Voice Assistant

The Voice Assistant enables hands-free survey completion through natural conversation with an AI. Speak your data, and the AI extracts and validates values automatically — no typing required.

When to Use Voice Mode

Voice mode is ideal when:

• Your hands are occupied (tools, equipment, climbing)
• You're wearing gloves
• Typing is impractical (tight spaces, outdoor conditions)
• You need to move quickly through a survey

Starting a Voice Survey

1. Open a survey
2. Tap the microphone button to enter voice mode
3. The AI will greet you and ask for the first field
4. Speak your response naturally
5. Continue until all fields are complete

Voice Recording Consent

Before your first voice survey, a Voice Recording Consent dialog explains how your speech is handled. It says that:

• Your speech goes to a speech-to-text service for live transcription, and that service doesn't keep your audio.
• The transcribed text, with your answers so far, goes to a third-party AI service to pick out the survey answers.
• Transcription and AI answers can be wrong, and you are responsible for checking them.
• Voice is optional. You can use manual input instead.

The dialog also links to the Privacy Policy. If your organisation has turned on the audio log, it says recordings are kept for up to 14 days. If the speaker lock is on, it describes the short clip of your voice that the lock keeps (see Speaker Lock).

Tap Accept & Continue to start. Tap Use Manual Input to decline: you leave voice mode and nothing is sent to the AI service. On the phone, the dialog appears once, and tapping Use Manual Input takes you out of the voice screen.

On the web, the dialog appears again when its wording changes, for example when your organisation turns on the audio log or the speaker lock.

The Voice Interface

Recording Button

• Tap to start recording — button turns red
• Speak your response — your words appear as a transcript
• Stop speaking — the system automatically detects when you finish
• An audio beep confirms your speech was captured

Chat Display

• Your speech appears in the transcript area
• AI responses are displayed and spoken aloud
• The conversation history scrolls as you progress

Progress Sidebar

• Shows completion percentage
• Lists all fields with their status
• Tap a field to jump directly to it

---

How End-of-Input Detection Works

The voice system uses a 2-second silence debounce to determine when you've finished speaking. Here's what happens:

1. You start speaking — your words are transcribed in real time
2. You pause briefly (under 2 seconds) — the system keeps listening, assuming you're mid-thought
3. You stop speaking (2+ seconds of silence) — the system finalises your transcript
4. An audio beep confirms your input was captured
5. The AI processes your response and replies

Tips for Pacing Your Speech

• Speak in complete thoughts — finish your answer before pausing. "The temperature is 75 degrees" works better than "The temperature is... 75 degrees" with a long pause in the middle
• Don't rush — the system waits 2 seconds after you stop, so there's no need to speak quickly
• Multi-field answers are fine — "The width is 1200 and the height is 800" is captured as one input
• If you pause too long mid-sentence — the system may process a partial answer. Just correct it by saying the full value again

---

Speaking Naturally

You don't need to use rigid commands. The AI understands natural language:

What You Say	What Gets Captured
"The serial number is ABC dash 123"	Serial Number: ABC-123
"Temperature is 75 degrees Fahrenheit"	Temperature: 75 degrees F
"It's in good condition"	Condition: Good
"Skip this one"	Field skipped
"The width is about twelve hundred millimetres"	Width: 1200 mm
"Next Tuesday" (for a date field)	Date: resolved calendar date
"Three out of five" (for a rating)	Rating: 3

Multi-Field Input

Provide multiple values in one utterance:

"The width is 1200 millimetres and the height is 800 millimetres"

The AI will capture both the width and height fields.

You can also answer across different field types:

"It's a roller blind, Nordic Grey fabric"

The AI captures the product type and fabric fields in one go. When answering multiple fields at once, comment fields are skipped automatically — see Comments & Observations for details.

Voice Shortcuts

For fields with long option names, use common abbreviations:

• "Arc weld" instead of "Electric Fusion (Arc) Weld / Spiral Weld (0.8)"
• "ERW" instead of "Electric Resistance Weld (1.0)"
• "Yes" or "No" for boolean fields

The AI matches your shorthand to the correct formal option. Voice shortcuts are configured by template builders — see Templates for details.

Dictating Codes and Serial Numbers

For fields configured as Code Input, you can use the NATO phonetic alphabet:

What You Say	What Gets Saved
"Alpha Bravo Charlie one two three"	ABC123
"Golf Bravo one two three Alpha Bravo"	GB123AB
"Sierra November dash one two three four"	SN-1234

Number words and common homophones are also converted:

• "zero" through "nine" → 0-9
• "niner" → 9 (aviation form)
• "dash" or "hyphen" → -
• "dot" or "point" → .

