How Many Support Messages Can AI Phone Control Answer?

How a day of messages accumulates

Someone I know handles support for a dozen store accounts. She says the most robotic part of her day is the first five minutes: signing into each account to check for overnight messages.

The real work starts at nine thirty. One asks whether an item is in stock, one asks about the courier, one asks for a discount. She switches between accounts, finds the right reply, confirms it sent, switches back.

She counted once: about a hundred and twenty replies by lunchtime, and fewer than twenty of them needed any thought. The other hundred had answers she already knew.

That is the problem. The answer is already known, but every one of them has to be sent by hand.

This article takes that apart: what can be handed over, what has to stay, and where the line sits.

---

Split it into three groups and the picture clears up

Pour a day of messages out and there are only three kinds.

• Fixed answers. In stock, price, shipping cost, lead time, returns, availability. The answer already exists; the only difference is retyping it.
• Almost fixed. “Which of these two is better”, “any discount if I take two”. The first half has a prepared answer; the second half needs a look at the specifics.
• Must be a person. Complaints, returns, anything involving a promise, and any conversation where the tone has already turned.

Once sorted, you notice the first group is the largest and the third is the hardest, with the second in between. And where does most of the day go? Into the first group, because it is the largest.

---

Group one: how to hand it over

The method is to store the answers and pick one when replying.

Write the common questions as one line each:

Yes, in stock
Free shipping, dispatched within 48 hours
Seven-day returns accepted
Price is X, or Y for two

Replying stops being typing and becomes picking. That sounds like a small change, but it turns “deciding how to phrase it” from something you do every time into something you do once.

The next step is letting a machine handle it. With iEasyRun you can describe it:

Open the support pages for these stores, find messages asking whether items are in stock, the price, shipping, or lead time, and reply using the matching line. Read prices and shipping from this sheet.

Three things sit in that description: which messages, which replies, and where the data comes from. Nothing about how to click.

Why can this be handed over? Because it meets three conditions — high frequency, fixed answers, and a cheap mistake.

One detail worth noting: a machine is more consistent at this than a person. A person answering for the thirtieth time gets bored, and boredom produces shortcuts and errors. A machine does not get bored.

---

Groups two and three: how to divide the work

• Group two should not be fully handed over. The sensible split is for the machine to prepare the material and the person to send it — the machine reads the relevant price, stock and specification and puts them in front of you, and you glance and send.

What that saves is not reply time but lookup time, and lookup is the slowest part of these conversations.

• Group three is a line that cannot move. Not because a machine answers badly, but because these conversations need something different: to be taken seriously rather than answered accurately.

I have seen teams automate complaints with templates in pursuit of full automation. The result is immediate — the customer escalates, then files a platform complaint. The minutes saved are paid for with the account.

• A practical boundary: if sending the wrong thing causes real loss, do not hand it over. Getting “in stock” wrong is awkward; getting “can I return this” wrong is a dispute.

---

How the day should be sequenced

Once the three groups are sorted, the rhythm follows.

• Fifteen minutes before work: overnight messages. I would not hand this over — it is also how you find out what today looks like. New categories appearing overnight are what tell you to adjust the replies.
• Morning: clear group one. Let the machine take the standard questions; people handle the parts of group two that need judgement. Mornings are when judgement is best.
• Afternoon into evening: watch group three. An overseas customer’s working day lands in your afternoon and evening, which is when complaints and after-sales arrive. Someone needs to be present.
• Wrap-up: read the day’s record. Which category spiked today, which reply got used repeatedly — that is what decides tomorrow’s adjustments.

The point of this arrangement is putting people where judgement is needed and machines where volume is highest but complexity is lowest.

---

An easy thing to overlook: where accounts live

Everything above is about replying, but multi-account support has an unavoidable second question: where do the accounts sit?

Rapid sign-in switching on one device is itself an anomaly. The normal arrangement is to spread accounts across devices, each handling its own set, and to work by device group when centralised handling is needed.

With a handful of accounts you will not notice. With more, it becomes a bottleneck — and it is the kind of problem you only see in hindsight: while nothing is going wrong it just feels “fine”, and when something goes wrong the cause turns out to be here.

---

A test you can reuse

Take any category of message and ask three questions.

• Is the answer fixed? If it has to be thought through every time, handing it over just moves the thinking.
• Would a wrong answer cause real loss? If so, it stays with a person. This one is not negotiable.
• Does the customer want accuracy or tone? Accuracy can be handed over; tone cannot.

Pass all three and it can go. Fail one and it should wait.

• What first? The highest-frequency, simplest category — stock and price questions. Largest volume, cheapest to get wrong, and once it runs steadily you can add harder categories. Start with the hardest conversation and two failures will make you abandon the whole idea.

The software is completely free and runs on your own computer. For device-side troubleshooting, read what to do when a phone keeps going offline; for arranging many devices, see managing 20 phones solo.