Cut parcel-tracking calls by sending updates at the moments people usually ring: out for delivery with a realistic time window, when they're next in the queue, and straight after a delay or failed attempt. Send those from your delivery software, let AI write the exception messages, and let an assistant answer leftover "where is it?" questions from live tracking data.
Before changing anything, find out why people are actually calling. In a small courier firm a handful of causes usually produce most calls, and they need different fixes: a missing time window is a notification setting, a failed delivery with no explanation is a message-writing problem, and business senders chasing proof of delivery is a reporting problem. Only some of these need AI at all.
Tag two weeks of calls before you touch any settings
For ten working days, whoever answers the phone writes one line per call: who rang (recipient, sender, someone else), what they asked, and what they needed to be told. A shared sheet or a notepad both work. At the end, paste the lines into an assistant and ask it to group them:
Below are notes from calls to a courier company over 10 working days,
one per line. Group them into no more than 8 reasons. For each reason
give: a short name, the number of calls, two example lines, and
whether the caller could have been answered by an automatic message
(yes / partly / no). Don't create a reason for fewer than 5 calls;
put those under "other".An illustrative result for the six-van firm costed further down, which logged 450 calls:
| Reason | Calls | Could a message have answered it? |
|---|---|---|
| "What time today?" | 170 | Yes |
| Missed delivery, what happens now | 102 | Yes |
| Sender asking for proof of delivery | 63 | Yes |
| "It says delivered but I can't find it" | 42 | Partly |
| Change of address or safe place | 32 | Partly |
| Damaged or missing items | 22 | No |
| Other | 19 | n/a |
Check a few of the groupings against the original notes; assistants sometimes lump "missed delivery" and "delivered but can't find it" together, and those need different responses. In this example, three reasons account for about three quarters of calls, and all three can be answered before the phone rings.
Four messages that answer the question before the phone rings
Most delivery management platforms can send recipient notifications without any AI. Spoke Dispatch (previously Circuit for Teams), for instance, sends automatic updates when the driver starts the route, when the recipient's stop is next in the queue, and when the stop is marked successful or failed. Its tracking link can show just an ETA window, the window plus position in the queue, or live driver tracking once they're next. Onfleet offers similar triggers, including one that fires when the ETA drops below a threshold you set. Whichever you use, this sequence covers the common call reasons:
| Moment | What the message must include | Call reason it removes |
|---|---|---|
| Route starts | A time window (two hours is often safer than one), the tracking link, and how to leave a safe-place instruction | "What time today?", safe place changes |
| Next in queue | "Your driver is about 15-25 minutes away," with live tracking | Late-morning "is it still coming?" |
| Delivered | Where it was left, a photo if taken, and who signed | "Says delivered but can't find it" |
| Failed or delayed | What happened, in plain words, and the options: redelivery date, collection point or new safe place | "Missed delivery, what now?" |
The safe-place link on the route-start message answers most of the 32 change-of-address and safe-place calls, with one exception to plan for: parcels the sender has marked signature-required. A recipient who taps the link and writes "behind the green bin" for a signed-for parcel will be annoyed when it goes back to the depot anyway. Hide the safe-place option on those drops, and have the message say plainly "this parcel needs a signature, so it can't be left", which heads off the call before it starts.
For senders, the fix is usually a proof-of-delivery email or an end-of-day summary per account, not more messages to recipients. In the example above, that alone removes 63 calls a fortnight.
The delivered message carries most of the weight for the 42 "says delivered but I can't find it" calls, and the detail in it decides whether the phone still rings. Compare two versions for the same drop:
Before: "Your parcel has been delivered. Thank you for choosing us."
After: "Your parcel from [retailer] was delivered at 11:52 and left with your neighbour at number 9, who gave the name [surname]. Photo: [link]. Not there? Reply to this message and we'll check with the driver today."
The second version answers the next three questions (when, where, who) and gives a route that isn't the phone. It relies on the driver capturing the neighbour's house number and name at the door, so the change is as much a driver-app habit as a message template. Where the driver's app has a required field for "left with" or "safe place used", switch it on.
Where AI earns its place: exception messages
"Your parcel has been delivered" doesn't need AI; a template does it perfectly. The messages that go wrong are the exceptions, because drivers write notes in their own shorthand and the template can't cover every case. A realistic before and after:
Driver's note: "NA x2. side gate locked. dog. card thru door. back to depot"
What the template sends: "We were unable to deliver your parcel. Reason: NA x2."
What an AI step can send instead, drafted from the note and the redelivery options in your system:
Sorry we missed you today. Our driver knocked twice and couldn't get to the back door because the side gate was locked. Your parcel is back at our depot. You can choose a redelivery tomorrow, collect it from the depot after 2pm, or tell us a safe place: [link].
