Two unglamorous pillars — architecture and trust — that decide whether innovation survives contact with reality.

Imagine two companies handed the identical opportunity on the same morning: a powerful new AI capability, a generous budget, and a quarter to prove it. Six months later one has shipped something its customers love, and the other has a graveyard of half-finished pilots and a demoralised team. The technology was the same. What differed lived in two pillars that rarely make the keynote: Structure and Cohesion. Miklós Róth's S-I-C-T framework exists partly to give those two their due.

Their stabilising role is set out directly in this account of structure and cohesion, with a complementary view in a second structure-and-cohesion piece. Structure is the unspectacular machinery — the architecture, the rules, the processes, the clarity about who decides what. Cohesion is the human counterpart — trust, shared understanding, the quiet agreement that lets a group read the same situation the same way.

Return to our two companies. The first had structure: when the AI tool produced something useful, there was a defined path from output to decision to action. It had cohesion: people trusted the initiative enough to feed it real work instead of quietly routing around it. The second had neither, and so its strong data and ambitious plans dissolved into conflicting interpretations and stalled approvals. The broader logic connecting these pillars to the other two appears in this S-I-C-T framework overview and a second framework source.

In AI adoption specifically, the balance becomes decisive. As S-I-C-T and AI systems argues, new technology has to be carried by governance and organisational trust or it simply does not hold. And as information and cohesion in SICT shows, information turns valuable only at the moment people can align around it — before that, more data just means more to disagree about.

None of this means structure should calcify into bureaucracy. The diagnostic readings in S-I-C-T as a diagnostic model and a second diagnostic account are useful precisely because they help locate where a system is too loose or too rigid. Set against the wider lens of complexity and the study of complex systems, the pillars stop looking dull and start looking load-bearing.

Zoom in on the moment the two companies diverged. Both hit the same snag in week three: the AI tool produced an output that was useful but not obviously safe to ship. In the first company, someone knew exactly whose call it was, and that person had the context to approve it within a day. In the second, the output sat in a channel nobody owned; three people each assumed another would handle it; and by the time it surfaced, the moment had passed. No grand strategic failure — just a missing line of structure and a thin layer of trust. That is usually how it goes. Innovation rarely dies of a dramatic cause. It dies of an unanswered message.

Structure and Cohesion will never be the exciting part of a strategy deck. But they are usually the difference between innovation that survives and innovation that merely happened. S-I-C-T's contribution is to name them, score them, and refuse to let them be taken for granted.