Getty Images/iStockphoto
5 agentic AI case studies for CIOs
Agentic AI can help organizations automate complex workflows. Explore real stories of what worked, what didn't and what challenges kept CIOs up at night.
Executive summary
TechTarget interviewed five IT leaders about how they deployed agentic AI in enterprise environments:
- PromptCare: Automates patient onboarding workflows to reduce referral-to-appointment time.
- General Assembly: Uses a multi-agent system to accelerate curriculum development, cutting first-draft creation time.
- DXC Technology: Applies agentic AI in the SOC to triage alerts and reduce response times.
- Rimini Street: Implements agentic AI with strong governance controls, using an AI steering committee, access policies and monitoring to manage enterprise-wide adoption.
- Info-Tech: Highlights the need for standardized workflows, multi-agent architectures, customization and ongoing evaluation.
Agentic AI is reshaping how organizations execute end-to-end workflows.
Across industries such as healthcare, education, cybersecurity and IT services, CIOs have applied agentic systems to automate multi-step processes. These deployments consistently show that success depends less on individual models and more on how well organizations design workflows, define roles between humans and AI, and establish governance frameworks.
TechTarget interviewed CIOs and other tech leaders from various industries about their agentic AI deployments. These case studies highlight how enterprises can translate agentic AI into measurable improvements and navigate challenges such as cultural resistance and system complexity.
Common themes from these case studies include the following:
- Start with clearly defined processes.
- Be upfront with employees about how AI will affect them.
- Don't wait for perfection before deployment.
- Maintain a human in the loop for critical decisions.
- Measure outcomes in terms of both efficiency and ROI.
IT leaders can explore the following case studies to learn how these principles play out in practice across different industries and use cases.
PromptCare: Streamlining healthcare operations with agentic AI
PromptCare, a New Jersey-based healthcare provider, implemented agentic AI to automate repetitive workflows in patient onboarding. The system orchestrates AI agents and robotic process automation tools to ingest referrals, process documents and analyze patient eligibility for therapy.
CIO Phil Merrell emphasized a process-first approach, in which organizations map end-to-end workflows in detail before introducing automation. The deployment also faced cultural resistance, requiring clear communication about how employees would interact with AI. While the system maintains humans in the loop for final validation, Merrell noted that some jobs previously performed by employees were automated, with staff shifting toward exception handling and monitoring. PromptCare also adopted a phased deployment strategy guided by the 80/20 rule -- deploying at roughly 80% completeness and refining the final 20% through iteration.
Key lessons:
- Start with process mapping before introducing automation or agents.
- Use the 80/20 rule to avoid over-engineering before deployment.
- Maintain human-in-the-loop oversight for compliance and trust.
- Address cultural concerns early with clear communication about roles.
- Measure both efficiency gains and monetary ROI.
General Assembly: Accelerating curriculum development with multi-agent systems
General Assembly, an education and training company, adopted agentic AI to help its curriculum development team keep up with demand. The company built a proprietary system called GAIA -- the General Assembly's Intelligence Application -- to orchestrate specialized agents aligned to real roles such as instructional architect, QA architect, subject matter expert and learning experience designer.
Rather than replacing human expertise, GAIA accelerates the early stages of content creation and has reduced first-draft development time by roughly 90%. This enables teams to move from a blank page to a structured draft in minutes. Humans stay in the loop to focus on refinement, contextual judgment and client-specific customization.
The implementation also highlighted the importance of engineering involvement, evaluation frameworks and governance around content quality. Initial assumptions about simple, plug-and-play deployment proved optimistic. The actual implementation required iterative design and close collaboration between engineering and curriculum teams.
Key lessons:
- Use multi-agent architecture to mirror real-world roles and workflows.
- Target high-friction tasks -- such as creating first drafts -- for automation.
- Expect significant engineering involvement for production-grade systems.
- Establish clear evaluation criteria for quality and consistency.
- Maintain human review checkpoints for all customer-facing outputs.
- Start with a focused pilot team before scaling broadly.
DXC Technology: Transforming the security operations center
DXC Technology, a global IT services company, applied agentic AI to its security operations center (SOC) to improve threat detection, triage and response at scale. Under CISO Mike Baker, the company implemented agentic capabilities across tier-one SOC functions, automating the initial handling of security alerts and allowing human analysts to focus on more complex activities such as threat hunting and incident response.
