Every company collects data.It rarely reaches the people who need it.
Most AI agents fail because of the data, not the model. We build a clean data foundation first, then bring the answers straight to your employees through an agent.
The real payoff
Answers across all your data sources
A single system only ever shows one piece of the picture. Our agents link sales, service, IoT telemetry, production and customer reviews, and find connections that no single source reveals.
Why is revenue from series B7 falling in southern Germany?
- SalesRevenue down 18% year over year, mainly among 9 existing customers.
- ServiceThese customers had three times as many service calls.
- IoTThe trigger is error E-217 in 70% of cases.
- ProductionAffected units come from batch 24-11 with a changed supplier part.
- ReviewsCustomers describe the same problem 38 times.
Use cases
Agents for every department
What has happened on unit M-0412 since the last maintenance?
14 × error E-217, temperature recently above limit, ticket 48214 still open.
Service & Support
Before a visit, technicians see the machine's error history, usage and open tickets.
Which errors pile up on series B7 after 1,500 operating hours?
E-217 and E-104 account for 61% of the messages. Reviews describe E-217 38 times.
Product management
Field data and customer feedback show how devices are used and where errors pile up.
Which customers in southern Germany have reduced their volume compared with last year?
12 customers, 4 of them with open service cases on series B7.
Sales
Revenue, customers and service history in one answer instead of three exports.
Build the service cost report by region as slides and send it to division management every month.
Report with 6 slides created. From now on it goes to division management on the first of each month.
Controlling & Reporting
Metrics with one shared definition. Reports on request or automatically by email.
Notify me when the error rate of a series rises above 2%.
Alert set up. The error rate is checked daily at 6:00 a.m.
Monitoring & Alerts
Agents watch thresholds and report anomalies before anyone asks.
Example questions and answers for illustration.
One question. Three layers.
This is how an agent works on a clean data foundation: raw data becomes a documented data model, and the agent answers the question on top of it.
Which machines had a cluster of E-217 errors last week?
Three machines are above the threshold. All belong to the same series and have more than 1,900 operating hours.
| machine_id | series | week | error_code | count | operating_hours |
|---|---|---|---|---|---|
| M-0371 | B5 | 2026-45 | E-217 | 1 | 842,0 |
| M-0388 | B7 | 2026-45 | E-217 | 11 | 1.904,2 |
| M-0412 | B7 | 2026-45 | E-217 | 14 | 1.932,5 |
| M-0655 | B5 | 2026-45 | E-217 | 0 | 310,7 |
| M-0907 | B7 | 2026-45 | E-217 | 9 | 2.011,8 |
Services
From data source to agent, all from one team
Join at the layer where your company stands today, or hand us the entire build.
Start
Readiness Check
How agent-ready is your data, which use cases are worth pursuing, and where do you begin?
Layer 0
Platform & Enablement
Connect data sources, set up the cloud platform, and put clean access rules in place.
Layer 1
Data Foundation
Clean and link raw data, then turn it into documented, semantic data models.
Layer 2
Agent Foundation
Agents that answer questions, create reports and raise alerts when thresholds are crossed.
Run
Operations & Scale
Support in operation, cost control and rollout to more departments.
The starting point
Data & Agent Readiness Check
You find out how well your data suits AI agents, which use cases are worth pursuing and where to begin. Scope and schedule are agreed individually with your company.
Request a Readiness Check- Inventory of data sources and infrastructure
- Assessment of how agent-ready your data is
- Two to three prioritized use cases with expected benefit
- Roadmap with a concrete starting point
Principles
Reliable in operation
Experience from industry
We have built data platforms and data agents in industry that are in production today. Together we bring 12 years of professional experience in data engineering and AI.
User permissions
Every query runs with the permissions of the person asking. The agent never sees more than the user.
Semantic layer
Metrics with a single definition instead of queries on raw data. The same numbers everywhere.
Cost limits
Fixed caps per query keep cloud costs predictable.
Tests before launch
Before an agent goes live, we test it with questions whose answers we already know.
EU hosting
Hosting and AI models in European data centers.
Google Cloud
Focus on BigQuery, dbt, Vertex AI, Gemini and ADK. The approach carries over to Azure, AWS, Snowflake and Databricks.
How a project works
Intro call
We talk about your data, your goals and your constraints.
Readiness Check
Assessment of data sources and data quality, prioritization of use cases.
Data Foundation
Build the data layer for the first use case, documented and tested.
Agent pilot
A first department works with the agent. We support the pilot closely.
Operation and rollout
More data sources, departments and user groups are added.
The team behind 2ndLayer
We are two data & AI engineers from Stuttgart, and we work full time at Kärcher in IoT. Together we bring 12 years of experience in data engineering and AI. Peter builds the data foundation, Malte builds the agents on top of it.
This collaboration is how 2ndLayer came about. Your project gets exactly this division of work.

Layer 1Data Foundation
Peter Müller
Builds data models that people and agents can rely on.
Peter has worked in data engineering for eight years and was co-responsible for building an enterprise-wide data lake. His focus is data modeling: turning scattered raw data into dependable models that others can work with.
His mechatronics studies mean he knows machines from the technical side too. He knows what a sensor value or an error code means on the machine. That helps most with IoT data.
Master's in Mechatronics & Data Mining
LinkedIn profile (opens in a new tab)
Layer 2Agents
Malte Reßler
Builds agents that let departments analyze their own data.
Malte started in third-level support for connected machines. There he saw that IoT data existed but was hardly used: regional subsidiaries had to request every analysis centrally, and for service technicians, getting at machine data took too much effort.
Through data analytics he moved into building AI agents. What drives him is how everyday work changes when employees have direct access to data: people ask questions that nobody could ask before. What convinces him most is when users report that an agent has changed their work, and more colleagues ask for access on their own.
Cooperative degree in Industrial Engineering at ABB (B.Eng., DHBW Mannheim), M.Sc. in Business Informatics (Hochschule Reutlingen)
LinkedIn profile (opens in a new tab)
Our standard
Built for real use
We only build solutions we would use ourselves every day.
Measured by usage
A project succeeds when the department uses the system in daily work and gets real benefit from it.
Strengthening people
Agents do not replace employees. They improve processes and put decisions on a reliable data basis.
The era of AI agents has begun.Start with your data.
We start from your goals and end with an agent that your department uses every day.
Book an intro callStart with an intro call
Briefly describe your project, or book a time directly.