Skip to content

Troubleshooting AI and dictation

Ask Dockli (the AI assistant) and voice dictation are part of the Spark plan. If the AI Mode button is missing, or the AI or microphone isn't behaving, this page covers each case.

Last updated: July 2026

Ask Dockli and voice dictation belong to the Spark plan and need a live seat. If a feature is missing or misbehaving, match your symptom below.

There’s no AI Mode button at all

Symptom: You don’t see the AI Mode toggle in the search bar, and there’s no way to chat with Dockli.

Cause: This is expected on the Essential plan. Essential gives you the full File Explorer file experience but not Ask Dockli — the AI is a Spark-only feature. When your organization is on Essential, the AI Mode toggle is hidden entirely. This is not a bug or a broken install.

Fix

  1. If you expect AI, confirm your organization is on the Spark plan (not Essential) — ask your IT administrator.
  2. If your organization has Spark but you still don’t see AI, confirm you have a Dockli seat — AI needs a live seat/entitlement. See Access groups and seats.
  3. Restart Dockli (close all File Explorer windows, then reopen) so it re-checks your license.

Essential vs Spark

Essential = the full file experience in File Explorer, no AI. Spark = everything in Essential plus Ask Dockli (chat, voice, document diffing, and plain-language search). The AI Mode toggle only appears on Spark.

AI returns an error or refuses a request

Symptom: The assistant shows an error message (often starting with “AI service returned…” or similar), or declines a prompt.

Cause: A few different things surface here:

  • Content filtering. Prompts (or results) that trip Microsoft’s content safety filters are blocked, and the assistant surfaces an error rather than answering.
  • No live entitlement. The AI gateway is fail-closed and enforced server-side: if your seat isn’t active, the AI won’t run even if the interface still shows the toggle.
  • A transient service error. An upstream hiccup returns an error frame like “AI request failed” with a status code.

Fix

  1. If the message looks like a content-safety block, rephrase your prompt and try again.
  2. Confirm your Dockli seat is active (see Access groups and seats) — a removed or expired seat stops AI server-side.
  3. For a transient error, wait a moment and resend the prompt.
  4. If errors persist across different prompts, check your network and see Performance and network and Endpoints to allowlist.

The AI isn’t responding, or seems stuck

Symptom: The assistant is “thinking” but nothing comes back, or a request appears frozen.

Cause: A long-running or stuck request. Dockli always gives you a way to cancel: while the AI is working, the microphone button becomes a Stop button.

Fix

  1. Click the Stop button (the rounded-square icon that replaces the mic while the AI is working) to cancel the in-flight request.
  2. Try your prompt again — often a simpler or shorter prompt resolves faster.
  3. If every request stalls, check your connection and see Performance and network.

The microphone / dictation isn’t working

Symptom: Clicking the mic does nothing, dictation never starts, or you see a “Speech error” on the mic button.

Cause: Dictation uses Azure Speech and needs a working microphone plus microphone permission. Dockli runs inside WebView2, which grants microphone access automatically, so the usual culprit is Windows itself — no mic, the wrong default device, or Windows microphone privacy settings blocking apps.

The mic button shows its state as you go: “Connecting…” while it’s setting up, then “Listening — speak now” once it’s genuinely capturing. If it never reaches “Listening”, something is blocking capture.

Fix

  1. Confirm your microphone works elsewhere in Windows (test it in Settings → System → Sound → Input, or in Voice Recorder).
  2. Check Settings → Privacy & security → Microphone and make sure microphone access is on and apps are allowed to use it.
  3. Make sure the correct microphone is selected as your default input device in Windows.
  4. Hover the mic button — if it shows a “Speech error” tooltip, that’s the underlying reason; a “could not start the microphone” error points to a device or permission problem.
  5. Retry: click the mic, wait for “Connecting…”, then “Listening — speak now”, and speak.

Two ways to use voice

In plain search, the mic just dictates into the box. In AI Mode, the mic is conversational — after a short pause it submits what you said as a new prompt and keeps listening for your follow-up. See Voice dictation.

Attachment limits when chatting with AI

Symptom: A file you tried to attach to an AI conversation is rejected, or the attach option is disabled.

Cause: Dockli caps AI attachments at 5 files, each up to 50 MB. Extra files or oversized files are rejected with a note on the chip.

What you’ll see

  • A file beyond the count shows “Limit is 5 files”.
  • A file over the size cap shows “Over 50 MB”.
  • The attach menu hint reads “Up to N more files · 50 MB each”, or “File limit reached (5)” when full.

Fix

  1. Remove attachments you no longer need (the “x” on each chip) to free up slots.
  2. Keep each file at or under 50 MB.
  3. You can attach files from your PC (upload) or drag existing Dockli files straight onto the AI bar — both count toward the same limit of 5.

Still stuck?

If AI or dictation still isn’t working after these steps, email support@dockli.io with the exact error text (a screenshot helps), whether your organization is on Essential or Spark, and — for microphone issues — whether your mic works in other Windows apps. See also Ask Dockli overview and Voice dictation.

Still stuck? Email support@dockli.io — a human replies within one business day. Or book a demo.