GlobSynk
GlobSynk Academy™
Module 2 of 9~32 minutes

Designing Your Assistant's Role & Boundaries

A role says what an assistant is for, what it may help with, what it must never decide, and when to stop — written down, because drift is only invisible when nothing was fixed.

Objectives

  1. State, in one sentence, what a specific assistant is for.
  2. List the tasks an assistant is allowed to help with, in their own words.
  3. Name at least two decisions they will not delegate, and say why.
  4. Write a stopping rule that tells the assistant when to hand the task back.
  5. Tell the difference between a role and an instruction for a single task.
  6. Recognize role creep in a real example, and name what caused it.
  7. Explain why authority stays with them even when the assistant is right.

Introduction

Module 1 asked whether a task should be delegated at all. This module answers the next question, which arrives the moment the answer is yes: what exactly is this assistant for, and where does it stop?

Most disappointment with assistants is not a capability problem. It is an undefined-role problem. A person opens a chat window with a vague hope, gets something vaguely useful, and slowly starts handing over more than they meant to — not by decision, but by drift.

A role written down once prevents that, because drift is only invisible when nothing was ever fixed.

By the end of this module you will have a written role you can hold a real request against — and you will have found at least one place where your own list contradicted itself.

Lesson 1

A Role Is a Job Description, Not a Personality

Dislodge "you are a helpful assistant." A role says what work this thing does for you, for whom, and to what end — not what character it plays.

Lesson 2

Naming the Work It May Help With

Move from "help me with admin" to a short list of actual tasks. Specific enough that someone else could tell whether a request belongs.

Lesson 3

The Decisions That Stay Yours

The heart of the module. Some decisions are not delegable at any quality level — because the consequence lands on a person, not on a draft.

Lesson 4

Stopping Rules: Telling It When to Stop

A boundary that only exists in your head is not a boundary. Teach the learner to write the condition under which the assistant should stop and hand back.

Lesson 5

How Role Creep Actually Happens

Nobody decides to over-delegate. It happens one convenient exception at a time. Show the pattern so the learner can catch it in themselves.

Lesson 6

Authority Is Granted, Never Inherent

Carry Module 1's principle into practice: a correct answer is still not a decision, and being right does not transfer accountability.

The four parts of a written role

A role worth writing has four parts, and most people write only the first.

  • Purpose — what is this assistant for, in one sentence?
  • Allowed tasks — what may it help with?
  • Non-delegable decisions — what must never be decided by it?
  • Stopping rule — when should it stop and hand back to me?

Write all four or none of them work. A purpose without a stopping rule is an invitation to drift; a stopping rule without allowed tasks has nothing to bound.

Where authority actually sits

Read this as a sequence. The assistant occupies exactly one position in it, and never the last one.

You define the role → the assistant works inside it → it produces a result → you review it → you decide → the outcome is yours.

Two things are worth noticing. The assistant never touches the last two steps. And the first step is the only one that is genuinely optional — which is precisely why skipping it causes the rest to blur.

Where authority actually sits

The assistant occupies exactly one position, and never the last one. The first step is the only optional one — which is why skipping it blurs the rest.

A six-step sequence showing where authority sits. You define the role, the assistant works inside it, it produces a result, you review it, you decide, and the outcome is yours. The assistant occupies exactly one position and never touches the last two steps. The first step is the only one that is genuinely optional, which is precisely why skipping it causes everything after it to blur.

  1. You define the role

    The only genuinely optional step — and skipping it blurs all the rest.

  2. The assistant works inside it

    Its one position in the sequence.

  3. It produces a result

    A draft, an option set, a summary.

  4. You review it

  5. You decide

    The assistant never touches this step.

  6. The outcome is yours

    Nor this one.

Practical exercise

≈7 min

A role you will actually use. Pick one assistant you have or plan to use — one, not a category — and write its purpose in a single sentence. If you need "and" twice, you are describing two assistants; split them.

List three tasks it may help with, drawn from your own week. Then write two decisions you will not delegate, with one sentence of reasoning for each. The reasoning matters more than the decision: it is what you will reuse next time.

