The layer between your systems.

Systems we work in

The list reflects the typical system landscape of mid-sized companies; it is not a list of partnerships. We are not tied to any vendor. Under each family is what we typically build there: with a client case, or explicitly marked as possible.

• CRM and salesSalesforce, Pardot, HubSpot, Pipedrive, Microsoft Dynamics 365 Sales, Zoho

  What we typically build thereReading enquiries, classifying them and creating them in the CRM with all details, cleaning up duplicates, enriching company data. Connecting to Salesforce often starts with sorting out junk data. At a spare parts dealer, a workflow filters placeholders, drawing numbers in the name field and discontinued products out of the Salesforce data before pages and texts are generated from it.

• ERP and inventoryNetSuite, SAP S/4HANA and Business One, Microsoft Dynamics 365, weclapp, Odoo, Sage

  What we typically build thereClassifying enquiries by market and country of origin and passing them to the responsible team in the sales pipeline. That is how AI runs with NetSuite at a technology distributor across twelve locations in Europe. Possible, with the same check before booking: reading orders and delivery notes from emails and creating them as records in the ERP.

• HR and recruitingPersonio, SAP SuccessFactors, Workday, softgarden, d.vinci, applicant management portals

  What we typically build thereBringing vacancies from the HR system onto the website, like the import of job ads from SAP SuccessFactors at a fuel cell manufacturer. Possible, not yet built for a client: a Personio integration that reads applications from the inbox and creates them in structured form, without assessing them. The decision stays with a person, and the works council receives the documents for it.

• Product data and PIMAkeneo, Pimcore, manufacturer databases, ETIM and BMEcat catalogues, Excel and CSV files

  What we typically build thereGenerating structured product data and texts from manufacturers' PDF data sheets, checked before they go live, as at a technology distributor. Possible with the same checks: checking ETIM or BMEcat catalogues for gaps and contradictions and proposing missing attributes.

• Communication and serviceMicrosoft 365 and Exchange, Google Workspace, Postmark, Brevo, Freshdesk, Zendesk, phone systems with an interface

  What we typically build thereTranslating service content according to a technical glossary, at a fuel cell manufacturer in eleven languages, filling only empty fields and leaving checked texts untouched. In our own operations, calls are captured and created as records with customer, request and next step.

• Web, shop and contentWordPress, TYPO3, Shopware, Magento, custom portals, Cloudflare

  What we typically build thereManufacturer texts in a weekly rollout, with eleven checks before every go-live and a kill switch, as at a spare parts dealer. Checking text inventories across many domains against rules and proposing rewordings for a person to approve, as in the EmpCo check of ten domains.

• Data and analyticsMySQL, PostgreSQL, Microsoft SQL Server, Google Search Console, Matomo, Plausible, Power BI

  What we typically build thereMonitoring that reports only deviations: counters, sitemaps, daily clean-up. In our own operations, a check runs every night across the interfaces of dozens of client systems. Only deviations are reported.

• AI modelsAnthropic Claude, OpenAI, Google Gemini, Mistral, open models such as Llama run locally, vector databases for knowledge bases

  What we typically build thereThe right model for each workflow: large cloud models for difficult language, small or open models for volume and for data that stays on your own servers. At a technology distributor, the knowledge base sits on its own infrastructure, while the chat in five languages runs on a cloud model under contract. If the model changes, the rules and checks stay.

Your system is not on the list? If it has an interface, we can work with it. We have been doing that since 1998: reading systems, understanding them, connecting them. Getting to know your system is part of the build, not a separate item.