CSC Network WebinarCSC Network

AI 102 Resources

From good prompts to repeatable workflows. Steve Cooker, Founder, Logic Sphere.

Six workflow design questions

  • What starts the work?
  • What information is needed before the AI task?
  • What exactly is the AI asked to do?
  • What does a good output look like?
  • Who reviews it, and against what checklist?
  • What happens after approval, and how is it tracked?

The prompt formula defines the AI task. The workflow defines what happens before and after it.

From AI 101 to 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

Missed mat service workflow

  • 1Customer issue received
  • 2Relevant account context gathered
  • 3Issue categorized
  • 4Missing information identified
  • 5Customer response drafted
  • 6Internal note prepared
  • 7Human reviews
  • 8Approved task assigned
  • 9Follow-up tracked

Illustrative training example. Not an actual customer case. Always separate proposed actions from actions actually completed.

Review checklist

  • Facts match the account record
  • Nothing is promised that we cannot deliver
  • No pricing or contract commitments
  • Tone fits the relationship
  • Next step and owner are clear
  • Proposed actions are marked as proposed, not completed

Guardrails

  • Boundaries: what the workflow may never decide on its own
  • Escalation: who gets it when the case is outside the boundaries
  • Failure tests: cases you run on purpose to see how the workflow breaks

Approved knowledge pack

  • In: approved service descriptions, standard policies, common issue resolutions
  • In: tone and language your team already uses with customers
  • Out: confidential customer data and private contract terms
  • Out: employee records and anything HR sensitive
  • One named owner keeps it current

Pilot plan

  • One workflow, one owner
  • A small set of real cases
  • A defined review step before anything goes out
  • Success criteria agreed before you start
  • A date to review results and decide: keep, change, or stop
Open the AI 102 deck

More repeatable workflow ideas

The missed mat service example is one workflow. Here are others that fit the same pattern: a clear trigger, gathered context, a defined AI task, a named reviewer, and a tracked follow-up. Start with one, not five. Illustrative examples for training, not actual customer cases.

Customer service and accounts

  • Service complaint response
    Issue received, account context gathered, response drafted, internal note prepared, reviewed, task assigned, follow-up tracked.
  • Weekly account health summary
    Recent service issues and credits summarized for each account manager, reviewed before the Monday meeting.
  • New customer welcome packet
    Service schedule, item list, and contact details turned into a clear welcome letter, checked against the signed agreement.
  • Credit and adjustment request write-up
    A credit request turned into a consistent write-up with the reason, the amount proposed, and the backup named, held as proposed until a manager approves.
  • Service cancellation save attempt
    A cancellation notice paired with recent service history so the account manager gets a draft save plan and talking points before the call.

CRM and pipeline

  • Field lead intake cleanup
    A lead sent in from the route or the plant floor turned into a clean CRM record with company, contact, location, opportunity type, and a suggested next step for the rep to approve. Field capture tools like Wingmate are where this starts.
  • CRM note and activity summary
    Call notes, emails, and visit logs on an account summarized into a short status the sales manager can read before a pipeline meeting.
  • Stale opportunity nudge
    Opportunities with no activity for a set number of days flagged, with a draft follow-up message and a proposed next action the owner approves.
  • Pipeline review prep
    This week's deals grouped by stage and risk, with the missing information named for each one so the review is about decisions, not data entry.
  • Win and loss write-up
    Closed deals turned into a short, consistent record of why it was won or lost so the pattern shows up across the branch.

Route and operations

  • Route stop notes cleanup
    Driver notes turned into consistent, readable service notes with a flag when something needs a supervisor.
  • Missed or short delivery follow-up
    Shortage logged, cause categorized, customer message drafted, make-up delivery proposed for approval.
  • Daily exception recap
    Yesterday's exceptions grouped by cause so the operations meeting starts with the pattern, not the list.
  • Service schedule change request
    A customer request to move a day or frequency turned into a clean change request with the route impact named, approved by the route manager before anything is entered.
  • Holiday and shutdown service notice
    Route-by-route schedule changes turned into clear customer notices, checked by the service manager before they are sent.

Sales and quoting

  • Quote follow-up sequence
    Draft follow-up notes after a quote, with pricing and terms pulled only from the approved quote document.
  • RFP and bid question drafting
    Standard answers pulled from an approved response library, gaps flagged for a human to fill.
  • Win and loss notes
    Short structured summary after each decision so patterns are visible by quarter.
  • Prospect research brief
    Public information on a prospect pulled into a one-page brief with likely garment, mat, and linen needs and three questions for the first call.
  • Price increase letter drafting
    An approved increase turned into a clear, consistent customer letter per account, with the amount and effective date confirmed by a human before sending.

