A man reviews AI-assisted data and workflow dashboards on dual monitors in a modern office.
AI’s most useful role inside a managed service provider may be preparing people to make better decisions—not making those decisions for them. According to CRN, AllSafe IT chief operating officer Nathan Phinney made that case at The Channel Company’s XChange NexGen 2026 event, arguing that AI should turn knowledge workers into “wisdom workers.” His distinction: employees remain responsible for judging the output.

For MSPs, that is a more useful proposition than simply asking how many tasks an agent can automate. The practical question is whether the finished work becomes faster, more accurate or more useful after someone has checked it.

A man reviews AI-assisted data and workflow dashboards on dual monitors in a modern office. From information gathering to human judgment​

Phinney’s examples span project-digest emails, customer-support briefings, candidate interview preparation and discovery-call transcripts converted into Markdown notes and proposals. He also described personal news and music agents, alongside potential uses such as log analysis, translation and PowerShell generation. He credited the proposal workflow with improved sales close rates, although CRN provided no numerical results or comparison baseline.

The common thread is preparation: assembling material before a person acts on it. That suggests a sensible starting point for MSP experimentation—have AI produce something reviewable before giving it authority to change something consequential.

Nigel Ziyad, CEO of Ziyad Technical Resource Consulting, told CRN he intended to use Phinney’s examples in customer discussions, while emphasizing the need for controls rather than automatic trust in generated answers.

These are workflow anecdotes, not reproducible implementation guides. CRN does not identify the underlying models, permissions, retention settings or accuracy tests. Nor should Phinney’s description of models as producing a weighted average of answers be mistaken for a comprehensive technical explanation.

Adoption is not the same as a measured return​

Nerdio’s 2026 State of the MSP Business Report, released October 20, supplies a useful reality check. Among respondents:

  • 32% had adopted AI for internal operations.
  • 12% reported measurable productivity gains.
  • 14% operated a revenue-generating AI service line.

The methodology matters. Thirdside surveyed 153 MSP decision-makers managing Microsoft 365 and Azure, using an independent research panel between June 22 and July 7, 2026. These are self-reported findings from that sample, not a census of the industry. The headline percentages also do not establish what proportion of AI adopters achieved gains; that requires a specific cross-tabulation.

The report adds a particularly relevant finding beyond adoption: only 14% consistently applied a documented AI governance framework across every customer using Copilot or other AI tools. Meanwhile, 17% reported customer inquiries about those tools that they were not yet ready to address.

Nerdio sells management software, so its conclusions deserve that commercial context. Nevertheless, the survey supports a narrower, useful observation: deploying AI and demonstrating its value are separate accomplishments. It does not prove that governance gaps caused disappointing productivity results.

What an MSP should measure next​

Our practical recommendation is to turn the “wisdom worker” idea into a small, auditable pilot rather than a company-wide slogan:

  1. Choose one bounded deliverable. A draft project digest is easier to assess than an open-ended instruction to improve service delivery.
  2. Define the evidence it must preserve. Require references to the underlying tickets, notes or records so a reviewer can check important claims.
  3. Name the reviewer and approval boundary. Decide who approves a proposal, customer message or generated script before it leaves the drafting stage.
  4. Count correction time. Compare the existing workflow with AI preparation plus human review—not generation speed alone.
  5. Set a stopping rule. Specify which omissions, unsupported claims or inappropriate data access should pause the pilot.

For a PowerShell experiment, we would make script review and testing explicit approval gates, not treat a convincing explanation as permission to execute. For a proposal workflow, we would separately check customer requirements, endpoint counts and commercial commitments.

Those are evaluation recommendations, not claims about safeguards AllSafe IT has implemented.

The useful version of “wisdom workers”​

The strongest interpretation of Phinney’s argument is not that technical knowledge suddenly matters less. It is that the person reviewing AI output needs a clear standard for accepting, correcting or rejecting it.

For Microsoft-focused MSPs, the next useful milestone should therefore be a workflow that demonstrably saves effort without weakening accountability. A polished draft is an intermediate result. A checked, dependable deliverable is the product.

 

References

  1. MSP COO: AI Should Turn Knowledge Workers Into ‘Wisdom Workers’ - CRN CRN Fri, 30 Oct 2026 22:17:00 GMT
  2. 32% of Managed Service Providers Have Adopted AI, but Only 12% Can Show It’s Paying Off - Nerdio getnerdio.com