Challenge
RealBytes operates a fully remote help desk of five without a dedicated dispatcher. Three frontline technicians assessed and routed new requests, while senior staff reviewed tickets and high-volume ConnectWise SOC Services™ and ConnectWise NOC Services™ alert traffic. Blake Bath estimates the business receives about 2,000 tickets and alerts in an average month.
Solution
CW AI Agents™ were deployed across the help desk, CW SOC Services™, CW NOC Services™, ConnectWise PSA™, and practice management boards. In CW PSA™, the agents categorize and enrich tickets, use prior service context to suggest next steps, prioritize alerts, and draft resolution notes.
Results
Time to first touch moved from as long as one hour to about five minutes, while resolution-note preparation moved from about 30 minutes to five to 10 minutes. There are also more first-touch resolutions at Tier 1, faster after-hours handling, and fewer missed calls.
A remote help desk without a dispatcher
RealBytes is a Brisbane-based managed service provider (MSP) with five people covering Tier 1 and Tier 2 triage, supported by senior and project engineers. Because the help desk is fully remote, frontline technicians cannot simply turn to a senior colleague for an answer when an unfamiliar issue arises.
“On a traditional help desk, you would have Tier 1 and Tier 2 technicians sitting near a Tier 3 technician who could answer a quick question. That is much harder with a remote team. Once we turned on [CW] AI Agents, our technicians could draw on the knowledge our senior people had built over time.” - Blake Bath, RealBytes
Blake estimates that RealBytes receives about 2,000 tickets and alerts per month, including roughly 1,000 active tickets. Three frontline technicians take the first contact by phone, Teams, or email, assess the request, and decide whether to resolve or escalate it. The company has never had a dedicated dispatcher.
Faster triage and clearer routing
Previously, the practice lead or a senior technician often had to review a ticket’s subject line, correct its category, and route it to the right team. RealBytes worked towards a one-hour service objective when the help desk was busy. CW AI Agents now categorize tickets, apply customer details and priority, and begin triage when the ticket arrives.
We were working to a one-hour SLO. Now, a ticket may be touched about five minutes after it hits the board. It is already starting to be worked on, whether a person has seen it or not.
The agents also add context to requests that arrive through informal channels. When a customer emails Blake directly for a password reset and he forwards the message without service-desk details, the agent categorizes the request and gives the technician a clearer starting point.
Using service context to support first touch resolution
RealBytes enabled the AI pod in CW PSA™ and set it to read tickets automatically. The agent searches prior tickets and available knowledge sources for relevant causes and resolutions. In one case, it suggested checking a Visual Basic script that the team had not documented or considered. The Tier 1 technician resolved the issue in about five minutes on the first contact.
“That ticket might have taken a technician an hour to resolve. Instead, it was fixed in about five minutes on the first contact. Otherwise, it may have required two technicians and a ScreenConnect session. The agent surfaced something we had never documented in an SOP.” - Blake Bath, RealBytes
The agent can also compare a customer’s suggested diagnosis with device telemetry in the RMM. Blake described a computer with 8 gigabytes of memory that had used 7.9 gigabytes over the previous 10 days. That evidence helped the team move from open-ended troubleshooting to a practical conversation about hardware.
“It helps us step back and look at the whole problem. Is more than one customer experiencing it? Have we seen it before? Is there a cause we already know about?” - Blake Bath, RealBytes
Separating signal from alert noise
RealBytes also deployed CW AI Agents on its SOC and NOC boards. Those boards receive large volumes of endpoint, SentinelOne, and data-loss-prevention alerts. The agents use ticket history and context to assign priority so technicians can focus on the items that require action.
One morning, we received almost 500 Windows Defender alerts for the same CVE across our customers. The agent deprioritized the duplicates and identified the separate malware ticket that required attention.” - Blake Bath, RealBytes
Blake says this context is also helping the Tier 1 and Tier 2 staff who cover evenings and weekends resolve more issues during their shifts. For example, the agent can surface the approval steps used for an earlier password reset, allowing the technician to follow the established process instead of waking a senior colleague.
Extending Tier 1 capability
Blake reports more first-touch resolution by frontline technicians. Some tickets that previously moved between Tier 1 and Tier 2 several times can now stay with the person who first spoke with the customer. The improvement is directional; RealBytes has not yet provided a before-and-after escalation rate.
“Why keep sending a ticket up and down the queue when we can give a Tier 1 technician the context and guardrails to resolve it? They can be confident in the answer and take ownership of the issue.” - Blake Bath, RealBytes
Assignment is becoming more consistent as well. Instead of relying solely on technicians to select the next ticket, the agent can route work based on criticality and service level. RealBytes is also testing how specialist knowledge, such as experience with a customer’s medical software, could inform routing. That work remains at an early stage.
[CW] AI Agents have given us the dispatcher we never had. The agent assesses each ticket’s criticality, SLA, and scope, then identifies what needs attention.
Faster notes and reusable knowledge
Technicians often have to document a resolution while managing several tasks. Blake estimates that a detailed note that might have taken about 30 minutes can now be prepared in five to 10 minutes with an agent-drafted summary. RealBytes is also adding SOPs to resolved tickets so the agent can surface that guidance when the same issue appears again.
“A resolution note that might have taken 30 minutes to write now takes five to ten minutes. That is valuable time back for our technicians.” - Blake Bath, RealBytes
Keeping people available for customers
RealBytes reviews missed calls in its weekly team huddle. Blake reports that the daily average has moved from about five missed calls to three, although the period and contributing factors still need to be validated. He expects further improvement as the dispatch workflow matures.
“Our customers want to talk to people. Let AI handle more of the administrative work so we can keep our people focused on our customers.” - Blake Bath, RealBytes
Looking toward Predictive IT
RealBytes had used general-purpose AI tools before adopting AI inside its service platform. Blake sees the difference in the platform context: access to ConnectWise RMM™ telemetry, ticket history, and alert data. His next goal is to use agent-triggered workflows for routine requests such as self-service password resets. Those capabilities are part of RealBytes’ future direction, not results demonstrated in this case study.
We need to serve our customers to a higher standard at the same price point. I believe Predictive IT can help us do that while giving technicians more time for complex issues and project work.
As an early Asia-Pacific adopter, RealBytes has applied ConnectWise AI Agents broadly and learned where operational guardrails matter. Blake’s advice to other MSPs is to test the technology in real workflows, stay curious, and build controls as experience grows.
“We have been talking about AI for years. Now it has a practical use. You have to stay curious, put the right guardrails in place, and keep learning, because your customers are already using it.” - Blake Bath, RealBytes
Key takeaways for MSPs
- Put AI where work enters: RealBytes applied ConnectWise AI Agents across help desk, SOC, NOC, and practice management boards, so triage begins when tickets arrive.
- Start with a clear operating measure: Time to first touch, routing time, and note-writing time gave RealBytes practical ways to assess early value.
- Use service context to support frontline resolution: Prior tickets, knowledge sources, and RMM telemetry can give Tier 1 technicians a stronger starting point.
- Treat alert handling as prioritization: The goal is to surface the ticket that needs action while consistently grading duplicate or expected alerts.
- Capture knowledge as work closes: Adding SOPs and clear resolution notes to tickets improves the context available when a similar issue returns.
- Pair broader autonomy with guardrails: Established approval steps helped after-hours technicians resolve routine work without unnecessary escalation.
- Measure workflow outcomes, not deployment alone: First-touch resolution, escalation rates, service-level performance, and missed calls will show whether the operating model is improving.
- Separate current results from the roadmap: Automated workflows and self-service fixes are part of RealBytes’ Predictive IT direction and still need future validation.
