Latest Insights: Data Integration | Process Intelligence | Trusted Agentic AI | Enterprise Architecture
Kai Waehner is a Global Field CTO writing about data integration, workflow orchestration, process intelligence, trusted agentic AI, and enterprise architecture. Articles cover real customer use cases across industries, from data streaming with Apache Kafka and Flink to workflow automation, process mining, and governed AI agents. Posts look at architecture patterns and vendor trade-offs across the market, not a single product.
Apache Kafka and the Database as Source of Truth for AI Agents
The source of truth for AI agents is a combination of systems in an event-driven architecture. The database is the system of record, where every transaction happens. Apache Kafka is the way in and out: events, a real-time view of the…
Why the Lakehouse Vendors and the Clouds Bet on Postgres for AI Applications
Five vendors with five strategies chose the same database underneath. What the Neon and Crunchy Data deals tell you, where Postgres reaches its limits, and why Oracle, SAP, IBM, and OceanBase build the same capabilities into their own…
Unified Orchestration in Aviation: One Control Plane Across the Airline
Airlines run data pipelines, infrastructure automation, application workflows and business processes in separate orchestration silos, each with its own scheduler and audit trail. Unified orchestration in aviation puts them under one…
Agentic Technical Debt and Where Orchestration Belongs in AI Agent Architecture
AI agents pile up debt when retries, approvals and audit land in the wrong layer. A guide from model and harness to durable execution and the control plane, with a worked invoice dispute….
My Lufthansa HON Circle Experience in 2026
My Lufthansa HON Circle experience in 2026 as a frequent flyer with around 30 trips a year. The new HON service, First Class upgrades and lounges worldwide work well. Boarding groups, feeder flights from Nuremberg and the group’s IT…
Why Physical AI Value Depends on Data Streaming and Workflow Orchestration
Humanoids get the headlines, but most physical AI value comes from what happens after a machine detects something: quality holds, work orders and approvals across MES, ERP and asset management….
How Decision Models Became a New AI Category After Jev
Within a month of Jev by TypeSafe AI, hyperscalers, AI labs, data platforms and open source projects shipped their own decision models. The request schema became the standard, the model became interchangeable, and benchmark claims…
Event-Driven Workflow Orchestration on Apache Kafka with Kestra
Kafka moves the events, but nothing coordinates what happens next. This post shows how Kestra runs on the Kafka cluster you already operate, turns topics into workflow triggers that reach ServiceNow and RabbitMQ, and where Flink or…
Industrial IoT Landscape Q4 2026: The Plant Floor Is Becoming Software
The Industrial IoT Landscape Q4 2026 maps vendors and open source from edge to cloud in four categories, explains the Unified Namespace shift, and shows where OT meets enterprise IT….
Unified Orchestration in Telecom: One Control Plane Across OSS, BSS, and IT
Telecom has more orchestration standards than any other industry, and the maturity stops at the network boundary. This post maps unified orchestration to the operator: the four silos, the tools in each, the billing cycle, copper…
Jev by TypeSafe AI in Workflow Orchestration and Process Intelligence
Jev by TypeSafe AI returns a calibrated probability instead of text. In a workflow that becomes a confidence gate: high confidence proceeds, medium goes to an LLM, low goes to a person, and every decision lands in a log. This post…
Why Jev by TypeSafe AI Fits Apache Kafka and Flink Better Than LLMs
Jev by TypeSafe AI returns typed decisions with calibrated probabilities in a few hundred milliseconds. This post shows how a System One model fits per-event decisions on Apache Kafka and Flink, how Flink prepares the state, how…
How System One Models Like Jev Change Enterprise AI Architecture
System One models such as Jev by TypeSafe AI answer typed questions with calibrated probabilities in a few hundred milliseconds instead of generating text. This post explains how they compare with rules, trained machine learning and…
From Cron Batch Jobs to Kestra: Migrating a Python Script Without a Rewrite