The instruction behind it keeps the AI on a short lead:
Turn this driver note into a short, friendly message to the
recipient (under 400 characters). Explain what happened using only
the facts in the note. Then list the options exactly as given below.
Never mention a delivery time, date or option that isn't listed.
Don't mention the dog or anything that could sound like blame.
Driver note: {note}
Options: {options from the system}The "don't mention the dog" line is there because an early illustrative draft read "our driver could not enter due to your dog," which is accurate and guaranteed to start an argument. Delay messages follow the same pattern: when a van breaks down with 40 drops left, the AI can write one clear explanation with the new window, which you then send to all 40 recipients at once, instead of waiting for 40 calls.
That breakdown message is worth a dry run, because the first draft usually overreaches. Given "van 4 broken down, replacement van arriving 1pm, remaining drops now 3pm to 6pm" and asked for a recipient message, an illustrative draft read:
We're sorry, our van has broken down and your parcel will now
arrive by 5pm today. A replacement vehicle is on its way and we
guarantee delivery today. Thank you for your patience.Two fixes before sending. "By 5pm" is the model's own tidy-up of a 3pm-to-6pm window, and the drops after 5pm would all ring in. And "guarantee delivery today" is a promise nobody authorised; if the replacement van is late too, you've made it worse. The version that went out said "now expected between 3pm and 6pm today" and "if anything changes we'll text you again", which is true, gives people a window to plan around, and commits you only to an update. Add "never shorten or round a time window, and never guarantee anything" to the prompt so the second breakdown doesn't need the same edits.
Connect this through your platform's API or an automation tool such as Zapier or Make: when a stop is marked failed, send the note and options to an AI step, then post the message back as the notification. Start with drafts that a dispatcher approves for the first week or two; once you trust the output, let it send automatically for routine cases and keep damaged-parcel messages manual.
An AI assistant for the calls that still come in
Some people will always ask rather than click a link. A web chat, WhatsApp assistant or AI phone agent that can look up a tracking record answers most of them. What it needs:
- A lookup it can call: your delivery platform's API, or a sheet refreshed every few minutes with tracking number, status, window and delivery notes.
- An identity check before revealing anything: the tracking number plus the house number or surname on the delivery. Delivery details are personal data, and "is anyone home at number 12 today?" is not a question to answer for strangers.
- Firm limits: it states only what the record shows and never promises a time the system doesn't have.
- Handover rules: damaged or missing parcels, complaints, and anything the record can't answer go to a person with the conversation attached.
An illustrative exchange on web chat:
Customer: where is my parcel it was supposed to come this morning
Assistant: I can check that. What's the tracking number, and the house number on the delivery?
Customer: CX48812, number 7
Assistant: Thanks. Your parcel is on today's route and you're 6 stops away; the current estimate is 12:40 to 1:20. You can follow the driver here: [link]. Would you like to leave a safe-place instruction in case you're out?
Watch how stale the lookup is. Suppose the sheet behind the assistant refreshes every 15 minutes. A customer asks at 12:14, and the assistant says "6 stops away", but the driver delivered at 12:09 and the record hasn't caught up. The customer, told the parcel is still coming, doesn't look on the doorstep and rings the office half an hour later. Two fixes: refresh more often if your platform allows it, and have the assistant say how old the information is ("as of 12:05, you were 6 stops away"). Reading one week's chats for timestamps like this is how the problem shows up, since the customer rarely says the assistant was out of date; they just ring.
The identity check has its own awkward cases. A common one: someone gives the right tracking number but a different house number, because they're the recipient's partner at a work address or a neighbour who took a card through the door. The assistant can't tell a neighbour from a stranger, so it shouldn't try. It says it can't match the details, offers to pass the question to the office, and a person takes it from there.
If you use WhatsApp for this, note Meta's pricing change from 1 October 2026: service replies within the 24-hour window become chargeable after the first 1,000 per business number each month, and rates vary by the recipient's market. WhatsApp customer service with AI covers setup and limits. Many of the same questions are covered for online shops in whether an AI chatbot can handle order tracking and returns, and when a chatbot should hand over to a human helps you set the handover rules.
An illustrative six-van courier, costed
Take a same-day and next-day courier with six vans doing about 650 drops a day. Before any changes, it takes around 45 tracking calls a day, each about four minutes once you include looking the parcel up. That's three hours of office time a day, or about $66 at $22 an hour.
The obvious fix, text messages for every update, has a cost many firms overlook. SMS is usually charged per segment and varies by country and provider; at an assumed 2 cents a segment, three texts per drop is about 1,950 texts a day, roughly $39 a day. That's more than half of what the calls cost. A cheaper mix:
- Route-start and delivered messages by email where you have an address. Spoke, for example, sends email notifications at no extra charge.
