
Sundaresan said enterprises continue to experiment with what kind of approach they want to take to coding and agent creation. He doesn’t want to close the door to fully autonomous agents actively completing tasks, but he believes enterprises will also want to take more precautions.
“If you tell me the ultimate answer will be OpenClaw, we’ll get there,” he said. “But it’s better to open the gate slowly than to say, ‘Oops, how do I close it now?'”
Bob reflects on the thought process that highlights the growing change for enterprises.
How does Bob compare?
BOB acts as a coding platform, but unlike similar products, it aims to standardize and control the agent workflows built on it.
Tools like cursors and cloud code put the user at the beginning of the task. They are writing signals, creating series of steps, and debugging. Langgraph does the same and also allows teams to define agent flows.
The difference is not about capabilities but about control, and whether the system explores enterprise potential solutions or provides predictable execution.
In this case, the human employee starts and ends the process. If the agent is unable to complete his task or makes a mistake, this fact is controlled later.
BOB, on the other hand, essentially pre-structures the development lifecycle into role-based stages. Agents will often check-in with the user for approval as a natural workflow checkpoint. Sundaresan said the idea is to combine human and automated workflows.
What is becoming clear is that the next phase of enterprise AI no longer depends on model power, but on how well tools are designed to balance autonomy and control.
Pricing and Availability
As mentioned earlier, BOB is now available to all areas where IBM does business. IBM’s pricing structure for Bob consists of four primary membership tiers for each user/seat and is built around its own internal credit system called "bobcoins," Which serves as the primary metric for transparency and predictability.
These are set at a fixed valuation of $0.50 USD per 1 Bobcoin. Users consume these coins by performing specific actions, such as generating code, running commands or performing file operations. If a user’s balance runs out, they will need to upgrade their plan to continue using the service.
Here are the plans currently offered and how many Bobcoins a user receives by subscribing to each tier.
- 30 day free Trial Awarding 40 Bobcoins
-
Pro Plan $20 per month with 40 Bobcoins
-
pro+ Plan $60 per month with 160 Bobcoins
-
extremely The tier costs $200 per month for 500 Bobcoins.
All Standard plans provide access to core features including exclusive agentive mode, literate coding, BOB Shell for intelligent CLI workflows, and Model Context Protocol (MCP) integration.
While all individual plans are limited to a single user enterprise The plan is available through a sales contact, offering centralized team management, flexible role assignments, and the ability to distribute Bobcoins across the organization.
Enterprise customers get additional benefits like priority support and a dashboard to track eligibility and usage awareness.
<a href