Use case

Find staff at short notice: how an AI phones your entire pool in 20 minutes

12 September 2026·8 min read·Suisse Line

06:10, Monday. The foreman calls in sick, the job starts at 07:00, and you have a list of 80 numbers of people who have worked for you before. So in theory the answer is right there. In practice you do not have 50 minutes to call them. That is exactly where an AI phone agent comes in: it calls everyone at the same time and tells you within minutes who can come.

The problem is rarely a shortage of people – it is time on the phone

Almost every business with fluctuating demand already has a pool: former temps, casual staff, students, pensioners, people from last season. That list is worth a lot – and it still sits unused, because it cannot be called through at the one moment when it would matter.

Let us do the arithmetic. A call takes about three minutes on average, including dialling, ringing, the conversation and the note afterwards. Two out of three people do not pick up on the first attempt, so many contacts need a second round:

These are example calculations, not promises – but the ratio holds: a human calls in sequence, an AI in parallel. And because time has always been the bottleneck, in practice only the top of the list gets called. The other 60 numbers did not decline; nobody ever asked them.

How a call run works

1. Bring your list Upload Excel or CSV, or hand the pool over from your CRM, ATS or scheduling tool via webhook. Name, number and a few attributes such as qualification, region or last assignment are enough.
2. Describe the job In two or three sentences: what kind of assignment, where, from when, for how long, what is required. The agent turns that into its call guide – no script editor needed.
3. Start the run The agent calls in parallel, in Swiss German or whichever language is stored on the contact. You see live who has said yes, who has declined and who is still open.
4. Take the result Confirmations arrive as a list, SMS, email or straight into your system. Whoever says yes automatically receives the assignment details.

What the agent clarifies on the call

A good run does not deliver a «yes» or a «no», it delivers an answer you can act on. Typically the agent captures:

• Availability: from when, until when, full days or only parts of them
• Qualifications and licences: what exists and is still valid – from a forklift licence to a nursing diploma
• Travel: whether the site is reachable and whether a vehicle is available
• Terms: confirmation of the agreed hourly rate, premiums and expenses
• Reason for declining: sick, abroad, on another job, no longer interested – that information is worth more for the next run than the refusal itself

All of it lands structured in the reporting. After three months you know which part of your pool is actually active – and you can stop carrying dead contacts around.

One call, the way it actually goes

The agent understands English, Swiss German and standard German, lets itself be interrupted and does not promise anything it is not allowed to promise. What it should not handle – salary negotiations, sensitive cases, complaints – it hands over to a stored number or flags for a callback.

Why parallelism is the whole point

The real lever is not that the AI makes calls, it is that it makes them at the same time. Anyone searching manually works down the list and loses another minute on every person who does not pick up. The agent dials the next twenty numbers while the first conversation is still running.

Which industries this works for

The pattern is the same across industries, only the details differ. The detailed articles:

Your list stays where it is

You do not have to switch systems. Three routes are common:

• File: upload Excel or CSV – the fastest start, good for a first run.
• Automation: pull the pool out of your existing system through n8n, Make or Zapier and write the results back.
• API and webhook: direct connection to your ATS, ERP or scheduling software. The run then starts with one click from the tool your team already uses.

More on the Integrations page and with the Outbound Agents.

What you need to get right legally

Calling people you already know is legally unproblematic – as long as the basis is sound. Three points to have in order:

• Consent and basis: you are calling your own staff, former temps or registered candidates – people who gave you their number for exactly this purpose. Put that in the pool sign-up form or the employment contract.
• Transparency: the agent says in its first sentence that it is your company's digital assistant. That is fair, avoids irritation – and in practice people react more calmly to it than many expect.
• Data protection (revFADP): phone numbers and call content are personal data. At Suisse Line the processing runs on Swiss servers, the company is ISO/IEC 27001:2022 certified, and a data processing agreement or an on-premise deployment is available. Recordings you no longer need are deleted on your own retention schedule.

What it costs

For comparison: four hours of dispatcher time costs a multiple of that internally – and that is only the direct cost. The expensive item is usually the assignment that cannot happen without a replacement: the site standing still for a day, the shift running short, the banquet you have to turn down.

Ready in 30 minutes

1. Create an account Sign up, use the CHF 20 starting credit, no credit card needed.
2. Set the agent up Define assignment type, questions, calling window and stop condition. Then test it with your own number and listen to the wording.
3. Upload the pool and go live Start the first real run with 20 numbers, read the transcripts, sharpen the phrasing – after that it just runs.

The decisive point: set the run up before you need it. When it matters you only press start.

Frequently asked questions

How fast does the AI find a replacement?

That depends on your pool, not on the technology. The agent runs up to 20 conversations at the same time, so 80 numbers are covered in roughly 15 to 20 minutes. The first confirmation usually arrives after just a few minutes, and you see it live without waiting for the run to finish.

Do the people called notice that it is an AI?

Yes, and that is intended. The agent introduces itself in the first sentence as your company's digital assistant. Hiding it would be neither necessary nor fair. The voice is natural, understands Swiss German and can be interrupted like a human.

What happens if somebody does not pick up?

The contact stays in the run and is retried after a defined interval, for example twice at 20 minute intervals. Anyone still unreachable optionally receives a text message with a callback number.

Does this only work for one particular industry?

No. Construction, logistics, care, hospitality, cleaning, manufacturing, events – the pattern is always the same: a list of known people, an urgent gap and too little time to make the calls. Only the questions on the call differ, such as licences, shift times or qualification levels.

What does a run across 80 numbers cost?

Only actual talk time is billed, from CHF 0.39 per minute. A run with around 40 talk minutes therefore sits between CHF 15 and 20, plus the platform package from CHF 1'980 per year including 100 free minutes per month.

Can the agent confirm the assignment right away?

Yes. Confirmations can be written by webhook straight into your scheduling tool, your CRM or a spreadsheet, and the person who said yes immediately gets an SMS or email with location, start time and contact person. More than 4'000 integrations are available for that.

In short

Finding staff at short notice rarely fails because nobody is available. It fails because in the decisive half hour nobody can make 80 calls. An AI phone agent removes exactly that bottleneck: it calls in parallel, asks structured questions, documents cleanly and gives you the answer before the job starts. Your pool turns from a list nobody can work through into a tool you actually use.

Related reading: The solution in detail · Construction · Logistics · Staffing agencies