Use your browser print dialog and choose "Save as PDF" (landscape, background graphics on).
Back to deck
CSC Network
CSC Network Webinar · Part 2

AI 102 for CSC Members

From Good Prompts to Repeatable Workflows

September 22, 2026 · 11:00 AM CT
45-minute session · 15 minutes of questions
Where we left off

Last time and this time

Last time, we looked at how to get useful AI output. Today, we are going to connect that output to a repeatable business process.
AI 101
Business context + clear task + human review = useful AI output
AI 102
AI is a tool. The workflow is the strategy.
AI 102 for CSC MembersSteve Cooker · Logic Sphere
The distinction

The prompt formula and the workflow

The prompt formula
Defines the AI task
Role + Task + Context + Rules + Output Format + Tone
The workflow
Defines what happens before and after that task
Who starts it, what context is gathered, who reviews it, who acts, how it is tracked
Before
AI task
After
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Agenda

What we'll cover today

  • From a good prompt to a working workflow
  • Featured example: the missed mat service, built into a process
  • Supporting examples: SOPs and internal knowledge
  • Choosing a workflow: value, complexity, then risk, data, and approval
  • Implementation: knowledge pack, reviewer, guardrails, pilot
  • What to do in the next 30 days
AI 102 for CSC MembersSteve Cooker · Logic Sphere
60-second reminder

The prompt formula still applies

Role
Who the AI acts as
Task
One clear job
Context
Account, history, what was promised
Rules
What to avoid
Output format
Email, checklist, table, SOP
Tone
Direct, warm, professional, calm
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Section 2

From Prompt to Workflow

7 minutes
Same progression

Where AI 102 lives

One-off prompt
Saved template
Team process
Reviewed workflow
Automation
AI 101 got you to a saved template. AI 102 is the team process and the reviewed workflow. Automation only comes after the workflow has been tested.
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Definition

A workflow answers six questions

  1. 1What starts it
  2. 2What information is gathered first
  3. 3What the AI is asked to do
  4. 4What the output must look like
  5. 5Who reviews and approves it
  6. 6What happens after approval, and who tracks it
AI 102 for CSC MembersSteve Cooker · Logic Sphere
The shape of it

The AI task is the middle, not the whole thing

Before
  • Trigger
  • Account context gathered
  • Approved knowledge pack
AI task
  • Reusable instructions
  • Defined output standard
After
  • Named reviewer
  • Approved action assigned
  • Follow-up tracked
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Same concepts, next step

What makes this AI 102

Context
Approved knowledge pack
Good prompt
Reusable instructions
Good output
Defined output standard
Human review
Named reviewer and review checklist
Guardrails
Specific boundaries, escalation, and failure tests
Useful experiment
Owned pilot with success criteria
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Section 3

Featured Example: The Missed Mat Service

12 minutes
Featured example

Last time we drafted the reply. Now the process.

Last time, we used AI to draft a customer response. Now let us build the process around it.
Illustrative training example. Not an actual customer case.
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Where AI 101 ended

A good draft is not a finished job

AI 101 result

A clear, calm customer email drafted in 30 seconds. Accurate tone. No overpromising.

Still missing

Who checks the account history? Who schedules the recovery service? Who confirms it happened?

Illustrative training example. Not an actual customer case.
AI 102 for CSC MembersSteve Cooker · Logic Sphere
The workflow

Steps 1 to 3: intake and context

1 · System
Customer issue received
2 · Human
Relevant account context gathered
3 · AI
Issue categorized
Illustrative training example. Not an actual customer case.
AI 102 for CSC MembersSteve Cooker · Logic Sphere
The workflow

Steps 4 to 6: the AI task

1 · AI
Missing information identified
2 · AI
Customer response drafted
3 · AI
Internal note prepared
Illustrative training example. Not an actual customer case.
AI 102 for CSC MembersSteve Cooker · Logic Sphere
The workflow

Steps 7 to 9: review and action

1 · Human
Human reviews
2 · Human
Approved task assigned
3 · System
Follow-up tracked
Illustrative training example. Not an actual customer case.
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Featured example

The missed mat workflow, end to end

1 · System
Customer issue received
2 · Human
Relevant account context gathered
3 · AI
Issue categorized
4 · AI
Missing information identified
5 · AI
Customer response drafted
6 · AI
Internal note prepared
7 · Human
Human reviews
8 · Human
Approved task assigned
9 · System
Follow-up tracked
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Critical distinction

Proposed actions are not completed actions

Proposed by AI
  • Draft reply, not sent
  • Suggested credit, not issued
  • Recommended recovery service, not scheduled
  • Suggested route note, not added
Completed by people and systems
  • Reply sent by the reviewer
  • Credit approved and issued
  • Service scheduled with a date and driver
  • Delivery confirmed on the route
Illustrative training example. Not an actual customer case.
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Reusable instructions

What the saved instruction block contains

Instruction block

Role: customer service lead for a mat rental service. Task: categorize the issue, list missing information, draft a customer reply and an internal note. Context: account type, service day, recent service history, what was already promised. Rules: no credits, no pricing, no dates unless supplied. Flag anything you had to assume. Output: three labeled sections. Tone: calm, direct, accountable.

