IBM deploys 6,000 AI agents across consulting workflows

4 tn in future revenue: Inside AI financing loop flagged by IMF IBM Consulting Advantage (ICA), the company's consulting platform, has a dedicated research and development team of more than 1,000 engineers. “IBM has 2,000+ live projects running on ICA globally, with over 500 deployments on client projects combining digital and physical labour,” Ali said. The company is using its experience of integrating AI into its own operations to help clients re-engineer business processes, moving beyond the deployment of AI tools to redesigning how the work gets done.

Ali said IBM's approach involves embedding agents into specific tasks within workflows, with the aim of improving productivity and translating efficiency gains into measurable economic value. IBM is seeing traction in transformation projects, helping enterprises rework processes across functions such as finance, human resources and banking. Its forward deployed units bring together architects, domain specialists and AI engineers to identify problems, redesign workflows, and build and deploy solutions.

These teams use tools on IBM Consulting Advantage to re-engineer processes and can work with AI models and infrastructure from various providers. The focus on transformation comes as traditional managed and operate services face pricing pressure, with clients expecting AI to lower the cost-of-service delivery. IBM is prioritising projects that use AI to fundamentally change business processes, rather than simply using the technology to make existing tasks faster.

An investigation into a security threat that typically took 45 minutes can now be completed by a group of AI agents in about two minutes, followed by 10 minutes of human review. The approach allows teams to handle three times as many threats with the same number of people, he said. To assess whether AI deployments deliver expected gains, IBM's Bangalore-based Productivity Measurement Lab tests workflows with and without agents before they are taken into production.

“And what they do is to test workflows with agents and without agents so we can see what the productivity is,” Ali said. He cautioned that introducing AI does not always improve productivity. “There are times where you build agents, put it into the workflow, and interestingly enough, you'll get negative productivity.” For IBM, the measure of success is not the number of agents deployed or the volume of AI tokens consumed, but whether their use translates into productivity gains and better financial outcomes.