Workflow template for customer support and service
Follow-ups promised on service calls
After service calls, the contact-center team leader gets every follow-up the agent promised, each quoted from its transcript line, with the deadline as said, any condition, who owns it and the record that matches it. Nothing is booked, assigned or sent, and no customer is contacted.
Listed under
Add it to a workflow in the Fully Powered Planner, beside the teams and roles that will run it. Book a fit call
The outcome
The team leader sees what customers were told they would get, in the agent's own words, before a customer has to ring back and ask. Every promise points to the line of the transcript where it was said, so anyone can check it against the call. A callback "by Thursday" shows the Thursday it means. A refund that "should" arrive is marked as hedged. "We'll sort that out" shows that nobody was named to sort it out.
Each promise is set beside the follow-up records your systems hold: a callback booked, a refund raised, a replacement sent. Where the record is due later than the customer was told, or there is no record at all, the list says so. Where the deadline as read has passed and no record shows the work done, the list says that too, and says plainly that this is not proof the promise was broken.
Lines that only describe how things usually go ("replacements normally go out within five working days") are kept apart from promises, because customers often hear them as one. Actions the agent finished while still on the call are kept apart as well.
The list ends with one question for each promise that has a gap. The team leader decides who owns each follow-up, by when, and whether a promise beyond the agent's limits stands. The AI assistant contacts no customer, books and assigns nothing, and changes no record. It describes calls, never people, and nothing on it is for an agent's file.
Who starts the work, and when
The contact-center team leader, or a quality analyst working for the team leader, starts it on request: on a day's calls for one agent, on calls picked for review, on calls about a known problem, or before answering a customer who says they were promised something. The calls must have ended and have a transcript with numbered lines.
The team leader reads the list, confirms or corrects each promise, deadline and owner, and settles each question. The people who own each follow-up then book or correct it in your own systems, and contact the customer where the team leader decides they should.
What they supply
- The call list. One row per call: the call's reference, date, start time, agent, your customer reference, the case number and the agent's wrap-up note.
- The transcripts. One row per spoken line, numbered, with the time into the call and who was speaking.
- The follow-up records, where you can export them: callbacks, refunds, deliveries, visits and tasks, each with the call or case it belongs to, its type, owner, due date, when it was created and its status.
- The run settings. Your working days and hours, the labels your transcription uses for agent and customer, who owns each kind of follow-up when the call names nobody, the names of your record types and statuses, and any limits on what an agent may promise without approval. The template carries sensible defaults for anything you have not set.
What comes out
One list in plain text, call by call, in a fixed order:
- A header with the counts: calls read, promises found, how many are hedged and flagged, and which settings were taken as defaults.
- Each promise, numbered, with its transcript line and time, its kind (callback, send, refund or credit, replacement or delivery, visit, escalation or task), the words, the deadline as said and the date and time it reads as with the rule used, any condition, the owner as said, the matching record, and whether the wrap-up note mentions it.
- Fixed flags where something is missing or out of line: owner missing, deadline not stated, hedged, conditional, needs approval under your limits, no record found, more than one record, record later than promised, deadline passed, outside working hours, speaker unclear.
- Expectations set, not promises, and completed during the call, for each call.
- Questions for the team leader, one for each flagged promise, the facts first.
- Could not be read, for any call or line that could not be used.
Good looks like this: the team leader can check every line against the transcript line it cites, no date appears that the call did not support, and nothing has been changed anywhere.
An invented example
Illustrative example with invented data. The company, its customers, its agents, the calls, the records, the dates and the amounts below are all invented. An invented online retailer of kitchen appliances runs a phone queue for orders and deliveries. On Thursday morning the queue's team leader picks two of Monday afternoon's calls for review, one taken by Maya and one by Theo. The team leader exports the call list, the transcripts and the follow-up records, and runs the workflow template.
The inputs come first, in full, so the template can be run on them. The list it returned follows.
