Growth brings opportunity, but it also creates new demands on the systems and teams that support it. For a US-based technology company serving thousands of customers and enterprises, increasing ticket volumes and service requests were making it harder to scale service delivery through people-led operations alone.
The company provides enterprise expense, mobility and cloud management solutions to more than 2,000 customers through an ecosystem of over 35 partners and approximately 1,500 enterprises. Its technology landscape spans cloud, infrastructure, enterprise applications, identity, collaboration, end-user computing and customer-facing platforms.
Persistent was helping keep this environment running across a broad set of services, teams and platforms, providing the people, processes and expertise needed for day-to-day operations. However, the next level of value required a different approach. The goal was to reduce avoidable work, automate repeatable activities and improve how demand was managed.
The scale of that pressure became clear across the data Persistent reviewed:
- 71.7% of incidents escalated to L2
- 65% average ticket reassignment rate
- 9.2 hours average incident resolution time
- 20% automation index against a target of 65% or higher
- High dependence on people-led delivery
Changing the Managed-Services Equation
To understand where effort was being consumed, Persistent analyzed more than 80,000 tickets, covering 22,021 Incidents and 66,030 Service Requests over a 17-month period. The analysis was done using Persistent AI accelerator, P(AI)Scout
The analysis revealed two distinct opportunities. Incidents required better detection, diagnosis and remediation, while Service Requests offered significant potential for self-service, deflection and knowledge-led resolution. The data also showed false-positive and reassignment patterns, including one high-volume performance category where 67% of tickets were closed as “Works as Designed.”
The findings made one thing clear: faster ticket processing would not be enough. Persistent needed to change what entered the system, how it was routed, what could be automated and where governance was required.
Persistent initiated a mid-contract shift from staffing and effort to AI, automation and measurable outcomes. The goal was to ensure that the productivity gains created by AI improved the client’s operations and economics, not just the provider’s margin.
Persistent proposed a deliberate reduction in its own run-rate revenue and committed to about 30% client cost savings over five years, including significant savings in the first year. It backed that commitment with investments in AIOps and agentic AI workflows, alongside a 25% reduction in the engineering team. The delivery approach also shifted engineers away from repetitive activity toward oversight, exception management and continuous improvement.
This was not a standard cost-reduction exercise. It was a way to pass the benefits of AI-led efficiency back to the client while aligning Persistent’s commercial model more closely with the outcomes the client wanted to achieve.
Building an AI-Led Operating Model
The roadmap focused on three areas where AI and automation could have the greatest impact.
Eliminating unnecessary demand
The first step was to separate tasks that needed human judgment from work that did not require any human intervention, which are typically reversible low impact on the production systems. This included reducing alert noise and duplicate incidents, correlating related events across infrastructure layers, removing unnecessary reassignment and escalation, addressing recurring sources of operational effort and rationalizing overlapping tools and workflows.
Automating repeatable work
Persistent then looked at which repeatable incidents and requests, that are consuming high toil, could be handled through the client’s existing technology investments. ServiceNow capabilities could support self-service, AI search, intelligent classification, routing and operational analytics, while Logic Monitor’s AIOps capabilities could support anomaly detection, alert correlation, root-cause analysis and approved automated remediation.
For service requests, the analysis identified highly repeatable activities, including:
- 15,800+ phishing and spam triage requests
- 2,000+ access and provisioning tickets
- 1,430+ device and software installation requests
- 640+ password-reset requests
- Alert to ticket ratio is 90 %, which meant high noise in the ticketing system
These patterns provided an evidence base for deciding which work could be prevented, deflected or automated.
Introducing agentic operations
For activities that existing platforms could not address alone, Persistent designed and deployed specialized Agentic AI workflows to support more complex scenarios These included a ticket summarization agent, cross-system triage agent, knowledge agent, proactive health-check agent and on-demand diagnostic agent.
Together, these capabilities would enable them to move from reactive support toward AI-assisted, predictive and increasingly autonomous operations.
Building Trust Through Governance
Governance was part of the design from the beginning. AI agents operate with human oversight, confidence thresholds and defined fallback mechanisms. Automated remediation is limited initially to approved, low-risk activities and actions are controlled, logged and reversible.
Client environment data is accessed through controlled environments and client information is not intended to be used for model training. Knowledge created by AI is reviewed before publication, while the scope of autonomy expands only after each use case meets agreed accuracy and safety thresholds.
Engineers remain central to decision-making, while AI surfaces context, handles repeatable work and coordinates activity across the environment. This allows the client to increase automation while maintaining accountability, control and data protection.
Sharing the Value of AI-Led Productivity
The shift went beyond technology, changing what the client paid for, how engineers contributed and how value was shared.
The AI-led operating structure was built to share value with the client over the five-year term. Persistent’s revenue is reduced by design so the client benefits directly from the change.
The engagement has already delivered several proof points, including:
- 50% facility cost reduction
- Sixfold improvement in network performance
- Query resolution reduced from four to five days to less than 10 minutes
The incident and service-request improvements below remain roadmap or projected benefits unless separately validated as realized results. The roadmap also points to additional opportunities:
- 74% of incidents identified as AI-addressable
- 61.6% of service requests identified as AI-addressable
- 35% to 45% potential deflection of L1 demand
- 20% to 30% potential shift of selected L2 service requests to L1
- 45% – 60% target for the automation index
Reimagining What Managed Services Can Deliver
By changing how work was delivered, governed and measured, Persistent helped the client move beyond incremental efficiency gains and build a more scalable foundation for managed services.
The client gains a clearer path toward lower demand, faster resolution and responsible autonomy. Engineers spend less time processing repeatable work and more time governing exceptions, improving systems and delivering business outcomes.
For Persistent, the engagement demonstrates how managed services can evolve in the AI era: use data to eliminate unnecessary effort, apply AI to repetitive work, introduce autonomy with appropriate safeguards and align commercial outcomes with client value.
By choosing to disrupt its own effort-based approach, Persistent helped create a more scalable, outcome-led path for managed services.
