From Chatbots to Enterprise AI Infrastructure for Agents
Enterprise AI infrastructure for agentic systems refers to the databases, workflow engines, developer platforms, and operations tools that let autonomous AI agents operate at scale across business-critical workflows, from marketing and manufacturing to incident response and in-house legal work, while meeting stringent requirements for reliability, governance, and integration with existing software stacks. The latest funding wave shows investors backing the unglamorous layers that make this possible. PhoenixAI has raised USD 80 million (approx. RM368 million) for an agentic AI database built to handle thousands of unpredictable, real-time queries from AI agents, unifying live and historical data in a single engine. Gradial secured USD 65 million (approx. RM299 million) to expand an AI agent platform that runs end‑to‑end marketing workflows inside enterprise constraints. Alongside them, vertical and horizontal platforms such as Limitless Labs, Niteshift, Sandstone, and Bland are pushing agentic AI into CAD/CAM, coding, legal operations, and voice interactions.
PhoenixAI and the Rise of the Agentic AI Database
PhoenixAI, rebranded from CelerData, is positioning its platform as an agentic AI database: an AI‑native data engine designed for workloads where agents fire off thousands of unplanned, real‑time queries. Traditional analytics stacks expect humans asking pre‑modeled questions; PhoenixAI instead joins live streams and historical records on demand, while tying into Apache Iceberg lakehouses and Kafka pipelines. Customers such as AppLovin, Coinbase, Conductor, and Demandbase already run PhoenixAI in production, reporting sub‑second response times across hundreds of millions of rows. According to PhoenixAI president Rick Underwood, “agents now fire off thousands of unplanned, real-time queries, often swarming systems with questions that weren’t anticipated when the data stack was designed.” The USD 80 million (approx. RM368 million) Series B led by Sky9 Capital signals that agent‑grade data performance and governance are becoming baseline requirements for AI agent platforms that have moved from prototypes to mission‑critical deployment.
Agentic AI in Marketing, Legal, and Voice Workflows
On the application side, funding is flowing into AI workflow automation where agents take over complex, multi‑step work. Gradial’s USD 65 million (approx. RM299 million) Series C backs a “system of work” for marketing in which agents author, QA, ensure compliance, tag, and publish content across existing CMS and media tools, shrinking time‑to‑market from days to same‑day for some customers. Sandstone’s USD 30 million (approx. RM138 million) Series A extends similar AI workflow automation to in‑house legal teams, triaging requests from Slack, email, and Jira before running drafting and analysis flows tuned for smaller legal departments. Bland’s USD 50 million (approx. RM230 million) Series C targets voice AI built for 30‑ to 45‑minute calls across regulated industries, handling more than 3.5 million calls per week for over 250 enterprises. Together, these AI agent platforms show investors backing deep domain workflows rather than generic chatbots.
Infrastructure for Coding and Manufacturing Agents
Below the application layer, new enterprise AI infrastructure startups are forming around specific kinds of agentic work. Niteshift has raised USD 7 million (approx. RM32 million) as a model‑agnostic routing layer for AI coding agents, arguing that enterprises will not want to send their most sensitive codebases through vendors that also sell vertical software products. Instead of replacing tools like Claude Code, Niteshift charges per‑minute infrastructure fees and positions itself as a cloud‑style substrate for coding agents. In manufacturing, Limitless Labs secured USD 20 million (approx. RM92 million) for an agentic CAD/CAM platform that embeds a CAM Agent into existing tools such as Creo, Siemens NX, and Mastercam. The company claims it can cut programming work in half by standardizing best practices and scaling expert machinists’ knowledge. Its “Physical AI Foundation Model” is trained on the physics of metal cutting, CAD geometry, and real machine constraints.
M&A Signals: Elastic, DeductiveAI, and the Next Phase of AI Ops
The funding surge is matched by a notable exit that underlines how fast AI‑native ops tooling is becoming strategic. Elastic is acquiring DeductiveAI, an AI site reliability engineering startup, for up to USD 85 million (approx. RM391 million) after the company raised only a USD 7.5 million (approx. RM35 million) seed round. DeductiveAI built agents that connect directly to code, logs, metrics, traces, and events, reasoning over a live knowledge graph to surface root causes and drive autonomous incident resolution. Early customers such as DoorDash and Foursquare reported up to 90% reductions in incident resolution time, with DoorDash alone saving more than 1,000 engineering hours annually. Elastic plans to fold these capabilities into its observability stack, which already includes earlier AI operations deals and agentic Kubernetes workflows. This acquisition, alongside funding rounds for platforms such as Default and other enterprise AI automation vendors, suggests AI‑native incident resolution and workflow automation are becoming acquisition currency.