The inputs
1. Run settings
Team: Orders and deliveries phone queue
For: Team leader, Orders and deliveries
Run at: Thursday 12 November 2026, 09:30
Working days and hours: Monday to Friday, 09:00 to 17:00
Speaker labels: Agent, Customer
Owner map:
Callback, Send, Task: the agent on the call
Refund or credit: Billing team
Replacement or delivery: Fulfillment team
Visit: Field scheduling
Escalation: Team leader
Record types (kind: type):
Callback: Callback; Send: Send; Refund or credit: Refund;
Replacement or delivery: Delivery; Visit: Visit;
Escalation: Escalation; Task: Task
Done statuses: Done, Closed
Promise limits: agents may offer a goodwill credit of up to $20 without
approval. Anything above $20 needs the team leader's approval.
2. Call list
Call ID,Date,Start,Agent,Customer,Case,Wrap-up note
C-4471,2026-11-09,14:05,Maya,Customer A,88213,"Dishwasher delivered with dented door. Customer emailing photos tonight. Replacement once photos received. Callback booked for Thursday. Call summary emailed."
C-4486,2026-11-09,16:40,Theo,Customer B,88240,"Refund for returned kettle not received. Raised with billing, TK-5531. Refund due within 48 hours."
3. Transcripts
Call ID,Line,Time,Speaker,Text
C-4471,1,00:00,Agent,"Thanks for calling, this is Maya. How can I help?"
C-4471,2,00:05,Customer,My new dishwasher arrived this morning with a big dent in the door.
C-4471,3,00:14,Agent,I'm sorry about that. Could you send us a couple of photos of the dent?
C-4471,4,00:22,Customer,"Yes, I can email them tonight."
C-4471,5,00:27,Agent,"Thank you. Once the photos are in, we'll arrange a replacement for you."
C-4471,6,00:36,Customer,How long does that usually take?
C-4471,7,00:40,Agent,Replacements normally go out within five working days.
C-4471,8,00:48,Customer,I had to take the morning off work for that delivery.
C-4471,9,00:55,Agent,I understand. I'll put a $25 credit on your account for the trouble.
C-4471,10,01:03,Customer,Thank you. Will someone let me know when the new one is coming?
C-4471,11,01:09,Agent,"Yes, I'll call you back by Thursday to confirm the delivery date."
C-4471,12,01:17,Customer,Could I have all of this in writing as well?
C-4471,13,01:21,Agent,Of course. I'll email you a summary of today's call by the end of the day.
C-4471,14,01:29,Customer,"Great, thanks."
C-4471,15,01:32,Agent,Thanks for calling. Goodbye.
C-4486,1,00:00,Agent,You're through to Theo. How can I help?
C-4486,2,00:04,Customer,I sent a kettle back three weeks ago and I still haven't had my refund.
C-4486,3,00:12,Agent,Let me look. The return reached our warehouse on 23 October.
C-4486,4,00:31,Agent,I'm going to raise this with our billing team now.
C-4486,5,01:02,Agent,"Right, that's raised with billing, reference TK-5531."
C-4486,6,01:08,Customer,So when do I get my money?
C-4486,7,01:12,Agent,The refund should be with you within 48 hours.
C-4486,8,01:18,Customer,And if it isn't?
C-4486,9,01:21,Agent,"If it's not there by then, I'll email you myself with an update."
C-4486,10,01:28,Customer,"OK, thanks."
C-4486,11,01:30,Agent,Thanks for calling. Goodbye.
4. Follow-up records, exported Thursday 12 November 2026, 09:15
Record,Call ID,Case,Type,Owner,Due,Created,Status,Description
CB-3099,C-4460,88190,Callback,Maya,2026-11-10,2026-11-09 11:02,Done,Call back about warranty registration
CB-3120,C-4471,88213,Callback,Maya,2026-11-13,2026-11-09 14:19,Open,Call to confirm replacement delivery date
SE-2208,C-4471,88213,Send,Maya,2026-11-09,2026-11-09 14:22,Done,Email summary of call
TK-5531,C-4486,88240,Task,Theo,2026-11-09,2026-11-09 16:41,Done,Raise refund delay with billing
RF-7702,,88240,Refund,Billing team,2026-11-13,2026-11-09 16:52,Open,Refund for returned kettle
The output
The list as the workflow template returned it, in plain text.
