AI

AI Automations

Let the systems handle the repetitive tasks.

From manual routines to workflows in operation

A customer enquires about their order. An invoice needs to be matched with a purchase order. The weekly report is missing figures from three systems.

We connect your systems and build AI agents and workflows that retrieve information, prepare responses and pass tasks on. You decide what can happen automatically and when an employee needs to approve.

We build on the systems you already use

  • Shopify
  • Make
  • n8n
  • Claude
  • ChatGPT
  • Zendesk
  • Gorgias
  • + your own systems

From data to a concrete next step

The AI reads the task, checks your systems and prepares the result. An employee always has the final say.

Customer service
  1. 01
    The task comes in
    Inbox
    A
    Anne K.to customer service
    Where is my order? #10482

    Hi. I ordered a rain jacket five days ago and haven't heard anything. Can you help?

  2. 02
    The data is checked
    Shopify · Order #10482
    StatusShipped
    ShippingPostNord
    TrackingReady

    The order has been found

  3. 03
    The result is prepared
    Gorgias · Reply draft
    Draft readyReply to Anne K.
    Hi Anne. Your order #10482 has been shipped with PostNord. You can track the package here: postnord.dk/spor …
    An employee reads and sends the reply
Product and inventory data
  1. 01
    The task comes in
    Shopify Admin · Products
    Cykel Pro rain jacketVariant: M / Black · SKU CP-M-BLK
    0 in stock
  2. 02
    The data is checked
    ERP · Inventory
    SKU CP-M-BLK24 units
    LocationMain warehouse
    WebshopShows 0

    The discrepancy can be described

  3. 03
    The result is prepared
    ClickUp · New task
    Draft readyCorrect inventory status for CP-M-BLK
    ERP shows 24 units, but the webshop shows 0. Synchronise the inventory for the variant.
    The person responsible approves the change
Invoices
  1. 01
    The task comes in
    Inbox
    L
    Supplier A/Sto accounting
    Invoice 2291

    Please find attached the invoice for delivery of 50 rain jackets.

    faktura-2291.pdf
  2. 02
    The data is checked
    e-conomic · Match
    Purchase orderPO-3318
    SupplierMatches
    Amountkr. 12.450 = kr. 12.450

    The invoice and order match

  3. 03
    The result is prepared
    e-conomic · Draft
    Draft readyDocument ready for approval
    Invoice 2291 matches PO-3318. The document and verification are attached.
    An employee approves it before it is posted

The agent prepares, you decide

The agent only works with the data you grant access to. We agree in advance which systems it may read from and what it may use the information for. Everything else remains closed.

Within those boundaries, the agent does the groundwork. It looks up, cross-references and prepares, so your employees don't have to search for information across systems. The task surfaces when it is ready for review.

The final decision always rests with you. An employee approves before anything is sent or changed. All events are logged, so you can follow what the agent has done and adjust the boundaries when you are ready to give it more responsibility.

Reads data
The agent retrieves the agreed information from your systems, e.g. order management, webshop or ERP. It only sees what you have granted access to.
Creates drafts
The agent compiles the information into a reply, a correction or an attachment, so the task is ready for review. If anything is missing, it does not guess. Instead, the task is flagged for manual clarification.
You approve
A team member reviews the draft and sends or approves it. All events are logged, so you can always see what has happened.

What we can help you with

Choose an area and see what the task looks like in the systems you already use.

Data and insights

Sales figures, conversion and deviations

Get notified as soon as sales or conversion deviates.

Daily monitoring with a notification when something deviates.

Example in practice

A drop in mobile conversion triggers a notification with the underlying data and possible explanations for the team to investigate.

# webshop-alerts
AI
Monitoring08:02

Mobile conversion has been clearly below normal since yesterday at 14:00. Desktop is unchanged. Possible cause: an error in checkout on iOS.

The team is notified immediately, with the underlying data included.

Analysis of return reasons

See why products are being returned.

Categorisation of return comments and patterns by product.

Example in practice

Return comments indicate that a model is often described as running small. The team receives a suggestion to clarify the size guide.

Cykel Pro · return reasonShare
“Runs small”
“The colour is different”
“Arrived late”
“Defective zip”

Customers' own words are grouped automatically. Here, the pattern points to the size guide.

Analysis of internal searches with no results

Find what customers are looking for but cannot find.

Finds searches with no results and suggests solutions.

Example in practice

Searches for a synonym with no results become new tags or an assortment suggestion.

Search with no resultsResultsSuggestion
shell jacket0Synonym
women's rain jacket0New filter
waterproof cycling trousers0Missing from assortment

Customer searches reveal what is missing from tags, filters or the assortment.

Automation and AI agents

Shopify Flow automations

Let fixed rules handle inventory, orders and tags.

Rules for inventory, orders, tags and customer segments.

Example in practice

Products with low inventory are tagged, and purchasing is notified automatically.

