From Sun to workflow code: Scott McNealy backs son's agent-orchestration play
Key points
Key takeaways from the theCUBE + NYSE Wired interview with LittleHorse founder and CEO Colt McNealy and Sun Microsystems co-founder Scott McNealy (October 2026):
Business as code. LittleHorse was founded in early 2022, before the current AI wave, to solve the fragmentation of business logic across the dozens of systems a real enterprise touches. Codifying that logic as code makes it possible to change, observe and optimise processes rather than bending the business to fit the software.
A four-year family loan. Colt McNealy started building the platform while working as a software engineer at a real estate company, frustrated by the tangled workflow from house viewing to closing. His father Scott McNealy funded him with a loan for roughly four years, a manoeuvre he compares to persuading James Gosling not to leave Sun Microsystems, which led to Java. Colt finished Stanford computer science in three years, coded solo for the first 18 months, then built a team around his first hire, Saul; the product has been live to market for about 18 months.
Agents need an action layer. Every board is telling its CEO to deploy AI agents, but software vendors only automate within single systems such as CRM or ERP. Orchestrating across the whole business is the wall customers hit, which is why agentic orchestration is behind LittleHorse's recent traction.
Deterministic core, intelligent steps. LittleHorse codifies the deterministic parts of a workflow in ordinary languages such as Java, Python, Go, C and JavaScript, with version control, testing and auditability, then uses agents as steps for the fuzzy parts, such as reading an email or generating a report. Calculating tax on an e-commerce order, by contrast, must stay deterministic and must never be left to an LLM spending tokens.
Real-time event capture. The engine, built on Kafka streams and databases, captures every event in real time across all workflows. Scott McNealy argues this event data can train a proprietary CEO chatbot delivering a red, yellow and green state of the business each morning, and that side-by-side implementations let a company compare a freshly coded process with the legacy software stack before switching the old systems off.
Proof at one customer. A logistics and vegetation-management customer started with a single workflow for work-order management of a truck fleet and reached 268 processes on the platform within 18 months, covering employee onboarding, tax reporting, job bidding and weather-event handling.
A Java moment? Scott McNealy likens LittleHorse to Java becoming the lingua franca for distributed network applications, arguing the company is building the vernacular for codified, observable business workflows, and that companies will eventually move all of their computing into workflow form.
Read more: theCUBE + NYSE Wired on YouTube
