Workforce intelligence · for the leaders who own how work is designed
Most companies put AI into old operating models and wonder why it isn't working.
They're adopting AI faster than they're redesigning the roles, decisions, and learning systems around it.
Verellax reads the hiring record. As at 14 September 2026 that is 24,944 live job specifications from 217 employers across ten markets, and we score every AI-central role, where AI is core or supporting the work as written, on four tests. Who works with the model. Who can overrule it. Who owns its output. Which decisions stay with the person.
The score tells a company where it sits. From there we redesign the roles, decisions, and workflows so people's time moves to the work only a human can do, and we build the governance layer that keeps it there.
217 employers on record · re-reads every MondayLive basis · W38 · as at 14 September 2026 · figures re-render each Monday from that week's pack
In AI-native, 29.3 points of 49.4% adoption carry no decision architecture: the widest gap of the six.
The task moved.
The role didn't.
Executives got AI written into their roles first. Graduates got it last. 24.1% of live director roles are AI-central; at entry level in Australia, 8 of 608 are.
Read the thesisWe measure the distance between the AI a company describes and the knowledge work it has designed.
Three ways to work with us, depending on what you already know. Most companies start with the measure. The ones that already know their gap start at Redesign, and the ones that have a design and need it to run start at Build.
We score every job specification a company publishes against four tests: how the person works with the model, who can overrule it, who owns its output, and which decisions the person holds. The Verellax Index publishes the market's answer every quarter, free, with every figure carrying its population and its date. The Diagnostic answers it for your company in three working days, A$18,000. The Benchmark answers it for your whole cohort.
Get your Diagnostic →Discovery builds a living model of how your company works: every recurring decision in scope, who holds it, what the model informs, and where your people's time goes. Strategy turns that into the order of workflows to address. Workflow redesign then delivers, for each one, the flow, the roles, the decision rights, and a named owner for every AI output.
Talk to us →Engineering wires the redesigned workflow and its decision register into your stack. Implementation runs it in production, and Enablement teaches your people to hold the decisions the new workflow gives them. Governance checks it monthly on two measures: judgement share, the proportion of a person's time in the work only a human can do, and accountability coverage, the share of AI outputs with a named owner.
Talk to us →Pick the question you couldn't get an answer to. The record has one.
Every line below is live basis, as at 14 September 2026, and re-renders each Monday
01 Chief Marketing Officer Our marketing teams use AI every day. Who answers for it? 57.3% of 5,674 AI-central live roles name nobody to overrule the model, own its output or hold a decision. Marketing, Australia→ 02 Chief People Officer Are the roles we're writing designed for the work they'll do? 75.6 per cent of 12,547 live roles that mention AI say nothing about who decides. Where you sit→ 03 Chief Financial Officer We cut roles and refilled them. Did we redesign them? Every live and closed specification we hold for an employer is scored on the same four tests. The Diagnostic reads yours against your cohort. Added while cutting→ 04 CEO or Chief Strategy Officer Are we building AI, governing it, or only using it? The user-economy line publishes with Edition one of the Index on 1 October 2026. The user economy→ 05 Transformation Lead Adoption is done. Where did the redesign stop? In every cohort a minority of roles score four of four. The rest is the gap. The designed minority→ 06 Board Member When the model is wrong, who is accountable in writing? 1,354 live roles name a governance framework. 36 name who holds decision authority. The four tests→ 07 Head of Talent What are we hiring graduates to learn? 8 of 608 live Australian entry-level roles are AI-central. At director level, 24.1% are. Entry level→Four instruments read the same record.
The Index is the quarterly measure. Skills reads what employers ask for, and what they've stopped asking for. Method shows how a specification is read. The Register is the named reading of one company, behind the gate. Every figure on all four resolves to the same pack, and every page states its population and its freeze date.
“Lead Automation Analyst, Claims”insurance · AU · posted August 2026 · employer not named
AI is in the title at the manager grade, and the role scores two of four. The specification describes how the person works with the model, and names who can overrule it. It never says who owns its output, or which decision is the person's to hold. It counts among the 42.7% of AI-central live roles that name someone to overrule the model but write no decision as the person's to hold, as at 14 September 2026.
Hired to use the tool.
Read the methodSkills are composition within this specification, not a trend; mapping coverage is not yet flat across the corpus. 10 skills named in this specification, as at 14 September 2026.
Read the skills map