---

Reference Images

When relevant, the AI can display reference images to help you:

• Equipment nameplates showing where to find serial numbers
• Condition examples (Good, Fair, Poor)
• Installation diagrams

Ask the AI:

"Show me the nameplate"

The AI will display the relevant reference image inline in the chat.

---

Photos & Signatures

Taking Photos

You can capture photos at any point during the survey:

"Take a photo of the serial number"

The AI opens your camera with a descriptive hint. Some fields have automatic photo prompts — the AI will ask you to take a photo before collecting the field value (e.g., "Before the serial number, take a photo of the nameplate.").

On a computer, the app shows the hint with Choose photo to upload a photo from your files, or Skip to carry on without one.

In hands-free mode (GoPro connected), photos are captured automatically without interaction.

Signatures

When a survey includes a signature field, the AI will prompt you to sign:

"Please provide your signature for the completion sign-off."

A signature pad opens on screen. Draw your signature and tap confirm. The AI handles the rest — you don't need to say anything to save it.

Location Pins

Some templates ask you to mark each observation on a drawing attached to the survey. Pins are placed by hand, in form mode — the assistant never asks for one and has no way to place one.

If you ask it to ("set the location pin next to the stairs"), it says pins are placed by hand on the sheet from the observation's card, then carries on with the current question. Anything else you said in the same breath is still saved.

Switch to form mode to place pins, then switch back. Completing the survey is refused while a required pin is missing, and the assistant names the observation and field when it refuses.

See Drawings & Pins →

---

Comments & Observations

Many survey templates include comment fields alongside checklist items (e.g., "Drive Shaft Condition" has a companion "Drive Shaft Comments" field). The AI detects these automatically and offers several ways to capture observations.

Explicit Comments

Use keywords like "note", "comment", or "add a note" to save an observation to the nearest comment field:

"Note that the housing is cracked"

"Add a comment: slight corrosion on the left side"

The AI finds the appropriate comment field in the current section and saves your observation there. If there's no comment field in the section, the AI will let you know.

Implicit Comments

When answering a checklist field (pass/fail, yes/no, etc.), any descriptive text you add is automatically treated as a comment:

What You Say	What Gets Saved
"Fail, missing bolts on the panel"	Condition: Fail, Comment: "Missing bolts on the panel"
"No, the nameplate is faded"	Field: No, Comment: "The nameplate is faded"
"Pass"	Field: Pass (no comment)

Short answers like "Yes", "No", "Pass", or "Fail" on their own are not treated as comments — just the field value.

Comments in Multi-Field Responses

When you answer multiple fields in one utterance, the AI skips comment prompting and moves straight to the next unanswered field:

"Filter condition good, last filter change was January 2026"

Both fields are saved and their comment companions are automatically cleared — no "Comment?" prompt follows. This keeps you moving through the checklist efficiently.

To include a comment within a multi-field response, use an explicit keyword:

"Filter condition good, note that the housing is cracked, last filter change January 2026"

The AI saves all three: the condition value, the comment, and the filter change date.

---

Navigating the Survey

Moving Between Fields

• The AI automatically moves to the next field after capturing your response
• Say "go back" to return to the previous field
• Say "skip" to move on without providing a value
• Say "next" or "continue" to advance manually

Jumping to Specific Fields

"Go to the equipment type field"

Checking Progress

"What's left to complete?"

The AI will summarise remaining required fields.

Section Focus

You can narrow the AI's attention to a single section:

"Focus on the equipment details"

While in section focus, the AI only asks about fields within that section. When the section is complete, it asks whether to continue to the next section. This is useful for long surveys where you want to work through one area at a time.

Bulk Completion

For checklist-heavy sections, complete an entire group in one command:

"All under vehicle checks: pass, no comments"

The AI sets all remaining fields in the section to the stated value and clears their comment companions. If it can't identify the section, it will ask you to continue one field at a time.

Navigating Repeatable Groups

When working with repeatable sections (e.g., rooms, equipment items), you can navigate between instances:

By label:

"Go back to the kitchen"

By number:

"Go to room 2"

Relative navigation:

"Previous room" or "Next room"

See Repeatable Group Commands below for the full set of commands.

---

Repeatable Group Commands

Repeatable groups allow variable numbers of items (rooms, equipment, tasks). The AI manages the flow automatically, but you can also use direct commands.

Adding Instances

After completing all fields in an instance, the AI asks:

"Room 1 complete. Add another room?"

• Say "yes" to create a new instance
• Say "no" to move to the next section

You can also add instances proactively:

"Add another room"

Zero-Instance Groups

Some groups represent items that may not exist (e.g., defects, issues). When configured with "Allow Zero Instances", the AI asks upfront:

"Are there any defects to record?"

Say "no defects" or "none" to skip the entire group. The group is recorded as "None recorded".