A working Python cron job moved to Kestra, the open source workflow orchestration platform, without a rewrite. Wrapping it unchanged surfaced a bug that had cost forty-five minutes every morning for weeks, and a later incident showed…
Data Streaming Trends Q3 2026: What Changes Through 2027
The Q3 2026 edition of the data streaming trends names ten trends through 2027, grouped by market, architecture, and practice: consolidation and migration tooling, sovereignty, diverging Kafka implementations, diskless Kafka, the…
Unified Orchestration in Financial Services: One Control Plane Across the Bank
Banks run four disconnected orchestration silos: data pipelines, batch schedulers, application workflows, and business processes, held together by homegrown glue code. With T+1 settlement, instant payments, and DORA enforcement…
What is Kestra? The Open Source Platform for Unified Orchestration
Kestra is the open source platform for unified orchestration: one control plane for data, infrastructure, applications, and business processes, with AI agents governed across all four. This article explains what unified orchestration…
Data Streaming Landscape Q3 2026: Who Controls Your Streams
The Data Streaming Landscape Q3 2026 maps every major streaming vendor and open-source project by workload and operating model, explains why data sovereignty now shapes deployment, and covers Confluent inside IBM and the lakehouse…
The AI Agent Harness: Where Vendor Lock-in Went After the Model Became Swappable
An AI agent harness is everything between the model API and the business outcome. Models became swappable, so vendor lock-in moved into the harness: context, definitions, and runtime. This post gives enterprise buyers the portability…
Unified Orchestration in Manufacturing: From Shop Floor to Cloud
Factories do not lack automation. They lack coordination between automations. This post maps unified orchestration to manufacturing and logistics: data pipelines from OPC UA and MQTT to the lakehouse, Python orchestrators like Airflow,…
When NOT to Use Stream Processing: Messaging and APIs Are Often Enough
Confluent rearchitected Control Center away from its Kafka Streams metrics pipeline. Kestra 2.0 rebuilt its queuing layer around plain queueing. Two companies that know streaming as well as anyone reached the same conclusion about the…
OceanBase vs. Databricks vs. Snowflake: Which AI Database for Transactional Workloads, Analytics, and AI Agents?
Every vendor now sells an AI database. This deep dive compares OceanBase with Databricks and Snowflake across transactions, analytics, and AI agents, including when not to use each….
Multi-Model AI Orchestration: Which Layer Should Pick the Model?
Every router can pick a model. The question that decides whether a multi-model strategy survives production is which layer chooses, on what grounds, and with what audit trail. Decide high, route low….
Why Enterprises Need a Multi-Model AI Strategy: Cost, Compliance, and Resilience
Single-vendor AI has become a single point of failure. Five drivers now make the case for a multi-model AI strategy across SaaS frontier models and self-hosted open weights, and the architecture underneath decides whether it works….
Trusted Agentic AI Landscape Q3 2026: Enterprise Vendor Selection, Sovereignty, and Lock-in
The Q3 2026 edition of the Trusted Agentic AI Landscape maps enterprise AI vendors on two axes, trust and lock-in, and explains why sovereignty, open weights, and agent-layer lock-in now decide the vendor choice. Free PDF with the full…
Data Integration vs Workflow Orchestration: Connecting Systems Is Not Coordinating the Work
Data integration and workflow orchestration get confused because both ship hundreds of connectors. This post draws the line: integration moves and reshapes data, orchestration coordinates what runs across categories and recovers when…
Process Intelligence Landscape 2026: Mining, Orchestration, and the Agentic AI Shift
Process intelligence has become three things, not one: mining, orchestration, and a decision gate. Here is how they fit, why agentic AI raises the stakes, and the four choices that keep your options open. Full vendor landscape as a free PDF….
When to Use AMQP, JMS, Kafka, or MQTT: Trade-offs, Not a Winner
AMQP, JMS, Kafka, and MQTT get compared as rivals, but a message broker, a log, and a device protocol sit on different layers. A clear guide to what each is, how they differ, when to use which, and why one platform often supports…