AI Form Fill is designed to help teams capture site information naturally while reducing manual form entry across Dashpivot and Sitemate Mobile. By combining voice recordings, photos, form context, and site data, Storm generates accurate draft form responses.
What Information Can Storm Process?
Storm reviews multiple data inputs together to generate draft responses for supported fields. It can process:
voice recordings
uploaded photos
typed notes
weather information already added to the form
location information already present in the form
existing content already inside the active form
previously completed forms in the same folder (referenced by name in your directive — note that photo files from historical forms cannot be referenced)
When multiple inputs are provided together, Storm merges them into a single form entry — for example, a voice note describing a defect and a nameplate photo will combine, with the asset ID and serial number coming from the photo via OCR and the description coming from voice.
How Storm Processes Information: Step-by-Step
When AI Form Fill starts processing, Storm examines the provided information and maps it into relevant form fields and sections where possible through the following stages:
1. Voice Recording Evaluation
Storm first reviews the voice recording. During this stage, Storm may:
transcribe spoken audio
detect activities and events
identify keywords and points of interest
recognise names, companies and equipment references
detect delays, issues, defects or incidents
identify instructions spoken by the user
output speaker labels when multiple speakers are detected in the recording
Examples of user instructions:
“Add another row to the workers table”
“Rewrite this as a paragraph”
“Summarise this as meeting minutes”
Note: Storm attempts to distinguish instructions for the AI and actual form content during processing.
2. Photo Examination
Storm inspects attached photos one by one. During this stage, Storm may:
detect equipment and machinery
identify visible issues or defects
identify brands or registration plates
review documents visible inside photos
associate photos with relevant form sections or table rows
For OCR-critical photos such as dockets, nameplates, and labels, Storm reads text directly from the image. OCR results override audio for names, asset IDs, serial numbers, and docket numbers.
Framing determines how Storm routes a photo — a close-up of a defect reads as defect evidence, while a wide shot of the same area reads as site context. Multiple angles of the same defect attach to a single entry rather than creating separate rows. Sequential photos of an ordered process (such as ITP steps) align to numbered rows when capture order matches step order.
Storm attempts to map photos based on photo contents, surrounding context, the order of uploaded images, and related form entries. (Photo mapping is best-effort and may occasionally require manual review).
3. Existing Form Context
Storm reads information already inside the active form to improve generated responses and field mapping. This includes:
Weather fields
Dates and timestamps
Existing notes
Linked dropdowns and lists
Table content and previously entered values
Previously completed forms in the same folder — when referenced by name in your directive (e.g. "use yesterday's site diary")
What Storm Can Read | What Storm Cannot Currently Read |
✓ Active form content | ✗ Other Dashpivot forms |
✓ Active table & dropdown data | ✗ External forms |
✓ Weather & timestamp fields | ✗ PDF, Word, or Excel attachments |
4. Linked Lists and Dropdown Matching
Storm uses linked lists and dropdown values as contextual vocabulary to improve accuracy when filling dropdown fields, linked lists, tables, and structured workflows.
For example, Storm may:
match spoken subcontractor names
recognise equipment references
identify workers from linked people lists
map spoken terms to known dropdown options
5. Weather and Location Context
If weather or location information already exists in the form context, Storm may use it during processing to establish:
weather conditions for inspections
timestamps for site events
environmental conditions
project location context
Note: Storm does not automatically capture location data unless it is already available in the form or provided by the user.
6. Language Detection and Translation
Storm can process supported multilingual input. During processing, it may detect the spoken language, detect the template language, and translate content into the destination template language.
Example: A worker may speak in Spanish, but the destination form is in English. Storm will attempt to generate English responses while preserving the original transcript.
To keep content in its original language, users can state this explicitly during recording — for example, 'write this in Spanish' or 'don't translate.' Proper nouns, asset IDs, standards codes, and quoted content are always preserved verbatim regardless of language setting.
Multilingual behavior may vary depending on recording quality, workflow complexity, terminology, and regional language usage.
7. Draft Generation
Once processing is complete, Storm attempts to generate draft responses for supported fields. Storm will only fill fields when confidence is high enough; if Storm is unsure, fields may remain empty for manual review. Storm may:
populate text fields
complete tables
trigger logic-based fields
map photos into rows or sections
update dropdown selections
generate summaries or meeting notes
Human Review Is Still Required
While AI Form Fill is designed to assist with drafting and completing forms, users remain responsible for reviewing submitted content.
What Users Should Always Do:
Review all generated outputs
Verify important information
Check mapped photos
Confirm dropdown selections
Complete missing required fields
What Storm Does NOT Do:
Approve workflows
Sign forms
Complete signature fields
Automatically submit forms
Current Limitations
Storm currently does not:
work inside Dashpivot Actions
analyse PDF, Word or Excel files
Additional capabilities may be introduced in future Storm releases.