Customers are asking their managed service provider (MSP) a new question: how are you using AI, and what does it mean for my service?
It’s an important question, because many MSPs still spend most of their time reacting: responding to alerts, fixing incidents and working through the ticket queue. Proactive work, the improvements that prevent problems and move a customer’s systems forward, is what makes the biggest difference to a service, yet it’s often the first thing squeezed out. AI can help change that balance.
At Intelligent Co-Ops, Arinco’s managed services practice, we use AI in two distinct ways every day. There’s AI in how we work, which supports the tasks our people perform, and AI directly on customer systems, which runs on the applications, platforms and systems we support.
Proactive work has always been central to how we run our service. AI is helping us go further: higher-quality work, delivered more consistently and more quickly, with even more of our team’s time spent on proactive work. The point isn’t that a provider uses AI, or whether it makes the service cheaper, it is that the customer receives a better service because of it.
Two ways we use AI
1. AI in how we work
AI in how we work supports the tasks we perform to look after a customer’s systems. Using AI to augment our team means it can draft work for our consultants to review, give feedback on work our team has developed, or do a mix of both. Examples include writing and troubleshooting code, preparing changes, producing reports and keeping documentation up to date.
2. AI directly on customer systems
AI directly on customer systems runs inside the systems we support. It monitors activity, triages alerts and incidents, gathers the evidence needed for root cause analysis and answers questions about those systems while we investigate. When something goes down, it can also help us restore the service.
Where approved by our customers, both are now part of how we run our service. Here’s how each one works in practice.
AI in how we work
Here are some examples of where we’ve built AI around the work we do most often:
- Transition In: When we take on a new customer or a new service to support, our service mapping agent helps discover the customer’s systems and document what we need to support them. Knowledge gathering and support documentation happen faster, so our team gets up to speed sooner. Later, when a component needs work, the AI-generated service map puts the right information at hand, so the task gets done quickly.
- Service management: Our IT service management (ITSM) platform has AI built in. It reads each ticket and recommends similar tickets and relevant knowledge base articles, identifies colleagues who have solved these issues before and routes the ticket to someone with the right skills. Before AI, gathering that information meant manual searches through past tickets and documentation. Now it comes straight to the consultant working the ticket.
- Understanding applications: Our Agentic Context Framework (ACF) analyses a codebase and builds a structured picture of its architecture, patterns and conventions. With that context, our engineers and AI agents can identify code issues, and build new features and enhancements, faster and with better input.
- Writing and maintaining code: Quality code has always depended on skilled engineers, peer review and change control, and it still does: code is reviewed by another engineer and, where required, goes through a change request before it reaches production. What’s changed is that AI now supports the work along the way: GitHub Copilot helps our engineers write code, and it reviews code alongside their peers. The result is higher-quality code, delivered faster.
- Analysis and reporting: Our Secure Score Analyst and Cost Optimisation Analyst turn a customer’s security posture data (from Microsoft Secure Score) and Azure spend data into engineer-ready reviews. Reports take less time to produce, and every recommendation is grounded in our best-practice approach. Because the analysts keep the history of previous reports and the reasons behind past decisions, each report stays consistent, whichever engineer produces it.
AI directly on customer systems
AI directly on customer systems is where we see the most untapped value. Incident investigation and root cause analysis are at the heart of what an MSP does, and AI working directly in those systems helps our people find the problem faster. Two examples show what this looks like in practice: our SecOps Analyst and Azure SRE Agent.
SecOps Analyst
Our SecOps Analyst triages security alerts before our consultants pick them up. It carries out the evidence-gathering steps an analyst would otherwise do by hand: reviewing activity across Microsoft security tools, checking user and device information and identifying known-safe patterns. It then prepares a structured investigation summary with recommended actions.
The results have been significant:
- Investigation time cut by more than 90%: investigations that took 20 to 40 minutes now take under two minutes, and some finish in under 30 seconds.
- False-positive alert noise cut by up to 87%: its detection-tuning capability filters out alerts that don’t need a human.
Our analysts still approve significant actions such as closing or escalating incidents and deploying detection changes, and every change is recorded in a version-controlled audit trail.
Azure SRE Agent
Beyond security, we use Microsoft’s Azure Site Reliability Engineering (SRE) Agent to bring the same approach to the platforms and applications we support. When an alert fires, it gathers context from monitoring data, logs, recent deployments and resource configuration. It then forms theories about what went wrong and tests each one against the evidence. Instead of a raw alert, our engineers receive a likely root cause, the reasoning behind it and a suggested fix, which the agent can help apply once approved.
It also learns from experience, keeping context from previous investigations, including root causes and how they were resolved. Recurring issues are recognised sooner, and whoever picks up the next incident starts with the full picture.
From reactive to proactive
When people hear “AI in managed services”, the first assumption is usually lower cost. That isn’t a given. AI can save time on individual tasks, but running it, particularly directly on customer systems, comes with costs of its own. In our experience, the real value of AI isn’t measured in cost. It’s measured in the service.
The clearest gains have been in quality and consistency. Before AI, two engineers might carry out the same routine task in two different ways. Both might get it right, but the customer’s experience depended on who picked up the ticket. When AI supports routine tasks, they are performed the same way, to the same standard, every time. That consistency has been one of the most visible benefits we’ve seen.
On customer systems, the standout gain is speed to resolution. When every investigation starts with the evidence already gathered, mean time to resolution (MTTR), the average time it takes to fix an issue, comes down. And every alert gets the same depth of investigation, whether it lands at 2pm or 2am.
The biggest shift, though, is from reactive to proactive. We’ve always prioritised proactive work, and AI is accelerating it. Faster investigations, lighter routine tasks and the right information arriving straight to the engineer all free up time. Our engineers put that time into proactive work: fixing recurring problems at their source, enhancing what’s already in place and building new capability for customers. That’s where customers feel the difference. There’s a people benefit too. Most engineers would rather build than firefight, and a more engaged team delivers a better service.
Knowledge stays in the service as well. Because our tools keep the history behind past incidents, reports and decisions, that knowledge no longer sits in one person’s head, and it isn’t lost when people change roles.
None of this matters unless it’s done properly and we can show it works. AI, whether in how we work or directly on customer systems, needs to be operated with the right controls in place, within each customer’s security requirements. It also needs to earn its place. Implementing AI for AI’s sake just burns money and tokens (the units AI services charge by) without improving anything for the customer. So we measure the outcome: is the work faster, more consistent, easier to review or more useful to the customer? Where the evidence isn’t there, we refine the process or stop using the capability.
The measure that matters
So, back to the question customers are asking: how are we using AI, and what does it mean for their service?
Our answer is two ways of using AI, working together. AI in how we work helps our people deliver better, more consistent work, faster. AI directly on customer systems helps us find and fix problems sooner. Together, they give our team even more time to be proactive. Neither replaces the experienced people behind the service.
For our industry, that’s the real opportunity. Used well, AI should help managed service teams finally move the dial from reactive to proactive. And that’s the measure that matters: not whether a provider uses AI, but whether its customers get a better service because of it. It’s the measure we hold ourselves to, and the one customers should hold every MSP to.