FOLLOW-UPS PROMISED ON SERVICE CALLS: Orders and deliveries phone queue, run of Thursday 12 November 2026, 09:30
For: Team leader, Orders and deliveries. Calls read: 2 (C-4471, C-4486). Agents: Maya, Theo.
Promises found: 6 (5 firm, 1 hedged). Flagged: 5. Expectations set, not promises: 1. Completed during the call: 1.
Deadlines read using: Monday to Friday, 09:00 to 17:00 (your settings).
Follow-up records: 5 supplied, exported Thursday 12 November 2026, 09:15; used for these calls: 4; for other calls, not used: 1.
Promise limits: agents may offer a goodwill credit of up to $20 without approval.
Settings taken as default: none.
CALL C-4471 | Monday 9 November 2026, 14:05 | Agent: Maya | Customer: Customer A | Case: 88213
P1 | Line 5, 00:27 | Replacement or delivery | firm
- Words: "Once the photos are in, we'll arrange a replacement for you."
- Deadline as said: not stated
- Condition: "Once the photos are in" (line 5); depends on the customer; not met by the end of the call; customer's own timing: "Yes, I can email them tonight." (line 4)
- Owner as said: not stated on the call
- Proposed owner from your owner map: Fulfillment team.
- Follow-up record: none found
- Wrap-up note: mentions it
- Flags: OWNER MISSING, DEADLINE NOT STATED, CONDITIONAL, NO RECORD FOUND
- Question 1 below.
P2 | Line 9, 00:55 | Refund or credit | firm
- Words: "I'll put a $25 credit on your account for the trouble."
- Deadline as said: not stated
- Condition: none
- Owner as said: Maya ("I'll")
- Follow-up record: none found
- Limit: a $25 credit; your limit is $20 without approval.
- Wrap-up note: does not mention it
- Flags: DEADLINE NOT STATED, NEEDS APPROVAL UNDER YOUR LIMITS, NO RECORD FOUND
- Question 2 below.
P3 | Line 11, 01:09 | Callback | firm
- Words: "I'll call you back by Thursday to confirm the delivery date."
- Deadline as said: "by Thursday", read as Thursday 12 November 2026, 17:00 (rule 4: the first Thursday after the call date, at the end of working hours)
- Condition: none
- Owner as said: Maya ("I'll")
- Follow-up record: CB-3120, Callback, owner Maya, due 13 November 2026, status Open
- Wrap-up note: mentions it
- Flags: RECORD LATER THAN PROMISED
- Question 3 below.
P4 | Line 13, 01:21 | Send | firm
- Words: "I'll email you a summary of today's call by the end of the day."
- Deadline as said: "by the end of the day", read as Monday 9 November 2026, 17:00 (rule 2: the call date, at the end of working hours)
- Condition: none
- Owner as said: Maya ("I'll")
- Follow-up record: SE-2208, Send, owner Maya, due 9 November 2026, status Done
- Wrap-up note: mentions it
- Flags: none
Expectations set, not promises:
- Line 7, 00:40: "Replacements normally go out within five working days."
Completed during the call:
None.
CALL C-4486 | Monday 9 November 2026, 16:40 | Agent: Theo | Customer: Customer B | Case: 88240
P5 | Line 7, 01:12 | Refund or credit | hedged
- Words: "The refund should be with you within 48 hours."
- Deadline as said: "within 48 hours", read as Wednesday 11 November 2026, 16:40 (rule 6: 48 clock hours after the call's start)
- Condition: none
- Owner as said: not stated on the call; record RF-7702 names Billing team
- Follow-up record: RF-7702, Refund, owner Billing team, due 13 November 2026, status Open
- Wrap-up note: mentions it
- Flags: HEDGED, RECORD LATER THAN PROMISED, DEADLINE PASSED
- No record supplied shows it done. This is not proof it was broken.
