AI Consulting
After a workshop, you know where AI saves your team time.
- Workshop with your team
- AI policy
- Training per role
OpenAI
Claude
Gemini
Llama
Mistral
AI automation, from the first process to operation
ASES is an AI agency that automates the processes in your company, from sales and customer service to accounting. When recurring work runs automatically, you save time and costs. We start with one task, connect the solution to your systems and then take on the next.
Inbox
Phone
Chat
Your workflow
AI reads
Rules check
ERP
Calendar
CRM
Inbox
Phone
Chat
Your workflow
AI reads
Rules check
ERP
Calendar
CRM
How we bring AI into your company
Introducing AI starts with a task from your daily work. Choose where you stand today. Each phase ends with a result you can see and review.
- You are here
Initial consultation
- We
- We ask you to show us where your team reads documents, enters data or answers inquiries, and with which programs. Together we choose a task that takes a lot of time and lends itself well to automation.
- You
- You show us real examples, such as emails, invoices or forms, and name the programs involved.
- At the end
- The task we start with.
- You are here
Scope and quote
- We
- We define which steps run by fixed rules, where an AI reads, sorts or writes and where your data is stored. We also check which role and obligations you have under the EU AI Act.
- You
- You set the priorities and review the quote.
- At the end
- A fixed-price quote with a precisely described scope.
- You are here
Design and development
- We
- We set up the workflow, connect it with your systems and write the test cases we check it with. You see work in progress through a preview link.
- You
- You give feedback on the work in progress and make the decisions with us.
- At the end
- A workflow that works with your data and is ready for testing.
- You are here
Testing and go-live
- We
- We test the workflow against the test cases and the agreed scope. At launch, we train your team and everyone who works with the solution.
- You
- You review the result with us before the launch.
- At the end
- A running workflow and a team that works with it.
- You are here
Support and further development
- We
- On request, we monitor the workflow and install updates.
- You
- You tell us which task is next in line.
- At the end
- One workflow after another that takes work off your team.
Your first step
We look with you at the tasks your team repeats every day and choose one to start with.
Enough to get started
A task that costs your team time, and an example of it, such as a typical email or invoice.
Your first step
We look at what your team uses AI for today and with which data, and connect this work to your systems.
Enough to get started
A list of the tools in use.
Your first step
Then we start directly with this task and clarify the scope with you.
Enough to get started
A short description of the workflow and the programs involved.
Your first step
We check what is running today, even if another provider built it, and plan the next task with you.
Enough to get started
Access credentials for the running workflows, if you have them.
At the end of each phase, you decide whether we continue or stop here.
The comparison further down shows whether your task needs AI or whether fixed rules are enough.What is AI automation?
What is AI automation?
AI automation combines fixed workflows with AI for the steps where rules are not enough. An AI reads a freely worded email, recognizes the details on a scanned receipt or writes a suggested reply. Anything that can be described as “if A, then B” runs as a fixed workflow, fast and traceable.
The same request in three ways: a customer wants to know where her delivery is.
“Where is my delivery?”One request
| Fixed workflow | AI chatbot | AI agent | |
|---|---|---|---|
| What happens | Assigns the email to the right inbox and sends a confirmation of receipt. | Explains in the website chat, based on your shipping terms, how long a delivery takes and how to track it. | Looks up the order in the ERP, checks the shipping status and prepares the reply. |
| Who decides the next step? | The workflow as it was set up | The person in the conversation | The AI, with the tools it is allowed to use |
| With the same input | Always the same result | A similar answer | The path can differ |
| Fits if | the rules are fixed | the same questions keep coming up | a task has many small steps in several programs |
| Go to topic | Go to topic | Go to topic |
- What happens
- Assigns the email to the right inbox and sends a confirmation of receipt.
- Who decides the next step?
- The workflow as it was set up
- With the same input
- Always the same result
- Fits if
- the rules are fixed
An AI phone assistant is a chatbot or an agent on the phone.
We recommend what suits your task, even if it is the smaller solution.
AI use cases for businesses
Every department has tasks that repeat every day. AI automation takes over the following directly in the programs your team already uses. That leaves your employees more time for the work where people make the difference.