The system significantly improved operational metrics, including reductions of up to 80% in ticket acknowledgment and triage times. DXC also redefined workforce roles, upskilling tier-one analysts into more strategic positions while maintaining human-in-the-loop validation for AI-driven decisions. The implementation relied on continuous feedback loops between human analysts and the AI system to reduce false positives and improve accuracy over time.
DXC deployed iteratively and used active human validation, both of which were central to the deployment's success. Analysts worked alongside AI systems to continuously assess and refine outputs.
Key lessons:
- Apply agentic AI to high-volume, repetitive workflows like SOC triage.
- Have humans validate AI outputs in real time during deployment.
- Continuously feed human feedback into the system to improve accuracy.
- Upskill affected staff into higher-value functions.
- Focus on measurable improvements such as response and triage times.
- Partner closely with vendors and iterate jointly during implementation.
Rimini Street: Governing agentic AI with enterprise-wide controls
Rimini Street, an enterprise IT services provider, deployed agentic AI internally across customer research and service operations, including a "Deep Research" capability that aggregates data from CRM systems, financial platforms and service management tools. The system analyzes data from both internal and external sources to offer a 360-degree view of customers and significantly reduces employees' manual research time.
Rimini Street's approach strongly emphasizes AI governance. The organization established an AI steering committee composed of stakeholders from IT, HR, legal and business units. This committee evaluates all AI use cases through a structured intake process, so tools meet legal, privacy and operational standards before deployment.
The company also implemented formal training, access controls and ongoing monitoring mechanisms, including model drift detection and usage policies. Governance extends across multiple user categories, with differentiated access rights for developers, power users and general employees, which keeps agent development and usage controlled as adoption scales.
Key lessons:
- Establish a centralized AI governance body with cross-functional representation.
- Implement formal intake and approval processes for AI initiatives.
- Define clear policies, training and usage guidelines before deployment.
- Segment users by role with appropriate access controls and permissions.
- Monitor models over time to detect drift and maintain performance.
Info-Tech: Lessons on agentic AI in practice
Martin Bufi, principal research director at Info-Tech Research Group, works directly with enterprise teams to move agentic AI from prototypes into fully developed systems. With more than a decade of applied AI experience across fintech, healthcare and manufacturing, he advises organizations on how to design and roll out agent-driven workflows in real business environments.
Bufi's perspective carries weight because he has seen how enterprises actually deploy agents at scale. He consistently works with clients to "agentify" existing processes and integrate those systems into live workflows across areas such as IT service management, finance and HR operations.
He points to recurring patterns that slow adoption. Many organizations lack clearly defined process flows, which makes automation difficult from the start. He also notes that single-agent architectures rarely succeed in practice -- real deployments rely on multiple coordinated agents working across steps in a workflow. Additionally, he stresses that off-the-shelf agents typically fall short in enterprise settings and require code-driven customization to align with internal systems and processes.
Bufi also highlights cost and model selection as an often-overlooked factor. Not every step in a workflow requires the most powerful model, so teams must balance performance with efficiency as they design agent systems.
His broader takeaway is that enterprises must treat agentic AI as an engineered, end-to-end system rather than a plug-and-play tool. Success depends on disciplined workflow design, multi-agent orchestration, governance, evaluation frameworks and ongoing iteration.
Key lessons:
- Standardize workflows before introducing automation.
- Design multi-agent systems rather than relying on a single agent.
- Expect customization because off-the-shelf agents rarely work in isolation.
- Use stronger models selectively rather than at every step.
- Establish KPIs, evaluation frameworks and governance early.
- Treat agentic AI as a continuous lifecycle, not a one-time deployment.
Cross-case insights CIOs should remember
Across these implementations, several consistent patterns emerge. Successful agentic AI deployments begin with clearly defined workflows rather than technology-first thinking. Multi-agent architectures are the norm -- not the exception -- and human-in-the-loop oversight remains essential across industries.
Organizations that achieve measurable ROI tend to focus on full workflow improvements rather than isolated task automation. Cultural change and communication are just as important as technical execution, particularly when AI affects job roles and responsibilities. Finally, governance, evaluation and iteration are not optional -- they are continuous requirements for scaling agentic AI in enterprise environments.
Tim Murphy is site editor for Informa TechTarget's IT Strategy group.
