Enterprise AI OS: A New Layer Between Agents and Legacy Systems
An enterprise AI operating system is an infrastructure software layer that sits between AI models and existing business applications, translating data, logic, and workflows so that AI agents can understand, query, and act on legacy systems without a full system replacement. The back-to-back funding rounds for PhoenixAI and Conduct show how quickly this idea is moving from theory into production. Instead of telling enterprises to rip out decades-old ERP, CRM, and operational software, these AI operating systems focus on making current environments AI-ready. That means faster legacy system integration, more reliable data access for AI, and a way to govern how autonomous agents interact with core systems of record. In practice, this emerging enterprise AI infrastructure acts like middleware for the AI era, connecting agentic workloads to complex, customised software landscapes.
PhoenixAI: An Agentic AI Database Built for Real-Time Workloads
PhoenixAI, formerly CelerData, raised USD 80 million (approx. RM368,000,000) in Series B funding to grow what it calls an Agentic AI Database. The platform is designed for AI agents that fire off thousands of unpredictable, real-time queries across live and historical data. Traditional databases expect carefully modelled questions; PhoenixAI aims to support chaotic agent traffic with sub-second responses, high concurrency, and strong governance. Its engine combines streaming and at-rest data, making Kafka updates queryable within seconds while keeping the control enterprises expect. Customers such as AppLovin, Coinbase, Conductor, and Demandbase are already running production workloads on the system. As President Rick Underwood explains, agents now swamp systems with unplanned questions that strain existing data stacks, and PhoenixAI wants to close that gap with AI-native data infrastructure rather than incremental tuning of legacy warehouses.
Conduct: An AI OS That Makes Opaque Systems AI-Ready
Conduct has raised €51 million (USD 60 million; approx. RM276,000,000) in Series A funding for an AI operating system that helps enterprises understand, operate, and change their existing software systems. Built by former Palantir engineers, the platform maps the dense business logic buried inside SAP, Salesforce, Oracle, MES, WMS, and other customised systems, turning millions of lines of code and undocumented dependencies into legible, actionable structures for humans and AI agents. According to Conduct, teams using its AI OS see more than 30% acceleration in transformation workstreams and time-to-value for new features. CEO Jan Philipp Haas argues that decades of customisation have made core systems too opaque for people, and impossible for agents, to work with. Conduct’s partnerships with SAP, BCG, and NTT DATA Business Solutions underline its aim to become essential enterprise AI infrastructure rather than a bolt-on automation tool.
Why Investors Are Backing AI Infrastructure Over Rip-and-Replace
Taken together, the PhoenixAI and Conduct rounds point to a clear investor thesis: the next wave of enterprise AI will be won in infrastructure, not in wholesale system replacement. PhoenixAI focuses on the data plane, with an agentic AI database that can feed autonomous workloads at speed. Conduct tackles the application and process plane, making ERP and other systems comprehensible and operable for AI. Both startups target the same enterprise pain: legacy system integration. According to EU-Startups, Conduct’s round sits alongside funding for agent governance, ERP modernisation, workflow automation, and industrial AI, where capital is moving into software layers that let enterprises deploy AI against existing systems. This pattern shows strong confidence in AI-ready software approaches that respect sunk costs, minimise disruption, and give CIOs a realistic path from pilot experiments to production AI operating systems.
