Interactive workflow diagnostic
One Tuesday, twice.
Meridian Facility Services is a fictional 18-person service company handling about 45 field jobs a week. Bookings, changes, quote requests, and issues arrive throughout each day.
Follow the same 56 requests through two versions of the workflow: first through the original setup, then after six changes.
Run the three-part experience to feel the queue, add each change, and compare both workflows under the same demand.
The old Tuesday
Requests arrive through email threads, phone calls, texts, walk-ins, and sticky notes. Nobody owns intake, so everyone does it a little.
A request gets answered twice or not at all. Finding the status of a job means finding the person who last touched it.
Across the nine-and-a-half-hour day, 56 requests arrive while the operator works continuously:
| Measure | Old Tuesday |
|---|---|
| Requests arrived | 56 |
| Handled | 30 |
| Forgotten entirely | 3 |
| Carried into tomorrow | 23 |
| Duplicate requests | 11 |
| Angry customer calls | 10 |
| Average wait for an answer | 2h 28m |
| Operator time on coordination | 9.3 hours |
Even with the operator working continuously, the old workflow spends most of the day finding, re-keying, forwarding, and confirming requests.
Six changes to the workflow
Each change removes one source of repeated work, delay, or uncertainty from the day.
Week 1 — One front door
Every request goes into one intake form, no matter what channel it started in.
Five buckets become one queue, and the customer gets an answer right away: got it, you’re in the system. That confirmation removes most of the modeled duplicate requests.
Modeled duplicate rate: 12% to 2%.
Week 3 — One record per job
The spreadsheet, the whiteboard, and the sticky notes become one job record. Everyone reads the same record, and nobody re-types anything.
Watch the loose pieces on the map turn into one shared record.
Modeled misroute rate: 18% to 8%.
Week 5 — Assignment rules
New jobs route to the right crew member by zone and skill.
The dispatcher stops assigning work by memory and handles the exceptions.
The model routes jobs in about 25 minutes without manual coordination.
Week 7 — The customer hears first
Confirmations, arrival windows, and running-late notices go out automatically.
In the model, most angry calls come from customers not knowing what is happening.
Modeled complaints fall by more than 80%.
Week 9 — AI reads the inbox
Incoming requests get classified, key fields extracted, and replies drafted. A person confirms; the software does the typing.
The model includes classification mistakes, so every automated decision stays one click away from a person.
Modeled manual handling: 18 minutes to 4 minutes per request.
Week 11 — The weekly number
One screen shows the week: volume, response time, and stalled work.
Dropped requests used to be invisible. The modeled dashboard makes every stalled job visible.
Every modeled stalled job has an owner and a timer.
The same 56 requests, twice
Both runs receive the same 56 requests, and both operators work continuously. Only the workflow changes:
| Measure | Old Tuesday | New Tuesday |
|---|---|---|
| Handled | 30 | 53 |
| Forgotten entirely | 3 | 0 |
| Carried into tomorrow | 23 | 3 |
| Duplicate requests | 11 | 1 |
| Angry customer calls | 10 | 2 |
| Average wait for an answer | 2h 28m | 14 minutes |
| Operator time on coordination | 9.3 hours | 0.0 hours |
The redesigned workflow handles 23 more requests, carries 20 fewer into tomorrow, and spends 9.3 hours less on coordination.
The stress test
Double the requests, take the dispatcher out, and disable the scheduling API:
| Under stress | Old Tuesday | New Tuesday |
|---|---|---|
| Requests arrived | 94 | 94 |
| Forgotten entirely | 46 | 0 |
| Carried into tomorrow | 32 | 70 |
| Angry customer calls | 26 | 9 |
What drives the simulation
- Demand
- ~5 requests/hour; this run generates 56 requests
- Rush windows
- ×1.6 at 9 AM and 1 PM
- Manual handling
- 18 min/request
- Misroute rate
- 18% before, 4% after
- Duplicate requests
- 12% before, 2% after
- Forgotten after
- 4 hours untouched
- Auto-dispatch
- jobs route in ~25 min, 14 with triage
- Crew capacity
- ~65 requests/day
- Company
- Meridian Facility Services (fictional), 18 people
Which of these problems show up in your business?
Answer Yes or No for each statement.
Your workflow has fewer warning signs.
The answers suggest that coordination is not the first problem to investigate. Keep the pattern in view as the business changes.
A workflow review may be useful.
Several handoffs depend on memory or repeated work. A short audit can show whether process, tooling, or capacity is the real constraint.
Your answers point to operational risk.
These patterns can lead to missed requests and fragile handoffs. The right response may be a process change, software, more capacity, or a combination; the next step is to inspect the real workflow.
If these problems show up in your business, start with the workflow causing the most delay or repeated work.