- Question 4 below.
P6 | Line 9, 01:21 | Send | firm
- Words: "If it's not there by then, I'll email you myself with an update."
- Deadline as said: not stated
- Condition: "If it's not there by then" (line 9); depends on a time, read as Wednesday 11 November 2026, 16:40 (rule 6, as for line 7: 48 clock hours after the call's start); not met by the end of the call
- Owner as said: Theo ("I'll")
- Follow-up record: none found
- Wrap-up note: does not mention it
- Flags: DEADLINE NOT STATED, CONDITIONAL, NO RECORD FOUND
- Question 5 below.
Expectations set, not promises:
None.
Completed during the call:
- Line 4, done at line 5: raised with the billing team, reference TK-5531 said on the call.
QUESTIONS FOR THE TEAM LEADER
1. C-4471 P1: Maya said a replacement would be arranged once the customer's photos are in. The call names no owner and no date, the photos had not arrived by the end of the call, and no delivery record was found. Your owner map proposes the Fulfillment team. Have the photos arrived since, and who arranges the replacement, by when?
2. C-4471 P2: Maya offered a $25 credit; your limit for agents is $20 without approval. No date was given, no refund or credit record was found, and the wrap-up note does not mention it. Do you approve the credit, and who applies it, by when?
3. C-4471 P3: Maya promised a callback by Thursday 12 November 2026, read as 17:00; record CB-3120 is due Friday 13 November 2026. Does the callback move to Thursday, or is the customer told of the later day?
4. C-4486 P5: Theo said the refund should be with the customer within 48 hours, read as Wednesday 11 November 2026, 16:40. Record RF-7702, owned by the Billing team, is due Friday 13 November 2026 and is still Open. Has the refund been paid, and is the customer told of the later date?
5. C-4486 P6: Theo said he would email the customer himself if the refund had not arrived by then. That time, read as Wednesday 11 November 2026, 16:40, has passed and RF-7702 is still Open; no send record was found and the wrap-up note does not mention the email. Has the refund arrived since, and if not, does Theo send the email, and by when?
COULD NOT BE READ
None.
Prepared for the team leader. Nothing has been booked, assigned, changed or sent.
No customer has been contacted. This list is not proof a promise was kept or broken.
How the figures were worked out
- Promises: six. Lines 5, 9, 11 and 13 of C-4471 and lines 7 and 9 of C-4486. Five are firm; P5 is hedged by "should". Line 7 of C-4471 says how replacements "normally" go, so it is an expectation. Line 4 of C-4486 was done on the call, at line 5. Line 3 of C-4471 and line 3 of C-4486 are a request and a statement about the past.
- Flagged: five, every promise but P4.
- Records: five supplied. CB-3099 belongs to call C-4460, which is not in this run. The other four belong to these calls: CB-3120 and SE-2208 by Call ID, TK-5531 by Call ID (the action done on the call), and RF-7702 by its case, 88240, since its Call ID is blank. Each was created after its call started (14:19 and 14:22 after 14:05; 16:41 and 16:52 after 16:40).
- P3: Monday 9 November to the first Thursday after it is Thursday 12 November, read as 17:00. CB-3120 is due Friday 13 November, read as 17:00, one day later. Thursday 17:00 is after the run at Thursday 09:30, so the deadline has not passed.
- P4: the call date, Monday 9 November, read as 17:00. SE-2208 is due the same day and is Done.
- P5: 16:40 on Monday 9 November plus 48 clock hours is 16:40 on Wednesday 11 November, a working day within working hours. RF-7702 is due Friday 13 November, later than promised, and is Open, not a done status. Wednesday 16:40 is before the run at Thursday 09:30, so the deadline has passed.
- P2: a $25 credit against a $20 limit is $5 over, so it needs approval.