TriggerInventory quantity changed
ConditionInventory below 5 units
ActionAdd tag: reorder
ActionSlack message to purchasing

This is what the rule looks like in Shopify Flow.

MCP servers and AI agents on your own systems

Give Claude or ChatGPT secure access to your data.

We connect AI assistants to selected system functions via MCP and limit access, actions and logging.

Example in practice

Customer service asks an agent about an order status, and it looks it up directly in the ERP.

What is the status of order #10482?
Using tool erp.hent_ordrestatus(order: "10482")
→ status: shipped · carrier: PostNord

Order #10482 was shipped with PostNord yesterday. The customer can track the package via the tracking link.

The agent may only use the lookups you have granted access to via MCP.

AI agents for administrative tasks

Inbox, invoices and orders prepared for approval.

We build flows for invoices, inboxes, orders and documents with reply drafts and approval for agreed actions.

Example in practice

An agent prioritises the inbox, retrieves order information and prepares a reply. Invoices can likewise be matched with orders and sent for approval.

Supplier A/SInvoice 2291Invoice · matched
Anne K.Where is my order?Order · reply draft
Jonas P.The zip is brokenComplaint
NewsletterThis week's offersLow priority

The inbox is sorted, and the agent prepares replies and documents for approval.

Development and operations

Internal tools and dashboards

Bring webshop, ads and inventory together in one place.

Custom tools that bring together data from multiple sources.

Example in practice

A dashboard brings webshop, ads and inventory together in one overview for management.

dashboard.jereswebshop.dk
Management overview · yesterday
Revenue
Ad spend
Inventory at risk
WebshopShopify
AdsGoogle and Meta
InventoryERP

Data from multiple systems gathered in one place.

Automated reporting

This week's figures in your inbox, without manual work.

Recurring reports are generated and sent automatically.

Example in practice

Every Monday, the week's agreed KPIs and deviations are compiled into a report and sent to the selected recipients.

Inbox
Weekly reportMonday 08:00

In short: Revenue tracked expectations. Mobile declined midweek, and two products ran out of stock.

uge-rapport.pdf

The report is generated and sent automatically with comments.

Webshop apps and extensions

Functions the standard solution does not have.

Apps and functions for e.g. Shopify.

Example in practice

An internal app makes it possible to update hundreds of products in one step.

Shopify Admin · Bulk editing
Cykel Pro · MWinter tag
Cykel Pro · LWinter tag
Urban Shell · SWinter tag
Urban Shell · MWinter tag
Update selected products

Many products are updated in one step in an internal app.

Advisory

AI strategy and roadmap

Find out where to start.

Where AI creates the most value, prioritised by impact and effort.

Example in practice

A workshop results in a prioritised plan with the first three automations.

High impact · low effortOrder lookupInventory alert
High impact · high effortProduct data
Low impact · low effortReport
Low impact · high effort

After the workshop, the opportunities are prioritised. Start at the top left.

Rollout of Claude or ChatGPT

Get the whole team off to a strong start with AI.

Setup, guidelines, training and integration.

Example in practice

The team gets a shared setup with guidelines, onboarding and access to the data sources approved for the purpose.

Customer serviceInstructions: Reply politely and briefly. Always check the order status.
Retningslinjer.pdf
Returpolitik.pdf

Shared projects with instructions and approved sources for each team.

We start small and build on what works

Many AI projects stall because they start too big. We begin with one concrete task that takes up time in everyday work, and get it running within your own systems. Once it runs reliably, we build further.

You are involved every step of the way. We agree in advance on which data the solution may use, where a team member must approve, and what needs to be in place before it goes live.

From the first task to a solution in operation

We start with a well-defined task and agree on what needs to be in place before the solution goes live.

01

Analysis and bottlenecks

We review the task, data, and your current systems. Together we prioritise what is most relevant to build first.
02

Architecture and data

We select platforms and integrations, and clarify access, data processing, approvals, and responsibilities.
03

Development and integration

We build and connect the solution. Testing against agreed scenarios shows where it works and where a team member needs to take over.
04

Deployment and monitoring

We agree on handover, monitoring and maintenance, so that errors and changes in the systems are handled.

Which workflow should we start with?

Start with one specific task you want assessed. We look at systems, data and the most important dependencies.

Questions about AI Automations

Repetitive tasks with available data and clear rules are a good place to start. This could be order lookups, invoice handling, reporting or checking data between systems.

We build a path to manual processing. Thresholds for confidence, amounts, data quality, and actions are defined in the solution, so relevant cases are routed to a human agent.

We agree on responsibility for operations, monitoring, and changes before rollout. Access, error alerts, and maintenance are defined as part of the handover.

We clarify data sources, purposes, access, storage, and vendor terms before development begins. GDPR requirements and the use of data for model training are factored into service selection, configuration, and agreements.