Job Title-to-Medical Surveillance

Occupational health · UAE energy group · 204 roles

Turned an occupational-health team's spreadsheets into a one-search dashboard: any role, its required examinations, its hazards and their risk ratings.

Re-enacted role view: summary, surveillance factors and the activities-and-exposures table

Context

The occupational health risk assessment existed and was thorough — a surveillance sheet per role, a factor catalogue, and one exposure workbook per business unit. It was also unusable at the moment it mattered: a doctor, an HSE officer or HR asking what a specific role needs had to know which of a dozen sheets to open.

Surveillance factors, examinations and exposures were split across separate workbooks with no shared key
Whether a role counted as office or field staff was decided by whoever read the sheet
A new job title had no profile until someone assessed it from scratch
The same question — what does this role need? — was answered differently depending on who answered it

Outcome

204job titles, one search

A role's surveillance requirements went from several sheets and a specialist to one search for anyone
Office/field classification became a rule instead of a judgement
PLACEHOLDER — add a measured figure here (e.g. lookups per week, or time per profile before vs after)

Re-enacted: search a role, read its surveillance profile, open a factor for its tests — the dashboard's own flow, in the site's palette

What we built

First the join. Every workbook is read in the browser as-is — no migration, no new database — and matched on job title and business unit, so the surveillance sheet, the factor definitions and the exposure rows become one record per role. Office-or-field is decided by rule from the title and the exposures, not by eye.

Then the lookup: a searchable list of every onsite title across the group's business units. Choosing one opens the role's summary, its HR mapping, its active surveillance factors as chips, the health-examination table (test, frequency, pre-employment versus routine, field-specific) and the activity-and-exposure table with a risk rating on every row and High/Medium/Low counts at the top.

Then a second organisation. The same reader takes a different pair of workbooks, so a government dataset was added with a toggle rather than a second build.

Then intake. For a title that has no profile, an admin page finds the closest existing roles by fuzzy match and asks a language model to draft the full surveillance profile — constrained to factors that exist in the catalogue, with the best local match as the fallback if the model fails. The draft is editable before it is saved.

What stuck

The workbooks stayed the source of truth — the team keeps editing Excel, and the dashboard reads it
One answer per role, whoever asks
New roles start from a draft grounded in the existing rules rather than from a blank sheet

Snapshot

Client profile: Occupational health function of a UAE energy group; a second government dataset added
Engagement: Catalyst — scoped build
Timeline: About two weeks, April 2026
Scale: 204 job titles, 11 business units, ~2,800 activity-and-exposure rows, 19 surveillance factors
Scope: Workbook ingestion, role lookup, exposure and examination tables, AI-assisted intake for new roles

← All work