Navigating Between Instances

What You Say	What Happens
"Go back to the kitchen"	Jumps to the instance labelled "Kitchen"
"Go to room 2"	Jumps to the second instance
"Previous room"	Moves to the prior instance
"Next room"	Moves to the next instance
"Go back to window 1"	Jumps to a nested child instance

Nested Group Navigation

For surveys with nested groups (e.g., Room > Window), the AI tracks your position automatically:

Room 1: Kitchen
  ├── Window 1 (completed)
  ├── Window 2 (in progress) ← you are here
  └── Window 3 (not started)
Room 2: Living Room
  └── Window 1 (not started)

Example flow:

1. AI: "Kitchen, Window 2. What's the opening width?"
2. You: "1200 millimetres"
3. AI: "Opening height?"
4. You: "800"
5. AI: "Kitchen, Window 2 complete. Add another window?"
6. You: "Yes"
7. AI: "Kitchen, Window 3. Opening width?"
8. ...
9. You: "No" (declining more windows)
10. AI: "Kitchen complete with 3 windows. Add another room?"

Batches

A batch is a set of like items tested together, such as 12 bolts. The assistant asks one set of tests across every item, then the next set, so you can stay on one instrument. See Batches for how a template sets this up.

• The assistant names the item with each question: "Bolt 7: Length?". Give the value; you don't need to say the item.
• It announces each new pass: "Pass 2, hardness. Bolt 1?"
• When a pass has one question, the assistant asks it by the pass name: "Bolt 2: hardness?". So you know what to measure even if you switch from the form to voice halfway through a pass.
• It moves to the next item by itself and never asks "add another?".
• If the template asks how many items are in the batch, your answer sets the number. Answer it again to change it. Items that already have answers are never removed.

One value for every item:

"Grade AB1 for all bolts"

The assistant fills every item that has no value yet. An item that already has one keeps it, and the assistant tells you which.

Nothing to record on any item:

"No defects on any bolt"

The assistant records no defects on every item that has none yet: "No defects on all 12 bolts." An item with a defect already recorded keeps it.

If the assistant can't tell whether you mean one item or all of them, it asks: "Just this bolt, or every bolt?"

Repeated identifiers: if you give a serial that another item already has, the value is saved and the assistant says "A128 is already Bolt 1."

---

Pausing and Resuming

Say "pause", "stop", or "wait" at any point to put the voice assistant on standby. The microphone button turns amber and the system stops listening.

To resume, say any wake word:

• "Ready" / "Continue" / "Go ahead" / "I'm ready" / "Let's go"

You can also tap the microphone button to exit pause mode.

While paused, the AI plays gentle reminders at intervals (30s, 60s, 90s) so you know the session is still active. A short grace period after entering pause mode prevents the AI's own speech from accidentally triggering a resume.

---

Speaker Lock

The speaker lock stops other people's speech being saved as your answers, such as a homeowner talking nearby or a radio. It works in online voice only. Offline voice has its own check, which uses your voice print.

How It Works

1. When you start, the assistant asks you to say "start the survey". This tells it which voice is yours.
2. After that, speech from other voices is left out of what the assistant hears.
3. The assistant doesn't save answers until it has heard the start phrase.

These phrases all work: "start survey", "start the survey", "begin survey" and "begin the survey". Saying a little more around the phrase, such as "okay, start the survey", is fine.

What You See

Until the assistant knows your voice, a banner shows on the voice screen:

• Say "start the survey" so I know your voice. I'm not saving answers yet. Tap Repeat to hear the prompt again.
• I don't know which voice is yours, so I'm not saving answers. Shown if the phrase wasn't heard in time. Tap Fix to hear the start phrase again, then say it.

If the assistant heard speech it couldn't match to your voice, a Heard, but not saved label shows, with a count when there is more than one.

After a Reconnect

If the connection drops and comes back, the assistant normally recognises your voice again without being asked. If it can't, it says "Sorry, I lost the connection. Please say: start the survey." Say the phrase again to carry on. If you stop and restart the microphone yourself, it asks "Please say: start the survey."

Turning the Lock On or Off

An admin controls the lock under Settings › Data Management, in the Voice speaker lock card. The switch is Only save the surveyor's answers. It applies to web voice surveys and to the phone app, from the next survey started.

On the web with headphones, the lock works in any browser. On the loudspeaker, it works only in Chrome or Edge on a Mac. Turn the lock off if it ever withholds your own answers.

---

Switching to Form Mode

If voice isn't working well in your environment:

1. Tap the Form button in the header
2. Complete fields using the on-screen form
3. Switch back to voice mode anytime

Both modes share the same data — progress is preserved.

---

Headset vs Speaker Mode

The voice experience behaves differently depending on whether you're using a headset or the device speaker. See the Mobile App guide for full details.

