Skip to main content

October Hand-Off — Mandate Plans

Mandates PS-030 through PS-037: peak season supervision, hand-off, residual sizing, and permanent-seat recommendation.

PS-030 & PS-031: Supervise Agents During Peak Season​

What this means​

October is Madison Reed's busiest month: fall hiring ramp + open enrollment. This is the proof. If agents can carry the throughput during the heaviest month, the automation works. You supervise: monitor exceptions, intervene on failures, validate output.

Concrete tasks​

  1. Set up a monitoring dashboard before October starts: agent run status, exception queue depth, auto-clear rate, failed runs.
  2. Define escalation paths: when do you step in vs. let the agent handle it? What constitutes a failure that needs manual intervention?
  3. Track every intervention you make during October. Each one is data for the residual sizing.
  4. Compare agent output vs. manual: are the agents catching everything? Are they producing false positives that waste Syra's time?

People to meet​

PersonQuestions to ask
SyraDuring your busiest weeks, how many onboardings/job changes do you process? What's the error rate? What do you wish someone else handled?
ShlandaWhat does "peak season" look like for the People team? How many hires per week? What breaks?

PS-035: Transfer Platform Ownership to the Team​

What this means​

The Wednesday 9-10:30am CT working block is the transfer mechanism. By the end of October, Shlanda's team (Shlanda, Nickole, Syra) should be able to:

  • Run the agents themselves
  • Read and respond to the Exception Queue
  • Extend the Slack bot for new use cases
  • Make basic changes in the mr-people repo (edit rules, add config)
  • Understand what the agents do and don't do

Concrete tasks​

  1. Build a training curriculum for the Wednesday blocks (what to cover each week in September and October).
  2. Create operator documentation: one-page runbooks for each agent (how to check if it ran, what to do if it didn't, how to read the Exception Queue).
  3. Pair-program with Nickole on at least one agent extension (she's the AI-Ops candidate; building her capability is strategic).
  4. Do a "Daniel is out" simulation in late October: the team runs everything for a week with Daniel available only for escalation.

People to meet​

PersonQuestions to ask
ShlandaWhat does "your team owns this" mean to you? What would make you confident the team can run it without CMPRSSN?
NickoleWhat would you need to feel comfortable maintaining and extending the agents? What's your comfort level with GitHub and Claude Code?
SyraSame question: what would you need to run the recon agent review without Kyle or Daniel available?

PS-036: Document the Residual Human Work​

What this means​

After three months of lived experience (doing the work, building agents, supervising peak season), you produce a clear accounting of what still needs a person:

  • Which tasks could not be automated and why (genuine judgment, relationship work, field presence, legal sensitivity)
  • How many hours per week the residual human work takes
  • What skills the residual requires (is it an administrator skill set or a builder skill set?)
  • Whether the residual is one job or pieces of multiple jobs

Concrete tasks​

  1. Maintain a running residual log from Day 1: every time you do something that can't or shouldn't be automated, log it with time and reason.
  2. Categorize the residual at the end of each month: judgment calls, relationship work, exception handling, one-time projects, compliance gates.
  3. Quantify: total hours per week, by category, with trend (is the residual shrinking as agents improve, or stable?).
  4. Write the residual report in October: a clear document for Brad that says "here's what still needs a person, here's how many hours, here's what kind of person."

PS-037: Deliver the Permanent-Seat Recommendation​

What this means​

At the end of October, you deliver a recommendation to Brad (via Shlanda):

OptionWhen it makes sense
Permanent hireResidual is large enough to justify a full-time seat. Specify the scope and altitude (builder vs. administrator).
Fractional / contractResidual is real but not full-time. Specify hours per week and skill set.
NoneAgents carry the load; team absorbs the residual at a higher altitude. Specify what the team needs to keep doing.

Concrete tasks​

  1. Base the recommendation on data, not opinion: hours logged, tasks categorized, agent performance during peak season.
  2. Write two versions: one for Shlanda (detailed, operational), one for Brad (executive summary with cost comparison).
  3. Include the follow-on block structure: if the answer is "fractional," what does the next engagement look like? ATS go-live, LMS build, Workday migration?
  4. Be honest about what didn't work: if certain automations underperformed, say so. The recommendation is stronger with acknowledged gaps.

People to meet​

PersonQuestions to ask
MikyoWhat does CMPRSSN's recommendation look like? What's the format Brad expects? How do we frame follow-on engagement without being self-serving?
ShlandaWhat would make the recommendation credible to Brad? What data does he need to see?