[DigitalToday reporter Chi-gyu Hwang] "Productive individuals do not make productive organisations."

George Sivulka (조지 시불카) of enterprise AI search platform startup Hebbia recently posted on social media platform X (Twitter) under that title, stressing that organisational intelligence matters for proper AI innovation beyond the individual level.

To explain, he cited the Industrial Revolution era.

He said, "In the 1890s, New England textile mills replaced steam engines with electric motors. The technology was much better. But productivity barely rose for 30 years. The turning point came in the 1920s. Only when they redesigned the factory layout itself and introduced assembly lines did the benefits of electrification appear. The key was not swapping the motor but rebuilding the entire factory."

AI is in a similar situation. He said, "As of 2026, individual productivity for people who use AI well has risen 10-fold. Company-wide performance is standing still. We replaced the motor and left the factory as it was."

He describes this problem through the concept of "Institutional Intelligence". He explained that the core of institutional intelligence can be summarised in seven points, and that this will determine the next 10 years of B2B AI.

First is coordination. He said, "If every employee in an organisation uses ChatGPT on their own and produces outputs in different ways, it is like thousands of oars pulling at the same time in different directions. Agents, like people, need clear roles, responsibilities and communication methods."

Second is signal. He said, "In an era when AI can generate anything, the real problem is choosing 'the right thing'. Most AI output is useless slop. Organisational AI must find signal in those piles and operate in predictable and verifiable ways."

Third is objectivity. He said ChatGPT and Claude are trained to habitually agree with what users say. If the lowest-performing employee in an organisation falls into the illusion that the smartest AI in the world is on their side, it poisons the whole organisation. He said, "Organisational AI must reinforce truth, not agreement."

Fourth is edge. He said, "Personal AI aims for general-purpose use. Organisational AI aims for an edge that outperforms competitors in a specific domain. Just as Midjourney focuses on images and ElevenLabs on voice, only solutions specialised in fields such as finance, law and healthcare deliver real results."

Fifth is outcomes. He said personal AI focuses on saving time, while organisational AI focuses on growing revenue. He said, "If you ask a company's CEO what comes first between cost-cutting and revenue growth, most answer revenue. In M&A, personal AI helps analysts work faster, but organisational AI can pick the true acquisition target among hundreds of candidates."

Sixth is enablement. He said, "Providing a tool and teaching how to use it are different. This is why Palantir maintains a high share price even amid a major crash in technology stocks. Palantir is not a software company but a 'process engineering' company. The future of organisational AI lies not in selling tools but in designing processes so organisations can actually use AI."

Seventh is unprompted execution.

He said the most valuable AI work is what no one asked for. He said the goal of organisational AI is to spot first, without being asked, that cash-flow indicators at a specific portfolio company have worsened for three straight months and alert the person in charge. He said, "Personal AI is the channel through which most companies in the world first experience AI. That role is clear. Without institutional intelligence, it remains a half revolution. Just as factories in the 1890s wasted 30 years by merely installing electric motors, organisations today are repeating the same mistake. The electricity is already here. Now it is time to redesign the factory."