Threll.ai
Orchestration

Seven jobs you can move from the dashboard to the chat window

Connecting Threll to the AI you already use takes ten minutes. The hard part is knowing what to ask for afterwards. Here are seven jobs from an ordinary week – the sentence you type, what happens behind it, and what usually goes wrong.

ByEspen Hareide Nilsen10 min read
A man in a dark knit sweater stands at a standing desk in morning light holding a coffee mug in one hand and a phone in the other, with an open laptop and a matte black desk phone with a coiled cord on the desk in front of him

Connecting is the easy part. You paste in one address, sign in the way you always do, approve what the AI is allowed to touch – and then Claude, ChatGPT or Gemini is standing with a hand on your phone system. The steps for each of the three are in a separate article.

What is rarely written down anywhere is what to type next.

That is where people stop. The connector is green, the chat window is empty, and the first sentence tends to be «what can you do?» – which gets you a list of tools, not a finished job. Below are seven jobs from an ordinary week of running a voice agent. For each one: the sentence you actually type, what happens behind it, and the trap that costs the most time the first time round.

One thing first. The AI you are sitting in does not call anyone. It does what you used to do with the mouse: searches, reads, counts, proposes, prepares. The only thing that reaches the world outside is an outbound call – which is exactly why it sits behind a confirmation.

1. Monday, 8.40 am: what came in overnight?

What you type: «Show me every inbound call from 4 pm yesterday until now, local time. Which ones were handed to a human, and what were they about?»

What happens: the AI pulls the call history for the period, picks out the escalated ones, opens the transcripts and summarises them in plain language. You read five lines instead of clicking through twenty log rows to find the three that mattered.

The trap: two of them, and both are mundane.

Time is the most common source of confusion. Name the time zone and give the date. «Yesterday» is not unambiguous at 8.40 in the morning, and a window that is off by a few hours gives you a morning summary missing the evening before.

And: your own test traffic looks exactly like real traffic. The calls you place yourselves to check a change sit in the same list as the customers. Tell the AI to leave your own numbers out, once, and keep that sentence in a saved prompt.

2. Tuesday: why did we lose that one?

What you type: «Find the call that came in around 2.10 pm yesterday and ended in under a minute. Show me the last thirty seconds word for word.»

What happens: the AI finds the call, pulls the events and the transcript, and gives you the exchange as it happened.

This is the job that changes the most for people. Not because it saves time – finding one call in the interface is not slow either – but because the threshold for doing it drops to nothing. When it costs one sentence, you do it ten times a week instead of once a month. And one transcript tells you something a monthly average never will: it shows you the sentence where it broke.

As a rule that sentence is either the opening line or the moment the agent should have handed over and did not.

The trap: do not ask for «every call from last week, word for word». Transcripts are long, and you burn both your patience and the context window on material you were never going to read. Ask for the selection first and the detail second.

A desk seen from above with a printed list on white paper in the middle, a hand resting on the page holding a ballpoint pen just above it, the edge of an open laptop to the left and a phone face down to the right
Preparing a round calls no one. It produces a list you are meant to read before you say yes.

3. Wednesday: the calling round of eighty

What you type: «Here are eighty leads from the CRM. Create them as people in Threll, prepare calls with Harald tomorrow between nine and eleven, and show me the whole list before anything goes out.»

What happens: the people are created in one operation – up to a hundred at a time – and the calls are prepared in the next. Preparing calls no one. It produces a proposal naming the worker, the number it dials from, who is being called, when, and with what context. You read it. If you say yes, the whole round is queued together: either all of them or none. An approval can only be used once, so the same round cannot go out twice by accident.

The trap, and it is worth the whole article: anything that differs from customer to customer belongs in the context for that individual call – not in the instructions field.

The instructions field is not an addition to the script. It is the script. Write «remind them the contract expires on 1 November» there and you have effectively replaced Harald's entire script with that one sentence, for those eighty calls. The agent loses its opening line, the disclosure, the escalation rule and everything else somebody thought carefully about. Put the detail in the context instead, and the agent keeps its script while still knowing what it needs to know about this customer.

One small practical matter: scheduled calls go out within a couple of minutes of the time you set, not on the second. Schedule in windows, not in points.

Changed your mind? «Cancel the round at nine tomorrow» works right up until the calls are sent. After that it is no longer a scheduled job, and the agent is on the line.

A woman in a grey shirt sits at a desk in afternoon light wearing large over-ear headphones, bent over a stack of printed pages she is marking with a pencil, an open laptop beside her
The one job where the AI genuinely has the upper hand: it will read twenty transcripts against a twelve-page script without skipping any of it.

Which of the seven would save you the most time this week?

Fifteen minutes where we connect the AI you already use and try one of them on your own calls.

Book Meeting (opens in a new window)+47 24 12 66 30

15 minutes · no obligation · pick your own time

4. Thursday: the script that stumbled

What you type: «Fetch Nora's active script, read it against her last twenty calls, and propose what should change in the section on pricing. Show me the proposal, do not write it in.»

What happens: the AI finds the active script document for that worker, reads it against real calls, and points at the places where the agent actually hesitated – not where you assumed it did.

