Examples from practice

Local AI / AI as a Service

Local AI in the company for SMEs

Situation

Small and medium-sized enterprises want to use AI for knowledge bots, document analysis or assistant systems – but sensitive customer, production and contract data must not leave the company. Cloud APIs bring running costs per request, dependence on a single vendor and open questions about data protection.

Solution

We set up a company-owned, locally hosted AI: on hardware from Apple, Nvidia or AMD, from a powerful workstation to your own server rack. Even very large language models run this way inside the company. DATANOMIQ supports the setup of the environment, the fine-tuning of models on your terminology and documents, and the creation of model ensembles in which several specialised models work together. If you prefer not to operate hardware, the models are hosted in a data centre in Switzerland – or both are combined in a hybrid approach.

Impact

The company runs its own AI use cases with full data sovereignty and predictable costs – and can offer AI functions as AI as a Service, internally for departments or externally for customers. Independent of individual vendors, scalable from the first pilot to regular operation.

Data Science / Machine Sensor Analytics

Detecting anomalies in chemical production: machine sensor data analysed with machine learning

Situation

A global chemical manufacturer faced inefficiencies due to non-transparent production data and increasing process requirements. Sensor, control and order data existed, but it was spread across plants, the control system and the ERP and was never analysed together.

Solution

We consolidated the plants' sensor data together with control and MES data in a cloud data platform and analysed it with machine learning: predictive analytics detects process anomalies before they lead to scrap or downtime and derives optimised machine settings. The results flow back to the control room and plant operators as setpoints and alerts – closing the data loop from the field level to the cloud.

Impact

Plant efficiency increased by 10 to 20%. On top of that: lower costs, higher production volume and improved energy and quality management.

Data Platform / Logistics & Supply Chain

Supply chain for retail: a data lakehouse as the single source of truth for stock, routes and delivery times

Situation

A retailer with several warehouse locations and its own delivery fleet lacked a unified view of stock, orders, routes and suppliers. The data was spread across ERP, warehouse management, transport planning and shop systems; dispatching and route planning ran on manually maintained spreadsheets. The result: supply shortages, empty runs and unreliable delivery times.

Solution

We built a central data lakehouse on a cloud data platform and consolidated all internal and external data – ERP, warehouse management, transport planning, shop and marketplace interfaces, supplier data – in structured layers. On this basis run demand forecasts, route optimisation for picking and delivery tours, and a real-time reconciliation of stock, vehicles and staff. Dispatchers see bottlenecks before they arise.

Impact

Up to 30% cost savings and 40% time savings through route optimisation and better coordination between resources. As a result: fewer supply shortages, more reliable delivery times and one consistent data basis for reports and decisions across the company.

Data & Process Transparency

Optimisation potential in manufacturing logistics identified with process mining

Situation

For a mid-sized mechanical engineering company, production lead times are critical: customers choose primarily by the shortest delivery times. The processes ran across purchasing, manufacturing, assembly, quality inspection and logistics – where time was being lost could not be seen in the standard reports.

Solution

We built a purpose-developed data warehouse that generates an event log from ERP, MES and WMS – optionally extended with data from the CRM and from external systems such as logistics partners and suppliers. These event logs are far more granular than is common in business intelligence: every order, every operation and every handover is captured with a timestamp. On this basis, process mining runs as continuous process monitoring and analyses the manufacturing and assembly processes together with purchasing and logistics.

Impact

Unnecessary process loops were eliminated and bottlenecks identified; additional operations and waiting times were resolved in around 50% of cases. In quality inspection in particular, the processes also became more transparent – and thus audit-proof.

Enterprise AI / Knowledge Bot

Knowledge bot for industry and steel production: an LLM with access to ERP and document data

Situation

A large industrial and steel production company works with a large number of documents – requirement documents, specifications, purchasing and sales records – spread across file shares, document management and the ERP. With staff turnover, knowledge was lost; searching for comparable cases and checking new documents against existing ones tied up a lot of engineers' and managers' time.

Solution

We built a multi-layered enterprise AI: documents and ERP data are vectorised and stored in a knowledge base; a large language model answers questions and compares documents on this basis – always citing its sources. A data access and governance layer ensures that the permissions of the source systems apply and that every answer remains traceable. The model ran on Microsoft cloud infrastructure (IaaS); alternatively, the same architecture can be operated with locally hosted models (see the use case “Local AI in the company for SMEs”).

Impact

Managers and engineers compare new documents with earlier versions 75% faster, create drafts from keywords provided by the business units and find knowledge even when the responsible person has left the company or moved to another department.

CFO Advisory / Financial Data Quality

Digital CFO advisory: audit readiness and financial data quality in real time

Situation

From the perspective of the CFO and controlling, the quality of financial data worries many companies: postings from several systems, manual corrections and late closings leave it open whether the figures will stand up to a financial audit, a tax audit or a due diligence. Data transparency usually only emerges once the auditors are already on site.

Solution

DATANOMIQ relies on AUDAVIS, a software also founded by Benjamin Aunkofer and developed together with auditors. AUDAVIS gives auditors and CFOs a real-time overview of financial data and establishes audit readiness proactively – for financial audits, tax audits and due diligence. In addition to standard audit procedures, it includes trend analyses, benchmarks, liquidity KPIs and the reconciliation of financial and material movements. DATANOMIQ handles the connection to ERP and accounting, the data preparation and the interpretation of the results together with the CFO team.

Impact

Financial data quality becomes continuously measurable instead of being checked once a year: anomalies are detected before they turn into audit findings, closings and audits run faster, and CFO and controlling decide on a transparent, audit-proof data basis.

• AUDAVIS – www.audavis.ai