Prompts that work, grouped by the job — styling a map, writing popups, building a dashboard, drafting a report.
Copy a prompt, paste it into the Hyz AI panel for that job, and swap in your own field names. Three habits make them work harder:
status" beats "colour by status".status A = active, capacity is in MW. Turn on Share 3 sample rows if the model still guesses wrong — see Hyz AI.Make this an operations map. Colour the substations by status — normal green, shortage red, maintenance orange — size them by capacity_mw, label them by name from zoom 10, and do not cluster them.
A categorical colour, a size ramp, and labels that appear only close enough to read.
Animate the transmission lines one-way in the flow direction with arrowheads, colour them by status, weight 3.
Moving dashes in each line's digitising direction; say "two-way" for bidirectional flow.
Use icons by type: transmission as bolt, distribution as building, solar_plant as star, wind_farm as tower. Keep the points around 12 px.
An icon per value of a text field, up to about a dozen values.
Colour the regions by shortage_level — pale green for none, amber for moderate, red for severe. Light fill, dashed outline, label by name.
A categorical fill. Ask for "graduated in 5 classes" when the values are continuous.
Show only the problem: hide the normal substations, make the shortage ones large red bolt icons, and fit the view to them.
Visibility, symbology and extent at once — a good follow-up on a styled map.
Substations popup: title name, then Type, Status, Peak load (MW), Load %, Daily usage (MWh), Reading date — with those labels, in that order.
A field-list card; the labels you write become the field aliases.
Lines popup: a text card reading "{name}: {load_mw} of {capacity_mw} MW ({load_pct}%) — {status}".
A narrative template; every {field} token is replaced per feature.
Give every layer a popup with the four or five most useful fields and friendly labels, and hide the internal id columns.
Let the assistant choose; it can only pick fields that exist.
Turn popups off on the regions layer, keep them on the substations.
Per layer, in one request.
Build an operations dashboard for the week. KPIs: total daily usage on the latest day, substations in shortage, total shortage MW, average load %. A line chart of daily_usage_mwh by reading_date, a bar chart of peak_load_mw by region, a pie of status. The map with a legend. A table of substations with name, region, reading_date, load_pct, status.
One page in the order such a dashboard reads: selectors, KPIs, charts, map, table.
Colour the whole row red when status is shortage and orange when maintenance, and colour load_pct cells orange above 90.
Colour rules: row scope for status-like text, cell scope for numeric thresholds.
Add a date selector and a region selector at the top of the page.
Selectors filter every widget bound to the same source.
Add a second page called Regions with a table of regions coloured by shortage_level and a gauge of reliability_pct. Keep everything on page 1.
A revise prompt. Say what to keep, or "add a page" reads as "rebuild".
A4 landscape executive report: title, the map on the left, legend and summary on the right, a written analysis of where the shortage is and which lines are down, and a document info block. One page.
The analysis is written from the layer names, fields, counts and ranges — a section, not a caption.
One page per layer, each with a short analysis and an attribute table of that layer's most informative columns.
A multi-page report; page one stays the cover.
Portrait A3 with a large map, a scalebar, the CRS block and a north arrow, and no tables.
Page size, orientation and the blocks you want, in one line.
Write the analysis in Arabic and keep the block labels in English.
The text follows the language you ask for; the map does not change.
Describe this dataset for an electricity operations team: what each field means, the units (MW, MWh, kV, %), and which fields tell us about shortage.
A title, a description, an alias for every field, and ideas.
This is a municipal street-lighting inventory from a 2025 survey. Write the description for a non-technical reader.
The hint box is the fastest fix when it guesses the domain wrong.
The column names are codes. Suggest a readable alias for each one and say which fields are worth charting.
The strongest use of this job: an upload whose columns are f_01, f_02.
Suggest three dashboards and two maps this data would support, one sentence each.
The ideas alone, when your description is already fine.
reading_date and Reading Date are different columns — copy them from the field list.