How the change works: a new version is a new document, not an overwrite. The previous one has to be taken out of use, and it is the combination of active and linked to the worker that decides what the agent says on the next call. A new document takes a few seconds to finish processing before it applies. Always ask to see what changed, and ask for confirmation of which document is now the active one.

The trap: not every worker keeps its script as a document. Several inbound setups hold the lines in the worker's own configuration text instead. Ask where the AI found the script before you ask it to change anything – otherwise you can end up with a tidy new version that no call ever uses.

And what holds regardless: what the agent says can bind you. A proposed wording is a proposal to a human being, not a decision.

5. Friday: the numbers – and what is not being measured

What you type: «How many calls last month, split by inbound and outbound, answered and unanswered, and how long were they on average? Compare with the month before.»

What happens: you get the numbers without opening a dashboard and – more to the point – you can ask the follow-up immediately. «Which of the unanswered ones came outside opening hours?» is a question no report has a field for.

The trap: also ask «what call-analysis dimensions are configured on this worker?»

If the answer is none, nobody has decided what counts as a good outcome. You have volume and duration, and nothing at all about whether the matter was resolved. That is not a limitation in the AI, and no amount of better phrasing fixes it – it is a decision somebody has to make once. The numbers that are easiest to pull are also the ones that flatter the agent most.

6. The standing one: the questions nobody answered

What you type: «Go through the last thirty days of calls and find questions that keep coming up and that the agent had no good answer to. Rank them by how often they came.»

What happens: you get a list of the holes in your own knowledge base, drawn from what callers actually asked – not from what you assumed they would ask when the script was written.

This is the best raw material there is for three things at once: the next version of the script, the FAQ on your website, and the knowledge base the agent looks things up in. It is also a cheap way to discover that something completely trivial is missing – like the fact that you closed early before a public holiday, and the agent was the last to hear about it.

7. The actual point: one sentence, three systems

The first six jobs save you clicks. The seventh is the reason MCP exists.

What you type: «Check what slots we have free on Thursday afternoon. Call the top five on the waiting list, offer one of the slots, and book the meeting for whoever says yes. Show me the list first.»

That is the calendar, the phone and the customer list in one sentence. Nobody built that integration. The three systems use the same standard socket, and the AI in the middle is the glue.

It is also where you should be most awake. The more systems in one sentence, the less oversight you have of each individual step. Ask for the list. Every time.

Five sentences to try today

If you are only going to test one thing before you put this article down, take the last one.

  • «Summarise every escalated call from the last seven days, local time, with what each was about.»
  • «Which workers do we have, what languages do they speak, and which are active right now?»
  • «Find the three shortest answered calls this week and show me why they were so short.»
  • «Fetch the active script for [worker] and explain what it does when someone asks to speak to a human.»
  • «Prepare one test call to my own number in ten minutes, and show me the proposal.»

The last one is the best first. It shows you the entire confirmation loop on a number where no customer is disturbed.

Three rules that keep this enjoyable three months from now

Leave the confirmation on. It costs you one keystroke and saves you the one time the AI misread who «those five» were. Read the proposal – that is the entire reason it is there.

Give read access to whoever only reports. If the connector is there for monthly reports and summaries, it does not need to be able to place calls. Permissions are set at sign-in and cannot be widened along the way without somebody saying yes again.

Write down what the AI is not allowed to do on its own. One line will do: which lists are never called without a named person having seen them, and which offers the agent may never make. It takes two minutes now and is impossible to reconstruct later.

And what does not change because the call was planned from a chat: the agent still has to disclose that it is artificial intelligence, the escalation routine still applies, and what was said still has to be documentable. Orchestration moves your work. It does not move the responsibility.

Threll.ai builds voice agents in Norwegian, Swedish and Danish. The MCP server is part of the platform, not an add-on. If someone on the engineering side wants the detail, it is in the connection guide and the MCP reference.

Frequently asked questions

Up to a hundred calls are prepared at a time, and the round is approved as a whole: either all of them go into the queue or none do. Longer lists are split across several rounds, and it is worth asking to see each round on its own before confirming it. An approval can only be used once, so the same round cannot go out twice.

A change is always a new document rather than an overwrite of the old one, and it is the combination of active and linked to the worker that decides what the agent says on the next call. Ask to see what changed, and ask for confirmation of which document is now active. If you only want a proposal, say so in the sentence: show me the proposal, do not write it in.

Ask for it to be cancelled. That works right up until the calls are actually sent, and scheduled calls go out within a couple of minutes of the time they were set for. If the round has already started, the agent is on the line, and it is the escalation routine rather than the cancellation that matters.

Because the time zone is rarely stated. Name it and give the date explicitly instead of saying yesterday or today, particularly early in the morning and late in the evening. The same applies when you schedule outbound calls: state the time and the zone, and schedule in windows rather than to the minute.

It can be built, but we do not recommend it for anything that dials. The confirmation exists because an outbound call is the one action that reaches a person outside your organisation, and standing in it is cheap. For everything that only reads – summaries, numbers, searching transcripts – there is little reason not to let it run on its own, as long as the connector only has read access.

Related Articles

Continue reading more articles

See all articles