Sales
- Assess inquiries from the website and pass them to the person responsible
- Enter contacts and call notes into the CRM
- Prepare quotes from your prices and templates
Customer service
- Sort emails and tickets by request and forward them
- Answer recurring questions in the chat and on the phone
- Write suggested replies from your documents
Accounting
- Read incoming invoices and match them with orders
- Assign receipts to the right transactions
- Send payment reminders
HR
- Set up new employees in all programs
- Schedule job interviews
- Answer employees' questions about vacation and internal rules
Purchasing and logistics
- Read order confirmations and delivery notes from suppliers
- Create reorders as soon as a minimum stock level is reached
- Track shipments and inform customers about the status
Marketing
- Write drafts for texts, newsletters and social media posts
- Create product images and video variants for several channels
- Compile reports from your campaign data
AI solutions for your company
From the first overview to the assistant on the phone, the same team plans, builds and looks after your AI solutions. Often several of them interlock, such as a chatbot that creates cases in the ERP through an automation.
Local AI
Your own AI on your own servers. Your data stays with you.
- Llama, Qwen, Mistral and more
- GPU servers or on-premises
- Training on your data
NVIDIA- Llama
Qwen- Mistral
Ollama
AI Agents
AI that looks things up, creates cases and prepares replies.
- Reads incoming requests
- Works in CRM and ERP
- In your own software on request
- OpenAI
- Claude
- HubSpot
- Salesforce
- SAP
Process Automation
Recurring workflows between your programs, with no manual work at all.
- n8n, Make, Zapier, Power Automate
- Starts on a form, appointment or order
- AI for freely worded emails
n8n
Make
Zapier- HubSpot
Slack
AI Chatbots
An assistant answers questions, based on your own content.
- Website chat
- Internal knowledge base
- Answers from your documents
- OpenAI
- Claude
- Gemini
- Mistral
- Slack
AI Phone Assistant
Every call is answered, around the clock.
- Even when several people call at once
- Appointments in the calendar
- New contacts in the CRM
- OpenAI
- Claude
- HubSpot
- Salesforce
Pipedrive
Document Processing
Invoices and forms are read, checked and passed on, without retyping.
- Incoming invoices
- E-invoice (XRechnung, ZUGFeRD)
- Contracts and forms
DATEV
Lexware Office
sevDesk- SAP
- Odoo
What AI automation looks like in daily use
Choose a workflow: an inquiry in sales, a call after business hours, a new employee's start or an incoming invoice. Each one shows step by step what the automation takes over and what changes compared to today.
An inquiry from the website to the quote
A prospect writes through the contact form. The AI reads the inquiry, the CRM receives a complete contact, and your sales team receives a draft quote.
TodayInquiries stay in the inbox until someone has time.
With automationEvery inquiry is in the CRM immediately, with all details.
TodaySomeone types the name, company and request into the CRM.
With automationThe workflow creates the contact or adds to the existing entry.
TodayEvery quote starts with a blank document.
With automationYour sales team reviews a finished draft and sends it off.
Inquiry
The inquiry arrives through the form or by email.
Assess
The AI works out what the prospect needs and classifies the inquiry by your criteria.
Record
The workflow checks whether the contact is already in the CRM and creates it if not.
Quote
A draft quote is created from your prices and templates, and the person responsible finds it in the CRM.
A call after business hours becomes an appointment
A new customer calls in the evening. The phone assistant picks up, books an appointment and creates the contact in the CRM.
TodayAfter business hours, only the answering machine picks up.
With automationEvery call is answered and the request is handled in the conversation.
TodayWhen several people call at once, some hear a busy signal.
With automationThe assistant talks to several callers at the same time.
TodayCallback requests and appointments are on notes or in emails.
With automationThe appointment and contact are in the calendar and CRM immediately.
Answer
The assistant answers the call, even when your office is closed.
Understand
It asks about the request and takes down the name and phone number.
Book appointment
It suggests free times from your calendar and enters the chosen appointment.
Record
Contact, request and appointment are in the CRM, and your team sees everything the next morning.
New employees from contract to first working day
An employment contract is signed. The workflow sets up the new employee in your programs and prepares the first working day.
TodayThe same details are typed into several programs.
With automationThe details from the contract are in every program that needs them.
TodayOn the first day, access is missing because a request got stuck.
With automationThe tasks for access and equipment start on the day of signing.
TodayThe first day is coordinated by email among several people.
With automationThe welcome email and first appointments go out automatically.
Contract
The signed contract arrives by email or from your HR program, and the AI reads the name, start date, department and role.
Access
The workflow creates the email account and access to the programs that belong to the role.
