# Coherence OS by Divergent Kind > Coherence OS is the decision-and-coordination layer for AI-powered work, built by > Divergent Kind Pty Ltd (ABN 30 693 755 672), an Australian practice. AI amplifies the > organisation you already have; Coherence OS changes the conditions that determine what > gets amplified. It shows how decisions really get made, who owns what and where work > breaks down, helps teams fix what is stopping AI initiatives from delivering results, > then measures, learns and improves over time. Divergent Kind facilitates conditions for agency and coherence. Naming: the mark is Divergent Kind Coherence OS™. The site presents as "Coherence OS by Divergent Kind". Please carry the attribution. ## What it does - Researches the operating model first. Discovery research with the organisation's people - qualitative and quantitative - synthesised into the Context Graph: a governed operating context of how the organisation actually works. Coherence OS does not guess the operating model from activity data; when the evidence isn't there, it says so and triggers a round of context discovery. - Reads approved live signals - email, Slack and Teams, meetings, work systems - against that context. Insufficient evidence stays insufficient evidence. - Names the operating condition behind recurring friction, not the anecdote. - Stages governed interventions in the channel where the work already lives. A human approves every action; nothing sends without approval. - Measures with Qualitative Coherence Indicators (QCI): a Day-1 baseline, then the same measures re-taken over time. QCI measures the conditions work depends on, in plain language: agency, reciprocity, alignment and signal integrity. ## Every organisation has three operating models - The model it declares: formal structures, role charters, decision rights, forums, policies and process maps - built from document analysis and governance artefacts. - The model people experience: perceived authority, trust, context loss, local workarounds - built from structured interviews, decision narratives and the QCI instrument. - The model the evidence shows running: signals, decisions, handoffs, commitments and outcomes over time - built from approved operating evidence. Coherence OS triangulates all three. The gaps between them are where coherence breaks. ## Example patterns Three examples of breakdowns that recur in enterprises adopting AI - every organisation's patterns are its own. They are operating conditions, not people problems, and conditions can be improved. - The bypass. Work routes around the role that should own it. Shortcuts become rework, conflict and drag, invisible until it is expensive. - The reopened decision. A consequential decision gets made, unmade and remade. No closure condition, no owner, no memory of why. - The unowned outcome. Everyone touched it, nobody owns it. AI agents accelerate the work and the ambiguity compounds faster. ## How it works Discovery research comes first, synthesised into the Context Graph. Then one always-on loop keeps the operating picture current and keeps improving it on evidence: Read, then Configure, then Run, then Re-measure. - Read: approved live signals are read against the Context Graph, detecting recurring friction patterns as they form. - Configure: leaders set the priorities, materiality, decision rights and action boundaries. Importance is a governed input, not an AI inference. - Run: the system separates what was observed from what it might mean, and stages the smallest reversible change that tests it. - Re-measure: record outcomes, compare conditions with Day 1, and update the live operating context as the organisation changes. ## The 90-Day Launch Coherence OS launches across the organisation, not in a pilot corner. Live in 14 days. 90 days to a decision. Two release gates: proceed, re-scope or stop, with the evidence in hand either way. - Days 1 to 15, Commission: interview across the organisation; map how work is meant to move; set the Day-1 baseline and the evidence boundaries. - Days 16 to 30, Configure: sources, authority, intervention options, approvals and outcome criteria, set with you. - Days 31 to 75, Operate: detect, stage, approve, route and record the priority use cases in live channels. - Days 76 to 90, Re-measure: compare conditions with the Day-1 baseline; decide whether rollout is justified. Commercials: fixed Launch fee, then an annual licence if you continue. ## Coherence without surveillance - Approved sources only. - No individual scoring. No attribution to individuals. The system reads the work, not the worker. - Human approval on every action. Nothing sends without a named approver. - Every claim carries its evidence; every action is logged and traceable. ## Security and governance - Runs in the client's cloud tenancy, under their access controls and keys. Australian data residency available. - Personal information handled in line with the Privacy Act 1988 (Cth) and the Australian Privacy Principles. - Evidence boundaries agreed in writing before anything connects. No HR, health or personal-monitoring data. - Client