Output standard

1. Category and confidence 2. Missing information list 3. Customer reply, under 150 words 4. Internal note with proposed action only 5. Assumptions flagged at the bottom

AI 102 for CSC MembersSteve Cooker · Logic Sphere
Named reviewer

The review checklist for this workflow

Facts match the account record
No dates, credits, or pricing that were not approved
Assumptions are flagged and resolved
Proposed action is clearly labeled as proposed
Tone fits the customer relationship
Owner and due date are set before sending
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Section 4

Supporting Workflows: SOPs and Knowledge

7 minutes
Supporting example 1

SOP creation, built into a process

  1. 1Manager explains the process out loud
  2. 2AI turns it into a draft SOP in the standard format
  3. 3AI lists steps it could not verify
  4. 4Department lead reviews and corrects
  5. 5AI creates the checklist and quiz
  6. 6SOP is published, dated, and owned
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Supporting example 2

Context becomes an approved knowledge pack

What goes in
Approved SOPs, service rules, FAQ answers, standard language
What stays out
Customer data, employee records, contract terms, credentials
Who owns it
One named person per department, reviewed on a set date
How it is used
Pasted or attached as the context block for every workflow
How it stays good
Wrong answer means fix the source, not just the prompt
How you know
Every answer can be traced to an approved document
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Supporting example 2

Internal knowledge, built into a process

  1. 1Collect approved SOPs, FAQs, and service rules
  2. 2Remove anything confidential
  3. 3Organize by department and label each document
  4. 4Employees ask questions in plain language
  5. 5Wrong answers are logged against the source document
  6. 6Owner updates the source on a set schedule
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Section 5

Choosing a Workflow: Value, Complexity, Risk

8 minutes
Step 1

Value versus complexity, same as AI 101

High value / low complexity
Start here
High value / high complexity
Plan carefully
Low value / low complexity
Optional
Low value / high complexity
Avoid
Criteria
  • Repeats often
  • Takes real time today
  • Has clear inputs and a clear output
  • A human can review it
  • Mistakes are manageable
  • The team will actually use it
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Step 2

Then run the risk, data, and approval check

Risk
What is the worst realistic mistake, who sees it, and how fast can we fix it?
Data
What information does this need, and is any of it confidential or personal?
Approval
Who is the named approver, and are they available when the work happens?
Passes the matrix and clears all three checks. That is your pilot. Anything else waits.
AI 102 for CSC MembersSteve Cooker · Logic Sphere
From cautions to guardrails

Three things every workflow needs in writing

Boundaries
  • Never commits a date
  • Never issues a credit
  • Never quotes pricing
Escalation
  • Contract disputes to the account manager
  • Safety issues to operations
  • Anything unclear stops
Failure tests
  • Angry vague email
  • Wrong account number
  • Two issues in one message
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Tool flexibility

The method travels. The tool is your choice.

  • Today's demonstration happens to use ChatGPT
  • The same workflow design works in any approved assistant your company provides
  • Check which tools your company approves before you use one
  • If a tool is restricted, that restriction is part of your workflow, not an obstacle
  • Design the workflow on paper first. It will outlive whichever tool you pick.
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Section 6

Implementation: Making It Real

8 minutes
Implementation

Turn the experiment into an owned pilot

One workflow
The one that passed the matrix and all three checks
One owner
A named person, not a committee and not the IT department by default
A time box
Four to six weeks, with a decision date on the calendar
Success criteria
Agreed before you start. Time saved, consistency, or response speed.
A small group
Two or three people who actually do the work today
A written record
What worked, what failed, what you changed
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Implementation

How to stand up one workflow

  1. 1Write the workflow on one page, before any tool
  2. 2Name the owner and the reviewer
  3. 3Build the approved knowledge pack
  4. 4Save the reusable instructions and the output standard
  5. 5Run the failure tests with real past examples
  6. 6Pilot with a small group, log every correction
  7. 7Decide on the date: keep, change, or stop
  8. 8Only then consider automating a step
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Optional next step

Automate last, and only what you have tested

Automating a workflow nobody has run manually
Removing the human review step to save time
Letting the system send customer messages without approval
Automating before the knowledge pack has an owner
Automation makes a proven workflow faster. It makes an unproven workflow fail at scale.
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Section 7

Wrap-Up and Q&A

5 minutes
Resource

Other workflows worth building

Par levels and inventory
Par review proposals, shortage explanations, size and garment change requests
Production and PPOH
Daily PPOH recap, shift variance explanation, downtime note cleanup
RFID and asset tracking
No-read follow-up, scan exception triage, garment retirement lists
Routes and delivery
Route profitability review, stop sequence proposals, route handoff packets
CRM and pipeline
Field lead intake cleanup, account note summaries, stale deal nudges, pipeline review prep
Service, sales, plant, admin
Complaint response, quote follow-up, SOP drafting, training outlines
Where to find them
Full list with details on the AI 102 resource page. Pick one, not five.
Illustrative training example. Not an actual customer case.
AI 102 for CSC MembersSteve Cooker · Logic Sphere
Action list

What to do in the next 30 days

Pick one workflow that passes value, complexity, risk, data, and approval
Write it on one page with a before, an AI task, and an after
Build a small approved knowledge pack for it
Name the owner and the reviewer, and write the six-line review checklist
Run it manually for two weeks and log every correction
AI 102 for CSC MembersSteve Cooker · Logic Sphere
The takeaway

AI is a tool. The workflow is the strategy.

The prompt formula defines the AI task. The workflow defines everything around it.

AI 102 for CSC MembersSteve Cooker · Logic Sphere
CSC Network
Questions

Let's map your workflow

steve@logicsphere.ai703-870-8429linkedin.com/in/steve-cooker
Thank you to CSC Network for hosting.