Career Plans Need a Human Network

AI agents can draft a career plan quickly, but the plan still needs human review and company context. MentorStack's MCP server lets an authorized AI assistant query mentoring data and use matching and meeting tools. A person reviews the match and meeting actions.

A career plan can become stale between review cycles. The form may record an intention without connecting the employee to a relevant person, project, or opportunity. Treat that gap as an operating problem to measure in your organization.

What changed about career planning?

An employee can ask an AI assistant to draft a 12-month plan for a role transition. The quality depends on the prompt, model, context, and human review. Even a useful draft does not provide access to people and opportunities inside the company.

What is the actual problem?

Suppose a marketing employee wants to be a director in twelve months. A draft plan might include owning a cross-functional campaign, building management foundations, leading a small team, and preparing for the promotion conversation.

What's hard is everything from there. Who should this person learn from inside the company. Who's actually made the same jump in the last two years. Who's got political capital to sponsor them when the seat opens up. Who's running the kind of project they should shadow next quarter.

The proposed bottleneck is the step between the plan and the human network: an introduction, mentor relationship, shadowing opportunity, or sponsor conversation. Measure where employees stop before treating planning or execution as the cause.

The MCP thing, briefly

MCP (Model Context Protocol) is a standard Anthropic released in late 2024 that lets compatible AI clients call external systems through a common interface. It can connect an agent's reasoning to tools where authorized actions happen.

One proposed use is to connect an AI-produced career plan to the skills, aspirations, and experience data held in a mentorship platform. Test whether that workflow reduces the time from plan to a reviewed match or meeting action.

MentorStack's MCP server lets an authorized AI assistant query mentoring data and use matching and meeting tools. A person reviews the match and meeting actions.

This workflow does not replace a mentor with an AI coach. The AI output remains a draft, and a person reviews the match and meeting actions. The MCP server supplies the authorized connection between the systems.

What could change for HR teams?

If employees draft plans between formal reviews, HR teams can test a second question alongside the quality of their frameworks: can an employee reach a relevant person or opportunity when the plan calls for one? This may shift some effort from form design to network access, matching, and follow-through. It does not prove that annual reviews or development forms are obsolete.

The hypothesis is that easier access to relevant people can reduce the gap between a development plan and its first action. MentorStack has no customer retention result that proves this effect.

What is the bounded claim?

AI-generated career plans can be a useful starting point because they are fast to produce. Their quality depends on the prompt, the model, the available context, and human review. MentorStack has not tested them against the plans produced by most companies.

The hypothesis is that friction between a plan and the human network can be a material bottleneck. Connecting AI assistants to mentorship tools through MCP is one method to test. Measure time to a reviewed match, meeting follow-through, and participant feedback before claiming that it removed the friction.

So

If you run People or L&D, review where employees already use AI for career planning and what human support follows. That evidence can guide whether you need new infrastructure.

If your mentoring program has outgrown a spreadsheet, MentorStack runs it, and its MCP server lets your AI assistant read the program data. mentorstack.co (opens in new tab)

The practical question is what AI changes about career planning and which parts still require human judgment, access, and accountability.