Quick summary:

Mode	Mic During AI Speech	Best For
Speaker	Muted (prevents echo)	Phone speaker, shared spaces
Headset	Open (barge-in enabled)	Headphones, earbuds, AirPods

Tap the headphone icon in the header to toggle between modes.

---

Tips for Productive Voice Surveys

Before You Start

• Familiarise yourself with the template — skim the field list in form mode first if it's a new template
• Use headphones — headset mode is faster because you can interrupt the AI (barge-in)
• Check your connection — voice mode requires an internet connection

During the Survey

• Answer directly — "75 degrees" is faster than "the answer is 75 degrees Fahrenheit"
• Use multi-field input — "Width is 1200, height is 800" saves a round trip
• Use shortcuts — speak abbreviations instead of full option names when aliases are configured
• Dictate codes with NATO phonetic — "Alpha Bravo Charlie" is clearer than trying to say individual letters
• Don't wait for the AI to finish speaking — in headset mode, start speaking as soon as you know the answer

Dealing with Mistakes

• Say "go back" to return to the previous field and correct it
• Say the field name directly: "Change the serial number to XY-456"
• Tap a field in the progress sidebar to jump there manually

Completing Efficiently

• Say "what's left?" to get a quick summary of remaining fields
• Skip optional fields you don't need: "skip" or "skip this one"
• For repeatable groups with no items: "no defects" or "none"
• When all fields are done: "complete the survey"

---

Voice Commands Reference

Session Commands

Command	Action
"Start" / "Let's go"	Begin the survey from the first field
"Resume" / "Ready"	Resume a previously started survey
"Start survey" / "Start the survey" / "Begin survey" / "Begin the survey"	Tell the speaker lock which voice is yours (when the lock is on)
"Pause" / "Stop" / "Wait"	Pause listening — mic enters standby mode
"Ready" / "Continue" / "Go ahead" / "I'm ready"	Exit pause and resume listening

Navigation Commands

Command	Action
"Next" / "Continue"	Move to next field
"Go back" / "Previous"	Return to previous field
"Skip" / "Skip this one"	Skip current field
"Go to [field name]"	Jump to specific field
"What's left?" / "What's remaining?"	Get summary of remaining fields
"Focus on [section name]"	Narrow navigation to a single section

Repeatable Group Commands

Command	Action
"Add another [group name]"	Create a new instance
"No" / "That's all"	Decline adding more instances
"No [group name]s" / "None"	Skip zero-instance group entirely
"Go back to [instance label]"	Jump to instance by label
"Go to [group] [number]"	Jump to instance by index
"Previous [group]" / "Next [group]"	Relative instance navigation

Batch Commands

Command	Action
"[value] for all [items]"	Set a field on every item that has no value yet
"They're all [value]"	The same, for the field being asked
"No [list] on any [item]"	Record none on every item, such as "No defects on any bolt"

Bulk Section Commands

Command	Action
"All [section] pass"	Set all remaining fields in a section to Pass
"All [section]: pass, no comments"	Set all fields to Pass and clear comment fields

Photo & Media Commands

Command	Action
"Take a photo" / "Photograph this"	Open camera to capture a photo
"Take a photo of [subject]"	Capture a photo with a descriptive hint
"Show me the [image caption]"	Display a reference image in the chat

Survey Commands

Command	Action
"Complete the survey"	Finish and submit
"Note that [observation]"	Save observation to nearest comment field
"Add a comment: [text]"	Save comment to nearest comment field
"Change [field name] to [value]"	Update a previously captured field
"Correct [field name]"	Go back to a field to fix it

---

Audio Feedback

The system provides audio cues:

• Okay or Beep when your speech is captured
• AI voice reads back responses and asks questions
• Progress summary announced after completing sections

---

Troubleshooting

Voice not being captured:

• Check that the microphone button is active (red)
• Ensure the app has microphone permission
• Move to a quieter location
• Check that you're not in airplane mode

AI not understanding responses:

• Speak more slowly and clearly
• Use the exact option names or known shortcuts
• Try spelling out difficult words and proper names
• For codes, use the NATO phonetic alphabet

Feedback loop (hearing yourself):

• Enable Headset Mode if using headphones
• Lower the device volume
• Position the phone away from your face

AI cutting you off mid-sentence:

• Avoid long pauses (over 2 seconds) in the middle of your answer
• Speak in complete thoughts without extended gaps
• If cut off, just repeat your full answer

Survey stuck:

• Tap the progress sidebar to manually select a field
• Switch to form mode to troubleshoot
• Check your internet connection
• Pull to refresh on mobile

Voice not available:

• Voice mode requires an internet connection — check the connection indicator
• If offline, use form mode until connection returns, unless offline voice is available: your organisation has it switched on, your phone meets the minimum and you have turned it on in Settings
• Ensure microphone permissions are granted in your device settings