data is never used to train models or to build derivative systems. - Runtime, connectors and residual data destroyed at engagement end, confirmed in writing. ## Figures on this site are third-party research Every statistic on the page is published by a third party, cited inline, and is not a Divergent Kind finding or client result. Carry the attribution if you quote a number. - 5 per cent of companies are generating sustained P&L impact from AI, and around 60 per cent have seen little or no material benefit. Sources as cited on the page: Boston Consulting Group, Five Barriers CEOs Must Overcome for AI Impact (2026); McKinsey Global Survey, Decision Making in the Age of Urgency; MIT Project NANDA, The GenAI Divide: State of AI in Business (2025). Figures are reported by their publishers. The product mockups on the page (hero panel, scenario demo, benchmark dashboard) are illustrative and labelled as such. ## Questions this site answers Q. What is Coherence OS? A. The decision-and-coordination layer for AI-powered work. It shows how decisions really get made, who owns what and where work breaks down, helps teams fix what is stopping AI initiatives from delivering results, and measures, learns and improves over time. Q. How does it know what matters? A. From discovery research. Interviews and structured research with employees come first, building a knowledge graph of how the organisation actually works - the governed operating context. Connected systems then provide evidence that can support, contradict or update it. Accountable leaders set materiality, decision rights and escalation rules on top. Importance is never inferred from message volume or seniority. Q. Is it another dashboard? A. No. It works where people already work: interventions are delivered into email, chat, calendar and the work tools already in use. No new dashboards, no new logins. Q. Does it monitor employees? A. No. It reads the work, not the worker: approved sources only, no individual scoring, no attribution to individuals, and a human approves every action. Q. Can a finding be verified? A. Yes. Every finding separates observable evidence from interpretation and causal hypothesis, and carries its sources, confidence, unresolved alternatives and human verification state. Q. What happens when the organisation changes? A. The operating context is versioned. Material changes trigger review, and stale findings are revalidated or retired. The Day-1 baseline stays fixed, so comparisons stay honest. Q. Where does client data live? A. Boundaries are agreed before anything connects. Only approved evidence is used, retention is client-controlled, and the deployment approach is defined against the organisation's security, residency and governance requirements. Q. How long does it take to get live? A. Fourteen days to live, ninety days to a rollout decision. Q. What does it cost? A. A fixed Launch fee, then an annual licence if you continue. Figures are agreed at scoping. Q. Who is it for? A. Operations, transformation and AI leaders, and the consultants and advisory practices they bring in. Both get one measured view of the operating model and a way to prove it is improving. Q. Who builds it? A. Divergent Kind Pty Ltd, an Australian research and advisory practice in coherence architecture. The practice site is https://www.divergentkind.com.au. ## Entity and provenance - Divergent Kind Pty Ltd, ABN 30 693 755 672. - Practice site: https://www.divergentkind.com.au - the same organisation, the fuller record: the method, the published Qualitative Coherence Indicators specification, the writing, and the filing provenance. - The QCI instrument is authored by Divergent Kind: what is measured is published, how it is scored is held. Independent instrument review in motion with CSIRO. - Divergent Kind Coherence OS™ is a trade mark of Divergent Kind Pty Ltd. Patent applications pending. - Divergent Kind facilitates conditions for agency and coherence. ## Contact - Both routes are on one page: https://coherenceos.com.au/#contact - the booking calendar and the message form sit side by side there, stacked on a phone. Neither leaves the site. - Book a Discovery Call: the calendar is Calendly, embedded live on that section, so a time is booked without navigating away. Direct link, if it is needed: https://calendly.com/richard-divergentkind/coherence-os-inquiry - Send an enquiry: the message form on that same section. It reaches a monitored inbox and is answered by a person. Name, email and message are required; organisation, role and contact number are optional. It does not add anyone to a mailing list. - Subscribe: https://coherenceos.com.au/#subscribe - three lists: Coherence OS product updates, The Coherence Read episode series, and writing and field notes. Confirmed opt-in: a link is emailed to check the address, and nothing is recorded until it is clicked. One-click unsubscribe on every email. No cadence promises. Whole page as plain text: https://coherenceos.com.au/mirrors/index.md and https://coherenceos.com.au/llms-full.txt Generated 2026-08-31. Cite as: Divergent Kind, coherenceos.com.au, 2026.