Equipment
A laptop and work equipment are ordered, and your IT receives a task with all details.
First day
The new employee receives a welcome email with the first appointments, and the appointments are in the team's calendars.
An incoming invoice from the inbox to the ERP
The AI reads the invoice, your rules check it, and your ERP receives a finished posting proposal.
TodayInvoices arrive as PDFs by email, and someone types in the number, date and amount.
With automationThe details are already extracted in the system before anyone opens the invoice.
TodayTo match it, someone looks for the corresponding order in the ERP.
With automationThe workflow matches the amount and line items with the order.
TodayQueries go back and forth in long email threads.
With automationThe workflow flags deviations from the order.
TodayReceipts are scattered across folders and inboxes.
With automationEvery invoice is archived with its transaction.
Sample workflow: incoming invoice
Incoming
The invoice arrives as a PDF by email and is recognized as an incoming invoice.
Extract
The AI finds the invoice number, date, amount and order number.
Not needed, because the data is already machine-readable.
Check
The workflow matches the amount and line items with the order in the ERP.
The workflow checks the format and matches the amount and line items with the order.
Hand over
Your ERP receives a posting proposal with all details.
Archive
The invoice is filed with its transaction and can be found at any time.
If the invoice arrives as XRechnung or ZUGFeRD, the data is already machine-readable. Then extraction is not needed, and the workflow checks the format. Check an e-invoice for free
The systems and AI models we work with
An AI solution has to work with the systems you already use. For AI integration, we therefore first check which interfaces your systems offer. In the quote, we name the technologies we propose and give the reason for each choice.
AI models
Models for texts, receipts, images and videos, from the cloud or locally on GPU servers.
Cloud
- Claude
- OpenAI
- Gemini
Open models
- Llama
- Qwen
- Mistral
DeepSeek
Kimi
MiniMax
Xiaomi MiMo
Gemma
GLM
Images and videos
FLUX
Stable Diffusion
LTX-Video- Wan
AI development
Tools we use to train models, deploy them on GPU servers and integrate them into applications.
- NVIDIA
- CUDA
PyTorch
TensorFlow
Jupyter
vLLM- llama.cpp
- Ollama
LangChain
Your systems
ERP, CRM and accounting, which a workflow reads from and writes to.
- SAP
- Odoo
- HubSpot
- Salesforce
Workflows
Tools we use to set up recurring workflows between your programs.
- n8n
- Make
- Zapier
- Power Automate
Operations
Servers and cloud on which your workflows run.
AWS
Azure
Google Cloud
Hetzner
Channels
Where requests and documents reach you.
- Email inbox
- Website
- Phone
- Forms
- Scans
AI and data protection: GDPR and the EU AI Act
For the data protection of your AI solution, we look at the whole path of your data, from your system through the workflow to the AI model. We clarify with you where your data is stored, including the input an AI processes.
Data location
By default, your systems run on servers in the EU, preferably in Germany.
AI models
With external providers such as OpenAI or Anthropic, we check before use where they process your input. On request, we run open models locally.
Local AIContract
For personal data, we conclude a DPA under Art. 28 GDPR. Please direct questions about data protection to our data protection contact.
Access
A role and permission model defines who accesses which data and functions.
Protection
We encrypt data in transit and at rest, back it up regularly and test its security.
EU AI Act
For every application, we check which role you have under the AI Act and which obligations follow from it.
AI automation costs: what the price depends on
You know the price before implementation begins. For the implementation, you pay a fixed price per project; for support, a flat monthly fee.
Quote
Implementation, one-off
Fixed-price project
- Fixed price for the precisely described scope
- If the scope changes, we adjust the quote first
- No billing by the hour
Launch
After the launch, monthly
Support and further development
- Flat monthly fee for maintenance and support
- Phone support around the clock
- New tasks per work package, each with its own quote
What the scope depends on
- How many work steps and special cases the workflow covers
- Which systems are connected, such as ERP, CRM or accounting
- In what form your data and documents are available
- Which channels the solution works on: email, website, internal programs or phone
- In how many languages it reads and replies
- Where the AI model runs and whether we train it on your data
If you have a budget range, mention it in the initial consultation. We align the scope of the first task with it and plan further tasks for later steps.
Costs during operation
In operation, you pay for licenses, hosting and maintenance, and with a provider's model also for every request to the model. An agent often queries the model several times per task, a fixed workflow less often or not at all. In the quote, we calculate these items over several years and check whether there is a cheaper way to the same result.
