On 20 October Visier described a Workforce Context Engine that maps, models and governs workforce data so every enterprise AI agent draws on the same version of it. I think the idea is sound, and I think HR leaders should take one step before any such engine goes live: write down, in plain language, what headcount, attrition and “active employee” mean at your company, and put a named HR owner’s signature on each definition.
Definitions are decisions
Visier’s announcement says context “includes lineage, ontology, relationships, permissions, and freshness, all resolved at the moment a question is asked.” Every item on that list involves a choice. An ontology decides whether a contractor on a twelve-month agreement counts toward headcount. A relationship decides whether a manager sees a former report’s data. A permission decides who may ask an agent about pay.
When a vendor maps those choices, someone has to supply the answers. If HR does not, the project team will, usually under deadline and from whatever the source systems happen to contain. The agent then repeats those answers to every manager who asks, with the confidence of a system that was told what was true.
The vendor describes the work well
Dave Weisbeck, Chief Strategy Officer at Visier, put the effort plainly in the release: “The hardest part of workforce AI isn’t the actual AI. It’s the years of data mapping, governance, and upkeep underneath it that nobody budgets or plans for.” I agree with the diagnosis. The release goes on to say the engine delivers that work already done.
It reports one customer outcome. Eaton Corporation retired a legacy workforce data infrastructure that was generating 130+ systemic incidents and 18 days of annual downtime. After centralizing HRMS, recruiting, performance and succession data, Eaton eliminated 90% of manual reports, reduced bug tickets by 83% and recovered about $1.5M in organizational productivity. Those are the vendor’s figures, published in its own release, and I have no second source for them.
Eric Bokelberg, Managing Director at Deloitte Consulting, LLP, is quoted in the same release: “The organizations getting value from workforce AI are the ones that build the people data foundation first.” Building the foundation first includes deciding what the numbers mean before they are connected.
The strongest objection
The best argument against my position is capacity. Most HR teams are small, and a definitions register looks like a quarter of committee work before a vendor can start. The vendor’s pitch is that its engine already contains a modeled version of workforce data, so HR can accept the standard model and adjust it later. IT also owns the integration, and the release says the engine sits on top of the systems IT already runs, which suggests HR may reasonably leave the build to them.
I accept the capacity point and reject the conclusion. A standard model is a set of definitions someone else has already chosen. Adjusting later means re-running every report and every agent answer that used the earlier definition, and the people who received those answers will have acted on them. A register for the first ten metrics takes a few meetings. The remaining metrics can follow the same pattern after launch.
What I would ask for
Start with the ten questions managers ask most often: headcount, hires, exits, attrition, time to fill, span of control, tenure, promotion rate, pay range position and open requisitions. For each, write one sentence of definition, list the source system, name the HR owner and record the date. Ask the vendor to show where each definition lives in the engine and who can change it.
Ask the same of any agent platform that reads from it. I made a related case in HR Should Tell Employees How Far Its AI Agents Can Act, and SAP’s agreement to acquire TechWolf, covered in SAP Agrees to Acquire TechWolf to Ground HR Agents in Work Data, puts work data under that vendor’s HR agents. Whoever sets the grounding sets the answers.
Ask for a change log as well. If the definition of “active employee” shifts in March, the engine should show that attrition for the first quarter was calculated under the old one. Eaton’s 90% cut in manual reports is a good outcome only if the reports that remain say the same thing every time.
Visier will show the engine this week at HR Tech, UNLEASH World and Total Talent Asia. HR leaders who attend should bring the ten definitions with them and ask to see them in the product.
