AI assistants are moving beyond answering IT questions to performing tasks that once required technicians to swivel-chair between management platforms. The promise of greater productivity comes with concerns for managed service providers that want to use agentic tools without ceding oversight.
Earlier this month, software vendor GoTo added a Model Context Protocol server, AI-based insights and multi-tenant capabilities to its LogMeIn Resolve IT management support platform. The enhancements, now available in beta, give technicians tools to work with Anthropic’s Claude LLM and other third-party models in a secure environment protected by existing permissions and log actions.
For MSPs, the move addresses a longtime problem, Steve Shattuck, VP of global partner ecosystems at GoTo, told Channel Dive.
“MSPs are not short on tools — they’re short on people,” Shattuck said. “Most can’t hire technicians as fast as their client count grows, so the useful test for any AI feature is whether it lets a team take on more work without adding headcount.”
GoTo is aiming to give MSPs visibility into processes that have been managed across discrete consoles. Technicians already work across professional services automation, documentation and security tools. Returning to a remote monitoring and management console adds another stop.
“A platform that demands everything happens in its own separate console will lose to one that shows up where the work is already happening,” Shattuck said. “We want Resolve to be the system that MSPs trust for the parts that have to be right — the device data, permissions and record of what happened. Those need to hold up whichever tool the technician is using.”
Rather than supplanting RMM and PSA vendors, GoTo favors integrations. The company updated its HaloPSA integration in July 2025 to bring Resolve device context into the PSA and improve ticket synchronization. A multi-tenant software inventory now provides per-client install counts, filtering and audit-ready reports.
Through the MCP server, technicians can leverage AI to retrieve information on devices, alerts, tickets, assets and knowledge-base entries. They can use AI assistance to create or update records and initiate remote support sessions within existing permissions.
Shattuck cited a customer who had Claude compare device inventory against Microsoft and Apple end-of-support dates, organizing machines by support status and action needed.
“What would have taken their team a complicated spreadsheet and several hours was accomplished in just a few minutes,” he said.
As vendors open IT management platforms to AI, MSPs will have to scrutinize the underlying capabilities. Real-time device data, remote session performance and permission control reliability across multiple customer environments then becomes table stakes.
“Once everyone has an MCP server, the connection stops being the difference,” Shattuck said.“Even if a technician never opens the console, they still feel the difference in whether the answers are right and the access is safe.”
Autonomy anxiety
Authorizations get thornier when assistants move from reading records to changing them. Resolve requires a named technician to approve each AI action before execution and records that technician's identity in the audit log, Shattuck said. Even so, the audit trail has a limitation.
“Resolve doesn't currently document an ‘AI-initiated’ label in the audit log,” Shattuck acknowledged.
The audit documentation identifies the approving technician but doesn't separately flag AI-originated work. GoTo is seeking partner feedback during the beta on more specific AI logging ahead of general availability.
Authorization doesn't guarantee accuracy, either. An approved assistant could still pinpoint the wrong device within its permitted access. For higher-risk actions, Shattuck offered a litmus test in the form of a question: Does the action affect more than one customer?
“If so, a human should always make the call,” he said.
Shattuck recommended three safeguards tailored to multi-client operations:
- An AI assistant should never have more access than the technician managing it.
- Limit agent access to one client at a time, with strict tenant separation and technician confirmation before acting across customers.
- Start with small pilots, review AI output before acting, and check logs to trace and reverse mistakes. AI, Shattuck said, can be “confidently wrong,” including about which device a ticket concerns.
Looking ahead, Shattuck expects most MSPs to retain a core platform for device management and remote work, using AI as “the front door.” Others may connect several vendors through one assistant.
As MSPs gain freedom to choose AI interfaces without replacing their core RMM platforms, vendors face pressure to prove the reliability of their tools, the accuracy of their data and the strength of their security controls.
“Vendors are changing how they see their role, from ‘the place technicians work’ to ‘the system technicians can trust from wherever they work,’” Shattuck said.