What you own at the end
After full payment, we hand over everything you need to run your workflows yourself or have them developed further by another team.
- Code
- The source code developed for you
- Git repository with the complete change history
- Technical documentation
- Workflows
- The workflows we set up, for example in n8n or Make
- Instructions to the AI (prompts)
- Test cases
- Access
- Servers and hosting
- Domains
- Accounts for all services
- Rights
- Rights of use
We set up code, workflows and documentation so that a new team can also find its way around. With the prompts and test cases, your workflow can also be switched to another AI model. That way, you decide yourself who continues to look after your solution and which model it works with.
ASES in numbers
- 1 teamOur own team, no subcontractors
- 15+More than 15 years of experience on the team
- 10Conversations in ten languages
Questions people ask an AI agency
Answers on getting started, costs, data, the EU AI Act and running your AI solution.
With a task your team repeats every day, such as capturing receipts, sorting inquiries or transferring data from one program to the next. For the initial consultation, a short description of this task, a typical example and the names of the programs involved are enough. If you are not yet sure which task it should be, we find it together in a workshop.
If the rules are fixed, a fixed workflow is enough. If the same questions keep coming up and the answers are in your documents, a chatbot fits. If a task consists of many steps in several programs and the inputs look different every time, we use an AI agent. We clarify what suits your task in the initial consultation, using an example.
Above all, work in which your team reads documents, transfers data or answers similar inquiries. Such tasks exist in every department, from sales, customer service and HR to accounting. With local models, images and videos can also be generated.
It answers calls, understands the request and handles it in the conversation: it books, moves or cancels appointments and answers recurring questions. For this, we connect it to your calendar and your CRM. That way, it sees free times and records every caller with their request. For urgent calls, it connects immediately to a person you designate, such as your on-call service.
For the implementation, you receive a fixed-price quote as soon as the scope is set. The scope depends on the work steps and special cases, the connected systems and the form of your data. In operation, you pay for hosting and licenses, with a provider's model also for every request, and, if you like, a flat monthly fee for maintenance and support.
By default, your systems run on servers in the EU, preferably in Germany. For personal data, we conclude a DPA under Art. 28 GDPR. For every external AI model, we check before use where it processes your input, and we discuss the result with you. If your input should not be processed by any AI provider, we use local AI.
Yes. For many tasks, it is enough for a chatbot or agent to search your documents for every question and formulate the answer from them. The model itself stays unchanged. If it should learn your technical language, your formats or your style, we train an open model on your data, by fine-tuning with LoRA or QLoRA in PyTorch. Before use, we check it against test cases. Then it runs as local AI on a GPU server.
When we define the scope, we already set which steps run by fixed rules and where an AI reads or writes texts. Before the launch, the workflow runs through the test cases for its intended use. Every later change can be checked against the same test cases.
In the scope and quote phase, we check for every application which role you have under the AI Act, for example as provider or deployer, and which obligations follow from it. This includes, for example, that your customers are told when they are talking to an AI. We discuss the result with you before implementation begins.
Yes. First we check which interfaces your systems offer, which data should flow and which system is the source of truth for customer data and prices. Then we connect the workflow, for example, with SAP or Odoo, with HubSpot or Salesforce and with your accounting. After that, nobody has to type in details from invoices, inquiries and orders anymore.
In the cloud, we work with Claude, OpenAI and Gemini; locally, with open models such as Llama, Qwen, Mistral, DeepSeek or Kimi on GPU servers. We build the workflows so that the model can be changed later without redeveloping them. Since you receive the workflows, prompts and test cases from us, you are not tied to any provider.
Your team is involved from the start and gives feedback on the work in progress. At launch, we train everyone who works with the solution. Beyond that, we strengthen your employees' AI skills with AI training per role and write an AI policy for your company with you.
More on AI and automation
- APIs & IntegrationOrders, customer data and stock levels flow between your systems without anyone typing them in.
- Software DevelopmentYou get software that maps your processes, and the source code with it.
- IT Consulting & StrategyYou know which technology suits your project before you invest in development.
- E-Invoice ValidatorChecks your XRechnung or ZUGFeRD invoice for free against the official KoSIT rules and marks every error at the affected field.
Let's talk about your automation
Tell us which task costs your team time today and which programs it works with. We will get back to you as soon as possible and find a date for the initial consultation with you.
- Free initial consultation
- No obligation
- NDA on request