#Work Intelligence

Work Intelligence is TechWolf's task-level data layer. It breaks every job into the tasks that make it up, inferred from the job's data sources. Each task carries a score for AI impact and an estimate of the weekly hours spent on it, so a job's tasks together form a task-level map of the organisation's work.

#Data sources

Work Intelligence uses all job data synced into the tenant. Task profiles are built per job, from the data elements linked to that job.

Jobs

• Required: job titles and job descriptions. These are the primary anchors for task inference, so a job needs at least this to produce a profile.
• Optional: vacancies linked to the job. These broaden coverage and improve fidelity where they exist.

#Key attributes

Work Intelligence is organised around a set of connected pieces of data:

• Tasks, the individual pieces of work that make up a job, and the smallest unit in Work Intelligence.
• AI impact, a score on the Human Agency Scale (H1 to H5) for how much human involvement each task still needs once AI is applied.
• AI rationale, a short, plain-language explanation of why a task received its AI impact score.
• Job to task relationship, which tasks belong to which job. The same task can appear on several jobs when their work is similar.
• Weekly hours, an estimate of the time spent on each task within the job. A job breaks into 6 to 20 tasks, and the profile targets a full 40-hour week across them. Weekly hours make the map quantitative: they show where time goes and let higher-hour work carry more weight in any analysis.
• Custom tasks, tasks the organisation defines itself and adds to a job, alongside the inferred ones.
• Task hierarchy, the four-level structure that tasks sit within, from broad to specific: domain, subdomain, cluster, and canonical task. It lets work be analysed at any altitude, from a single task up to a whole domain.

#How it works

Work Intelligence turns a job's data into a task profile, then scores each task for AI impact and estimates the weekly hours spent on it.

The profile itself is built in three phases, applied to every job:

• Filter. A job often has many job descriptions and vacancies that repeat each other. Duplicates and near-duplicates are reduced to a representative set, so no single text outweighs the rest.
• Infer. Tasks are extracted from each representative text, together with the job title, the company name and, where available, the job family, which help read titles and terms the way the organisation uses them. Every task is written in the same structured format.
• Aggregate. Tasks that describe the same work are merged into one profile per job. Work that recurs across sources ranks higher, with job descriptions weighted highest, and the highest-ranked tasks make up the profile.

Every task in the profile is then scored on the Human Agency Scale (H1 to H5), grounded in Stanford research that captures how much human involvement a task needs once AI is applied, each with a short rationale.

Inference is not a one-off. Tasks are re-inferred as the underlying job data changes: once a tenant is live, a change to a job's description or to one of its linked vacancies triggers a re-inference of that job, so profiles stay current without a manual run. Inference is stable, so reprocessing the same data returns the same profile, and only new or changed data affect the output.

#Availability

Because Work Intelligence is a data layer, its value comes from being distributed. The same task data is available through several surfaces:

• Console: review and govern each job's tasks: validate or reject suggested tasks, adjust weekly hours, and add tasks, from the existing task inventory or as custom tasks.
• Exports: task and job-to-task data is extracted as CSV or Excel files, on demand or on a recurring schedule.
• API: the SkillEngine API exposes task profiles and search for programmatic access.
• Agent: the Analyst Agent answers questions about a job's tasks, their weekly hours, and their AI impact in natural language.
• Partner integrations: task data is delivered into Visier, where it powers dashboards on task hours and AI impact.