- Texts only for "next in queue" and failed or delayed deliveries, about 690 a day, roughly $14.
- Proof-of-delivery emails to business senders.
- An AI step for exception messages, and a chat assistant on the website.
Say calls fall to about 18 a day after a month, which is plausible when three quarters of calls were answerable by a message. Office time drops to about 1.2 hours, around $26 a day, so the saving comfortably covers the texts. The bigger gain is often failed deliveries: recipients who know the window stay in or leave a safe place, and every avoided re-attempt saves a driver's time and fuel. Measure both, because your figures will differ.
Better routes make windows easier to keep; whether a small courier firm can use AI to plan delivery routes is the natural companion to this work.
Where delivery updates create more calls than they stop
- Windows that are often wrong. A missed window triggers the call you were trying to prevent, plus annoyance. Widen windows on routes with frequent misses.
- Messages sent before the van is loaded. Tie the route-start message to the driver actually starting, not to the manifest being created.
- Branding confusion. Recipients who don't recognise your company name ignore the text or think it's a scam. Include the sender's name ("your parcel from [retailer]") where your contract allows.
- Duplicate messages when a stop is re-sequenced. Test re-routing on a quiet day and watch what gets sent.
- AI that fills gaps. If the note is blank, the message should say "we couldn't deliver today" and list options, not invent a reason.
Tracking calls per 100 drops, week by week
Calls alone rise and fall with volume, so track calls per 100 drops. Before the changes, the example firm sat at about 7; after, about 3. Keep the one-line call log going for a week each month so you can see which reason is growing. Add two more numbers: the share of drops delivered inside the stated window, and failed attempts per 100 drops. The window figure is quick to work out and says where to act. In an illustrative week of 3,250 drops, 2,860 landed inside the window, which is 88% overall. Split by route, five routes sat above 90% and one afternoon route managed 71%, because it crossed a school-run area between 3pm and 4pm. Widening that route's afternoon windows from two hours to three cost nothing and took most of its calls away. For a chat assistant, count how many conversations end without a handover and read a sample every week; how to measure whether an AI chatbot is working sets out what to look for.
Delivery update questions from courier owners
Should we send updates to the sender, the recipient, or both?
Recipients get the time-window and next-stop messages, because they're the ones deciding whether to stay in. Business senders usually want a proof-of-delivery email or a daily summary instead of every step. Ask your larger accounts what they want; some prefer that you contact their customers only under their brand, or not at all.
How accurate does the time window need to be?
Accurate enough that most drops land inside it. A two-hour window that is right nine times in ten generates fewer calls than a 30-minute window that is often wrong. Track how many deliveries arrive inside the window each week, and widen it for routes or times of day where you miss often.
Can we use WhatsApp instead of text messages?
Yes, if recipients use it, but check the costs first. Meta's WhatsApp Business Platform pricing changes on 1 October 2026: service replies inside the 24-hour window become chargeable after the first 1,000 a month per business number, and rates vary by recipient market. Compare that with your SMS rates before switching.
Will an AI assistant handle lost or damaged parcel claims?
It shouldn't decide them. Let it take the details, the tracking number, photos and a description, and pass the case to a person with everything attached. Claims involve judgement about liability and money, and customers with a damaged parcel want to know a human is dealing with it.
Further reads
- AI Route Planning for Small Delivery Fleets and Mobile Teams — Better routes make time windows easier to keep.
- How to Set Up an AI Phone Agent for After-Hours Calls — Set up an AI phone agent for calls outside office hours.
- Can AI Answer My Business Phone? What It Can and Can't Handle — What AI phone answering can and can't handle.
- AI Call Summaries: Log Every Phone Enquiry in Your CRM — Log every call automatically so reasons are easy to count.
- WhatsApp Business App vs API: Which Do You Need for AI Replies? — Pick the right WhatsApp set-up for automated replies.
- How to Set Up Missed Call Text Back for a Service Business — Reply by text when the phone rings out.
- How Dry Cleaners and Laundries Use AI for Orders and Collections — Where AI fits between the counter and collection: phone agents that book pickups, ready texts, an uncollected-garment ladder and calmer claim replies.
- How Independent Pharmacies Use AI to Shorten Phone Queues — Count the calls, then cut them: ready-to-collect texts, a phone menu that listens, a voice assistant with firm limits, and callbacks instead of hold music.
- How Garages Use AI to Send Repair Updates Customers Appreciate — How a small garage can turn technician shorthand into clear, timely repair updates with AI, and where those messages go wrong.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Spoke Dispatch help pages (recipient notifications); Onfleet support pages (notification triggers); Meta WhatsApp Business Platform pricing update notes.