Write one stopping rule by completing this sentence: "Stop and tell me when ___." Then read the four parts together and look for a gap — a task on your allowed list that could quietly produce one of your non-delegable decisions.

Nothing is submitted or stored. If you would rather not use your own assistant, or cannot right now, use the supplied worked roles below instead — they teach the same skill.

Three worked example roles are supplied below — a student's study assistant, a small-business owner's correspondence assistant, and a caregiver's appointment-organizing assistant. Each shows all four parts filled in, and each contains one deliberate flaw: one allows a task that could produce a non-delegable decision, one has a stopping rule too vague to trigger, and one states a purpose that quietly contains two jobs. Finding all three is the same work as writing your own role, and practices the harder skill of reviewing one.

Your progress

0 of 2 required activities complete in this module · course progress 0%

  • ○ GlobSynk Lab™ · optional
  • ○ Reflection
  • ○ Checkpoint

GlobSynk Lab™

optional, ≈5 min

About GlobSynk Lab™. GlobSynk Lab™ is the hands-on practice experience used throughout GlobSynk Academy. This Lab is optional hands-on practice: complete it now, skip it and continue the module, or return to it later. Skipping this Lab does not prevent you from continuing the course. For these activities you'll use an AI assistant that you already have access to, then return to Academy to continue your reflection and checkpoint.

Take the role you wrote and give an assistant only its purpose sentence and allowed tasks — not the stopping rule. Ask it for help with something just outside the list.

Then read what came back and answer one question: did it notice it was outside the role, or did it simply help? Most will simply help. That is the point: the boundary was never in the tool. Add your stopping rule and try the same request again.

Reflection

≈2 min

Think of something you have gradually handed over to a tool, an app, or another person — more than you originally intended. When did you last actually decide to hand it over? What would have had to be written down for the drift to be visible?

This reflection is yours alone — it is never sent to GlobSynk or stored. Only the fact that you completed it is saved.

Checkpoint

5 questions · unscored gate · instant feedback · retry as often as you like.

1. Four things a learner wrote down. Which one is a role rather than an instruction for a single task?
2. An assistant drafts a message declining a family member's request for money. Which part of this is not the assistant's to decide?
3. Two stopping rules for the same assistant. Which one can actually trigger?
4. A learner's assistant began drafting replies and now sends them. What most likely happened?
5. An assistant gives advice that turns out to be exactly right. What does that establish?

Answer all 5 questions to continue.

Key takeaways

  • A role has four parts: purpose, allowed tasks, non-delegable decisions, and a stopping rule.
  • A role is a job description, not a personality — "you are a helpful assistant" defines nothing.
  • Some decisions stay yours at every quality level, because the consequence lands on a person.
  • A stopping rule needs an observable condition, or it can never trigger.
  • Role creep happens through convenient exceptions, not decisions — which is why the role is written down.

Practice in Prompt Lab — optional

Optional

Practice defining one assistant role and one clear boundary.

Try one personal-assistant role in Prompt Lab: state its purpose, name two tasks it may help with, and give it one boundary it must not cross. Then make a request that sits just outside that boundary and see what happens. The interesting result is not obedience — it is noticing how much of the boundary was doing work, and how much was doing nothing.

Optional, and never required. Prompt Lab is not GlobSynk Lab™. Opening it is not part of module completion, the checkpoint, course progress, the Final Assessment, the certificate or Reward Points, and a learner who never opens it completes this course exactly as normal. Academy never depends on Prompt Lab being reachable.

Practice Prompting in the Real World

Opens in a new tab. Optional practice — never required for this module, the checkpoint, your progress, the Final Assessment, the certificate or Reward Points.

Before you move on

You now have a role you could hand to someone else and have them apply consistently. That is a real boundary, not a hope.

But a role describes what an assistant may do — not what it knows. The first time you use yours on something real, you will hit the same wall every time: it does not know your situation, your preferences, or the constraints that make your version of the task different from anyone else's.

Supplying that deliberately — and no more than that — is Module 3.