The Meridian centerpiece has the wrong mount. We need AX-17-B here before doors at six. Can you find a way to get the right part and someone to fit it?
An opening-night problem. A plan before doors open.
When plans break,
the show goes on.
A missing installation component threatens an exhibition opening. An operations lead makes one call to explore inventory, crew, and delivery options in the same conversation.
Step inside the storyFICTIONAL BUSINESS · ILLUSTRATIVE USE CASE
CONNECTION / 02
Afterlight ExhibitionsLive events & exhibition operations
PHONE CONVERSATIONGPT-Realtime-2.1
THE BUSINESSAfterlight Exhibitions
THE CONNECTIONGPT-Realtime phone agents
THE MODEL FITGPT-Realtime-2.1
EXPLORE THE POSSIBILITY 01 / THE MOMENT
16:42
Opening day
ONE MOMENT.
A DIFFERENT POSSIBILITY.
Guests arrive at six. The centerpiece is missing a mount.
Jonah is on the exhibition floor when a crate reveals the wrong adapter. The right component may be at the warehouse, but a courier alone will not solve it: the crew must be available and the installation supervisor must approve the revised plan.
- 01
A spoken part number needs to match the exact inventory item.
- 02
A workable plan depends on warehouse, courier, and crew availability together.
- 03
Urgency does not remove the need for approval before a spend or schedule change.
02 / THE CONNECTION, IN MOTION
From a problem on the floor to a coordinated plan.
Step through an imagined internal operations call. Tool checks and actions are mock results, with scripted times rather than performance measurements.
Afterlight ExhibitionsPHONE CONVERSATION / SCRIPTED SIMULATION
PREVIEW
NO LIVE CALLS, MESSAGES, OR BUSINESS ACTIONS
A scripted illustration, not a live agent. Select any moment or play the sequence. All people, businesses, messages, documents, and actions in this story are fictional.
Read the complete story
- Jonah / operations lead16:42 / Understand the constraint
The Meridian centerpiece has the wrong mount. We need AX-17-B here before doors at six. Can you find a way to get the right part and someone to fit it?
- Prodero / operations agent16:43 / Confirm the exact part
Confirming: A-X, seventeen, B. I’ll check the exact part, a delivery window, and the installation team. Is the installation supervisor still Ren?
- Mock operations systems16:44 / Find the dependencies
AX-17-B: two at North Warehouse. Courier estimate: arrival 17:25. Installation crew: next slot 17:30. Ren’s review is required. Courier quote: $85.
- Jonah / operations lead16:45 / Revise around a new constraint
One complication: the loading entrance closes at five. Use the east entrance after that. Don’t commit the courier until Ren has reviewed it.
- Prodero / operations agent16:46 / Present one reviewable plan
Here’s the draft: reserve one AX-17-B, request the quoted courier to the east entrance, and hold the 5:30 crew slot. Ren needs to approve the fit and sequence. Shall I send her the plan? Nothing is booked yet.
- Mock approval workflow16:47 / Prepare the handoff
Jonah approved sharing. Draft sent to Ren for review: exact part, proposed crew slot, courier quote, east entrance instructions, and booking links. No courier purchased; no final crew change made.
03 / A DIFFERENT ENDING
The team gets a plan they can actually approve.
Instead of relaying fragments between several systems, Jonah can review one connected proposal. In a production workflow, approval could trigger the connected bookings and return confirmations to the same conversation.
A part number lost between hurried callsAn identifier confirmed before a lookup
Delivery and installation planned separatelyA proposal that accounts for both dependencies
Urgent decisions without shared contextA supervisor sees the full sequence and quote
An imagined outcome to illustrate the workflow—not a customer result or performance guarantee.
04 / WHY THIS INTELLIGENCE
GPT-Realtime-2.1
Reasoning and tools, inside the voice conversation.
We propose GPT-Realtime-2.1 where the call itself needs to work through a changing operational problem. Its speech interaction, reasoning options, and tools can support this pattern when connected to reliable business systems.
OpenAI · GPT-Realtime-2.1 model documentation (opens in a new tab)- 01
Reason through constraints
Configurable reasoning supports a workflow that must consider several requirements together.
- 02
Use business tools
Function calling can request inventory, availability, and quotes from authorized integrations.
- 03
Handle spoken detail
Alphanumeric recognition and noise handling are relevant to a call from a busy exhibition floor. Identifier confirmation remains part of the flow.
THE CONNECTION RECIPE
The model is one part.
The business context completes it.
- Inventory API
Exact component availability at each warehouse.
- Courier quoting
Delivery estimates and a reviewable price.
- Crew calendar
Potential installation slots and assigned supervisors.
- Approval workflow
Human review before commitments.
Afterlight Exhibitions and this scenario are fictional. All tool responses, timing, inventory, and quotes are simulated. A real implementation would need connected systems, scoped permissions, availability checks at booking, and a qualified installation supervisor.
UP NEXT / Text message agents
Back to all four services Forma & Field
A better room. One text away.