Volt Electronics sells consumer electronics online across a national market: phones, laptops, TVs and small appliances, with several thousand parcels moving through carriers every week. Customer care is a small in-house team that covers chat, email and the phone line from one room, plus seasonal help at every promo peak. The phone is where impatience lands: once a parcel is late or a delivery slot moves, people stop typing and start dialing.
The challenge
Most inbound calls were the same question, where is my order, and they buried the few that genuinely needed a human.
Agents read tracking numbers all day instead of solving real problems.
The call log was monotonous. Order number, tracking number, carrier depot, estimated date, repeat. To answer that one question an agent kept two systems open, the order panel and the carrier tracking page, while the caller waited on the line and heard hold music instead of a date.
Most inbound calls were the same question, where is my order, and they buried the few that genuinely needed a human. A laptop that arrived with a cracked screen, a warranty dispute, a parcel lost between warehouses: all of it queued behind status checks. Agents read tracking numbers all day instead of solving real problems.
The peaks were predictable and equally painful. One late carrier run in the evening turned into a wall of calls the next morning, promo weeks flooded the line, and evening and weekend calls simply went unanswered because nobody was staffed to take them.
Why earlier attempts failed
IVR with tracking lookup Callers mis-keyed long tracking numbers on the keypad and pressed zero for a live operator within seconds.
Bulk SMS after dispatch The link showed a raw carrier status with no delivery date, so customers called anyway to have it explained.
Seasonal phone temps New hires needed weeks to learn the catalog and carrier rules, and left right after the peak, so the queue came back every season.
Callback and voicemail queue Messages piled up, return calls landed in voicemail, and a thirty second question turned into three attempts across two days.
The solution
A MyChatBot voice agent now owns the order-status line: it picks up, reads the live order record aloud in the caller's language, and puts a person on the call only when the parcel itself has gone wrong.
Put the agent on the line Inbound calls run on the MyChatBot Calls SDK, the agent answers on the first ring, matches the caller number to recent orders and asks for the order number only when the match is ambiguous.
Wire the live order lookup MCP custom tools and the Webhooks API pull the current order record, carrier handoff and pickup point straight from the order system, so the agent speaks from live data instead of a script.
Load the shipping rulebook Business Domain docs and the product feed sit behind Agentic Search, so cut-off times, pickup point changes, warranty terms and return windows get the same answer a senior agent would give.
Log every call in CRM create_crm_lead and edit_crm_lead update the customer card in KeyCRM with the call outcome and the source tagged AI-, and lead_chat_link attaches the recording and transcript so the next person sees what was already said.
Draw the hand-off line Hand-off Control routes damage, loss and warranty disputes to a human with silent hand-off so the caller never repeats themselves, and Flight Control alerts the care team in Telegram the moment a call is escalated.
Call first on delays When a carrier flags a delay, outbound calls go out through the same agent to offer a new date or a pickup point change, and Follow-ups with Lead Processing labels track who was reached and who still needs a call.
Knowledge Base: The agent's Knowledge Base holds the shipping matrix by carrier and region, order cut-off times, address and pickup point change rules, warranty and service center terms, the return window policy, and the product feed with model names and configurations, so callers hear the exact item they ordered rather than a generic line.
The voice agent resolves 78% of calls from live order data and support costs fell 40%.
What was deployed
Agents used
Voice AgentLead QualifierAppointment Booker
Channels connected
VoiceWhatsAppWeb widget
Rollout timeline
Days 1 to 3
Two weeks of recorded calls sorted by reason, the top order-status intents written into the first instruction in the Configuration Wizard.
Week 1
Order lookup tool connected, Knowledge Base loaded, battle-testing run against real past calls until the agent stopped guessing on split shipments and partial dispatches.
Week 2
Live on inbound during business hours, hand-off thresholds tuned daily from listened-back recordings.
Week 4
Full round-the-clock inbound coverage plus outbound delay calls, with the care team working the escalation queue only.
Results
Metric
Before
After
Calls handled without a human
Every inbound call went to an agent, status checks included
78% of calls auto-handled from live order data
Average call length
The caller held while an agent checked the order panel and the carrier page
2 min
Support cost
Full in-house shift plus seasonal phone temps at every promo peak
-40%
Answer speed
Queue and voicemail on peak mornings after a late carrier run
Answered on the first ring, no queue
Coverage
Business hours only, evenings and weekends went unanswered
Round the clock, including weekends and campaign nights
"Our people now handle the hard 20%, the AI handles the rest, instantly."
Nadiia Lysenko
Customer Care Lead
What comes next
Volt is moving the same agent onto warranty and service center intake, where the call needs a serial number and a repair appointment, and reusing the order-status Knowledge Base on WhatsApp so the identical question gets the identical answer in text.
Ship something like this
Set up your first AI agent in 10 minutes. No credit card needed.
