Read documents
A model reads a data sheet, quote or contract the way a person does and returns structured data, even when every document is laid out differently.
Service · AI automation
AI that worksinside your processes.
Our automations embed AI models directly in the process chain. They read documents, check content, decide according to your rules and complete the next step themselves.
✓Claude, OpenAI, DeepSeek, Gemini✓With rules and checkpoints
How it worksWhat happens
What happens
technically.
Four steps that run the same way for every transaction.
1Trigger
An event starts the process: an incoming message, a new record in the shop or a fixed time of day.
2Reading
The model reads the content as text. To do that it breaks it into tokens, the smallest units it works with. One token is about four characters, a page of text around 500 of them.
3Deciding
It classifies the case according to your rules and picks the next step. Anything that stays unclear goes to a person instead of into the system.
4Carrying out
The result is written where you need it, with a log of every step.
Why tokens decide what it costs you
Billing follows the amount of text processed, not the number of users or licences. A process that reads one page per transaction stays cheap, even when it runs ten thousand times a year. Which model takes on a step we choose by what the task demands: a small one for classifying, a large one for writing.
What AI handlesWork that used to be
Work that used to be
done by people.
Wherever content needs to be understood, assessed or created, AI plays to its strengths.
Create content
Product texts, FAQs, translations and social media posts are written in your tone of voice.
Check and classify
Assign categories, verify mandatory information and detect deviations, even in messy data.
Decide by your rules
Approve, hold back, adjust a price or ask a question: the model picks the next step and records why it decided that way.
Handle enquiries
Customer enquiries are classified, answered or forwarded to the right person.
Process images
Cut out and standardise product images and generate new visuals from product data.
The differenceClassic automation stops.
Classic automation stops.
Ours thinks further.
Rules work until an exception comes along. At that point, an AI model takes over in our systems and completes the process.
Classic automation
Works until an exception comes along.
InputNew data
→
RuleIf, then
→
ExceptionStop
→
Humansteps in
Automation by core.iq
AI takes over where rules fall short.
InputNew data
→
AIunderstands
→
AIdecides
→
Resultdone
Control
AI with clear guardrails.
Our AI doesn’t act freely but within fixed rules that we define together with you. Every decision is logged and remains traceable.
Where an approval makes sense, the system asks instead of guessing. This keeps quality high, even at large volumes.
- Rules and limits per process
- Automatic checks before every output
- Traceable logs for every decision
- Asking instead of guessing in unclear cases
- Human approval wherever you want it
ModelsThe right model
The right model
for every step.
We are not tied to any provider. For every task we choose the model based on quality, speed and cost. When a better model comes onto the market, we switch without anything changing for you.
In use
Claude
OpenAI
DeepSeek
Geminiand other providersTexts, image analysis, translations and decisions can be handled by different models depending on the requirements.
Use casesWhat already runs today
What already runs today
without anyone touching it.
E-commerce
From data sheet to a product ready to sell
Documents and manufacturer pages are read, product copy, attributes and nutritional values are written, translated into four languages, images cut out and the item created in the shop. Whatever is missing from the documents goes back to the supplier as a query.
Marketing
From the idea to the finished post
The model develops several approaches for a post, picks the strongest, writes the script, generates image and sound, checks the result for mistakes and puts the post in the calendar with a date.
Compliance
A whole range checked against the law in force
Ingredients, limits, claims and labelling are checked against the relevant regulations. Every deviation is explained and documented with its source, for hundreds of items in one run.
Sales
An enquiry becomes a finished quote
The enquiry and its attachments are read, the real need identified, the matching items found in the range, volume prices calculated and the quote put forward for approval.
Customer service
An inbox that works through itself
Every message is read, assigned to its case and answered with order and delivery details added. Anything that calls for judgement reaches the person who decides, with a summary and a draft reply.
Back office
The accounts run overnight
Receipts from the inbox, portals and photos are read, supplier, account and tax rate determined, matched against order and payment and posted. Unclear cases are left with a reason.
FAQ
Frequently asked questions.
The most important answers at a glance.
What is AI automation?
AI automation means that AI models take over tasks in a business process themselves: they read documents and emails, check content, write texts and decide according to defined rules. Unlike classic automation, they don’t stop at every exception.
Which AI models does core.iq use?
We work with Claude by Anthropic, GPT by OpenAI, DeepSeek, Gemini by Google and other providers. For each step we choose the model based on quality, speed and cost.
How do we stay in control?
The AI works within fixed rules and limits. Every decision is logged, and wherever you wish, the system asks for approval before the next step.
Which tasks are suitable for AI automation?
Tasks with unstructured content or judgement: analysing receipts and data sheets, classifying and answering enquiries, creating product texts and translations, or checking products against regulations.
What happens to our data?
Before the first run we define which data reaches a model at all and which stays in house. We use business accounts where the provider does not use your content to train its models. On request the automation runs in your own account, so access and data stay with you. As a processor we work under a contract based on Swiss data protection law.
Will we depend on core.iq afterwards?
No. On request the automation runs in your own account, you receive the access details and documentation of what happens where. Support is cancellable monthly, and the automation keeps running even if you no longer have us look after it.
What happens if the AI gets something wrong?
Every step has checking rules. Anything that fails a check lands on an exception list instead of in your system, and in unclear cases the automation asks instead of guessing. Every decision is logged, so any case can be traced. Mistakes that go back to our rules we correct as part of the support.
How long until something is running?
A first, clearly defined automation usually runs within a few days. We deliberately start with one single process, so you can judge from the result whether the next one is worth it instead of waiting months for a large project.
Do we need our own IT staff for this?
No. We set the automation up, connect it to your existing systems and look after the running operation. You need someone who knows the process, not someone who maintains technology.
What does AI automation cost at core.iq?
We work at a fixed price per process instead of by the hour. On top of that comes the use of the AI model, billed by volume. Ongoing support is optional and cancellable monthly. The binding quote follows after a short look at your process.
First contactLet us talk about
Let us talk about
your processes.
You have a task in mind that means a lot of manual work today? We show you how an AI model can take it over and where control points make sense.
✓Which of your processes can be automated
✓What effort and what timeframe are behind it
✓What the first step delivers, before you invest
✓Directly with a developer, not with sales
Write to us →