Plant, quality, and safety

  • SOP drafting and refresh
    A supervisor describes the task, AI drafts the SOP in your standard format, the supervisor edits and approves.
  • Quality inspection summary
    Inspection findings grouped by cause with a suggested corrective action for the plant manager to approve.
  • Safety incident write-up
    Facts collected in a set order, draft prepared for the safety lead, never filed without review.
  • Rewash and reject cause summary
    Rewash and reject tickets grouped by likely cause and shift so the plant meeting starts with one or two things to fix.
  • Maintenance work order cleanup
    Short operator notes turned into clear work orders with the machine, the symptom, and the urgency, reviewed by maintenance before scheduling.

Admin, HR, and training

  • Job posting and interview guide
    Role description turned into a posting and a consistent set of interview questions.
  • New hire training outline
    Approved SOPs turned into a week-one training plan for a specific role.
  • Internal knowledge answers
    Common questions answered from the approved knowledge pack, with an owner who keeps it current.
  • Policy and handbook plain language rewrite
    An approved policy rewritten at a reading level the floor actually uses, with HR confirming nothing changed in meaning.
  • Toolbox talk and shift huddle notes
    A short weekly talk drafted from recent incidents and quality issues, reviewed by the safety lead before it is read out.

Inventory and par levels

  • Par level review
    Usage, soil return, and delivery counts for an account pulled into one summary, AI proposes a par adjustment with the reason, the account manager approves or rejects it.
  • Shortage and overstock explanation
    When an account runs short or sits on excess, AI drafts the likely causes in plain language for the service manager to confirm before the customer call.
  • Size mix and garment issue requests
    Wearer change requests turned into a clean add, delete, and size change list checked against the account's approved garment program.
  • Seasonal par adjustment plan
    Accounts with seasonal swings flagged ahead of the change, with a proposed par plan per account for review.

Production and labor (PPOH)

  • Daily PPOH recap
    Pounds processed and hours worked by department turned into a short PPOH readout with the two biggest drivers of the day, reviewed by the plant manager.
  • Shift variance explanation
    When PPOH drops against target, AI drafts likely causes from downtime notes, staffing, and volume mix, and the supervisor confirms or corrects before it goes out.
  • Downtime note cleanup
    Maintenance and operator notes turned into consistent downtime categories so the weekly pattern is readable.
  • Weekly production summary
    Poundage, PPOH, reruns, and rewash grouped by department with one suggested focus area for next week.

RFID and asset tracking

  • Missing and no-read garment follow-up
    Items not scanned back for several cycles listed by account and wearer, with a draft customer message and a proposed loss charge for approval.
  • Scan exception triage
    Read errors grouped into tag failure, process miss, and wrong location, so the fix goes to the right person.
  • Garment life and retirement list
    Wash cycle counts turned into a proposed retirement and replacement list for the account manager to approve.
  • Mat and mop inventory reconciliation
    Scanned inventory compared with the service agreement, gaps explained, adjustments proposed rather than made automatically.

Routes and delivery performance

  • Route profitability review
    Revenue, stops, drive time, and service credits for a route summarized into a short profitability readout with the questions a manager should ask next.
  • Stop sequence and frequency proposals
    Stops that look mistimed or out of order flagged with a proposed change, always routed to the route manager for a decision.
  • Driver ride-along and coaching notes
    Ride-along observations turned into a consistent coaching summary with two specific actions for the next week.
  • Route handoff packet
    Account quirks, access notes, and recent issues for a route pulled into a one-page packet for a covering driver.

Uniform programs and wearers

  • New wearer onboarding packet
    Sizing sheet, garment selections, and start date turned into a clean setup request for the account, checked against the approved garment program before it is entered.
  • Wearer roster reconciliation
    Customer headcount compared with the active wearer list, adds and terminations proposed with the reason, account manager approves before billing changes.
  • Emblem and logo change requests
    Name, logo, and emblem change requests drafted into a single clear work order with the artwork reference and approval still coming from a human.
  • Garment damage and charge explanation
    Damage findings turned into a plain language explanation for the customer with the proposed charge marked as proposed until the service manager approves.

Loss, carts, and unreturned goods

  • Missing cart and container tracking
    Carts, hampers, and containers not returned from an account listed with the last known delivery, a draft customer message, and a proposed replacement charge held for approval.
  • Unreturned garment follow-up
    Terminated wearers with garments still out turned into a clear return request and a proposed loss charge the account manager approves before billing.
  • Mat and linen shrink review
    Delivered versus returned counts for mats, towels, and linen compared over several weeks, with likely causes drafted in plain language for the service manager to confirm.
  • Loss claim packet
    Service history, counts, and prior notices pulled into one packet so the loss conversation with the customer starts with facts, not a guess.

A note on CRM tools

CRM workflows only work when the lead and account information actually gets captured. Many CSC members use Wingmate, a CRM built around field teams sending in leads and service issues from the route. If you already use it, the workflows above sit on top of what your drivers and service reps are already reporting. If you use a different CRM, the same workflow design applies. Use whatever system your company has approved.

Run each idea through the same check before you build it: value versus complexity, then risk, data, and approval. If it fails one of those, it is not your first workflow.