Compare documents
- Incoming
A purchase order and confirmation are available.
- KIAI suggests
The AI matches items, quantities and dates and highlights differences.
- Your team checks
Your team checks discrepancies and approves the transfer.
SolutionsAI Consulting & integration
AI where it
takes work off your hands.
Review documents, route enquiries and gather information: we assess where AI can reduce repetitive work for your team. Then we implement a specific application. We decide beforehand which data it may see, and your team checks every result before it moves on.
On the left an order confirmation with item, quantity, price and a marked delivery date, on the right the check against the purchase order: item matched, quantity and price agree, the delivery date differs. Below, a note that this is the AI’s suggestion and approval rests with your team.
Typical situations
When AI pays off
AI usually does not pay off for the whole company at once, but for a single task.
Repetitive manual work
Your team repeatedly reads similar documents, sorts enquiries or copies information by hand.
Scattered knowledge
Answers are scattered across documents and systems, and finding them ties up experienced staff.
AI is already in use
Personal accounts or first experiments already exist. What is missing is managed access for everyone.
Use cases
From input to a reviewed answer
Three possible applications that we test on your real work - with clear review by your team.
Route enquiries
- Incoming
A message arrives in the shared inbox.
- KIAI suggests
The AI summarises the request and suggests the responsible team.
- Your team checks
Your team reviews unclear requests and confirms or corrects the routing.
Find knowledge
- Incoming
An employee asks a question about approved documents.
- KIAI suggests
The AI finds relevant passages and prepares an answer with sources.
- Your team checks
The employee checks sources and answer before using them.
In daily work
Ask a question, get an answer with its source
An assistant that only uses your approved data: every answer names its source, and everyone sees only what they may see.
Which orders are overdue?
- Permissions checked
- Searching: orders in the ERP
Three orders are overdue:
A-1167Customer Northoverdue
A-1171Plant Southoverdue
A-1175Trade Westoverdue
Source: orders · filtered by your permissions
What did we agree with Customer North about delivery?
- Permissions checked
- Searching: sales mailbox
Free delivery from the Hamburg warehouse, partial deliveries only by arrangement. That is what the order confirmation email says.
Source: sales mailbox · order confirmation Customer North
When is machine 3 serviced?
- Permissions checked
- Searching: technical network drive
According to the maintenance manual, every quarter and additionally after every tool change.
Source: maintenance manual machine 3 · section maintenance
Show me the payroll list.
- Permissions checked
You do not have permission for this. Please contact the HR department.
No approved source
Data & access
Not everything has to leave the building
For every task we decide together, in writing, where the AI runs.
Option 1
In-house
Models run on your hardware. We assess the requirements for the specific application.
Option 2
At a provider
We review the service’s terms on data processing, storage and training.
Option 3
Both combined
Tasks are split, and it is agreed in advance which data may go where.
Connectable, for example
- Network drives
- OneDrive & SharePoint
- Shared mailboxes
- ERP & business systems
Permissions are checked before retrieval: each user reaches only the sources approved for them, tested before rollout.
Approach
One task first.
Then the next.
We work out what the first task saves. Only then does the next one follow.
Collect and assess the tasks
We clarify which tasks are candidates and which AI tools your team already uses on its own initiative.
After this you have
Your tasks, sorted by effort and benefit, and the decision on where we start.
Test it on one task
The AI works on real cases, with an agreed location and data access.
After this you have
The comparison with today’s effort and a written record of which data the AI may see.
Bring it into everyday work
We set up permissions, logging and approvals and brief the department.
After this you have
A task in everyday work and a basis on which the next one goes faster.
Let us talk about which task to start with.
Book a callAI directly in the system
With ElbDesk, the foundation we build on, AI accesses your data directly, with defined permissions. An AI project does not require an ERP change: if you have an ERP, it stays. Accounts and warehouse keep running there.
Contact
Where could AI take work off your hands?
Name a task that AI should make easier. We will discuss how to test the value on your data.
Tobias Kietzmann
Managing Director · Your permanent point of contact
Send an enquiry
Tell us briefly where things are stuck. We will get back to you personally.
Thank you, your message has arrived. We will reply personally to discuss the next step.
A short description is enough. We will work out the details together.
Guides on this topic
Answers to the questions we are asked most often before a project - in detail, to read at your own pace.