- P6: the condition's time is P5's reading. The email itself has no deadline.
Nothing was booked, assigned, changed or sent; the team leader settles each question in the company's own systems.
Where it fits
- Customer support and service: the contact center's check on what its agents have promised, before the promises turn into repeat calls.
- Insurance: claims and policy lines where an agent promises a callback, a document or a decision by a date the customer remembers.
- E-commerce and retail: replacements, refunds and redeliveries promised on the phone and carried out by other teams.
- Energy: engineer visits, meter readings and bill corrections promised on a call and owned by field and billing teams.
- Growth & scaling: the phone queue has grown past the point where the team leader hears enough calls to know what is being promised.
- Established businesses: long-standing habits of saying "we'll sort it" that nobody owns.
It is separate from the account commitments check, which lists what both sides promised across an account's meetings and emails, and from the stale ticket sweep, which looks at the whole queue of open tickets. When one follow-up needs a specialist, the support escalation brief carries it on.
What has to change for your company
- Your working days and hours, so that "by the end of the day" and "within 48 hours" read as the times your team means.
- Your speaker labels, exactly as your transcription service writes them, including any label for a second agent after a transfer.
- Your owner map: which team owns each kind of follow-up when the call names nobody.
- Your record types and statuses, exactly as your case, callback and billing systems export them.
- Your promise limits, such as the largest credit an agent may offer without approval.
- The exports: their column headings, and which customer details are removed before they leave your systems.
- Your platform. It runs in ChatGPT, Claude, Copilot or Gemini, from files you supply; it needs no connection to your phone or case systems.
What has been checked
Fully Powered AI wrote this workflow template. It was editorially reviewed on 11 November 2026. It was tried on its invented example in Claude on 11 November 2026, and the record of that run is published with it. It has not yet been tested in a client environment. It has not yet been tested as a packaged version in ChatGPT, Claude, Copilot or Gemini.
Testing in your environment means running it on calls your team leader has already listened to, plus hard cases written in advance, each with its expected result: a promise with vague timing, a callback booked for a later day than the one promised, a line from a speaker the transcription could not name, a promise the agent took back two lines later, a credit above the agent's limit. The team leader checks every line against the transcript before anyone relies on the list. Pilot acceptance-test preparation sets those cases up, and what we test before we call an AI workflow ready describes the record we keep.
Who decides
Your people keep
- Who owns each follow-up, and by when
- Whether a promise beyond the agent's limits stands
- Whether the customer is contacted
It does not
- Contacts no customer
- Books and assigns nothing
- Changes no record
- Nothing on it is for an agent's file
Questions people ask
Who is it for?
The contact-center team leader, or a quality analyst working for the team leader. The calls must have ended and have a transcript with numbered lines.
What do I need to supply?
The call list, the transcripts, the follow-up records where you can export them, and the run settings.
Does it contact customers or book callbacks?
No. The AI assistant contacts no customer, books and assigns nothing, and changes no record. The team leader decides, and the owners act in your own systems.
Does it say whether a promise was broken?
No. Where the deadline as read has passed and no record shows the work done, the list says so, and says plainly that this is not proof the promise was broken.
What has been checked?
Fully Powered AI wrote this workflow template and it was editorially reviewed. It was tried on its invented example in Claude on 11 November 2026, and the record of that run is published with it. It has not yet been tested in a client environment.
The full workflow template
Everything above is the outline. The complete workflow template is the working version: the roles, the inputs, the procedure, the exceptions and stop conditions, the output format, the reviewer checklist, the acceptance tests, the instructions for an AI assistant and what to customize for your company. It comes with a download and the link to its folder in Fully Powered’s public repository.
Free with an account. Sign in by email: no password and no card.
Next step
Put this workflow template to work for your team: on your plan, or built with you in a sprint.
Build this in a sprint
In a sprint, Fully Powered adapts this workflow template to your rules, proves it on your own work and puts it in the hands of the roles that own it.
Two ways to begin: