M cenaly.ru
⚙️ Features

💡 Copilot: Analytics & Insights

Voice statistics, daily summaries, conversation context, change history, proactive AI suggestions

Documentation

Analytics and Smart Suggestions

Copilot is not just a command executor, but also an analyst. This section covers how to retrieve information by voice, maintain conversation context, and what suggestions Copilot provides on its own.

Why businesses need this. Business questions are resolved in seconds without opening reports: ask "revenue by channel for the month" — and the table arrives right in the chat. And Copilot runs data quality checks ("what is missing in the menu", "translation report") with a single phrase, immediately offering to fill the identified gaps.


Quick Voice Navigation#

Command What it does Examples
Open QR codes Table and device codes section "Open QR codes", "Show codes"
Open dish tags Dish card on the tags and allergens tab "Open Caesar tags", "Tags for Caesar"
Open labels Label and price tag printing section "Open labels", "Open labels"
Build chart Takes you to the dashboard builder with the prompt already filled in "Build a daily revenue chart for the month"
Access and roles Staff access settings "Access settings", "Permissions"
Open analytics Reports and statistics section "Show finances", "Open statistics"
Open locations Account venues list "Open locations", "Show my venues"

Why it matters. The more features there are, the harder it is to find the right section. Voice is the shortest path: 1 second versus 3–4 clicks through the menu.


Command What it does Examples
What's missing in the menu Scans item cards and shows a summary of gaps: descriptions, photos, allergens, translations, prices "What's missing in my menu?", "Check item card quality"
Translation report In-depth check: where and in what languages translations are missing "Translation report", "Where is the Turkish translation missing?"
Complete translations Finds gaps and fills them with auto-translation "Check translations", "Fill in missing translations"
Filter by criteria Shows a list of menu items based on a condition — without changes "What is under 5 GEL?", "Show all items in the Bar category"
Duplicates and structure Finds similar menu items and oddly organized categories "Find duplicates", "Fix grouping"

Example. "Translation report" — the copilot scans the entire menu and sends a breakdown: how many menu items and in what languages lack translation. Then, with a single phrase "fill in missing translations", the gaps are filled with auto-translation.


Order Information#

Copilot compiles the report directly in the chat: you specify the metric, breakdown, and period, and it returns a ready-to-use table.

What to ask Examples
Revenue “Weekly revenue”, “How much did we make last month?”
Order count “How many orders today?”, “Orders over 30 days”
Average check “Monthly average check”, “Average check by day of the week”
Top menu items “Best sellers this week”, “Top dishes of the month”
Shift summary and cash “Shift summary”, “How much is in the POS right now?”, “How much did we sell this shift?”
POS terminal check “Is the POS working?”, “What's with the POS?”

Breakdowns: by day, by hour, by day of the week, by sales source and channel, by payment method, by order type, by ad campaign. Periods: today, yesterday, week, month, last month, last 30 days, or explicit dates (defaults to the last 30 days if no period is specified).

Copilot does not make up breakdowns that the report does not support: if asked for “revenue by category”, “by waiter”, or “by table”, it will honestly reply that this breakdown is not calculated yet and suggest the closest alternative — for example, top dishes or revenue by sales channel.

Example scenario. After closing, the owner asks: “Revenue by day for the week” — Copilot sends a daily table right into the chat. Need a chart instead of a table? “Build a chart of daily revenue for the month”, and it will open the dashboard builder with the chart already generated.


Dialogue context — conversational intelligence#

Copilot remembers the last five messages in the dialogue — its own and yours combined, which is about two to three exchanges — and understands clarifications, corrections, and follow-ups.

Feature Examples
On-the-fly correction You said "Margarita 22" → did not confirm → "No, not 22, make it 24" — Copilot updates the number while preserving the dish context
Clarification You said "Khachapuri" (and you have 3 varieties) → Copilot displays a list of matches: "Imeretian, Megrelian, or Adjarian?"
Chaining "Hide Cheesecake", "now Tiramisu", "and Brownie" — each subsequent command operates within the context of "hiding"
Grouping in a single phrase "Hide all chicken dishes", "What is cheaper than 5 GEL?" — Copilot gathers the selection based on the criterion and shows a preview
Confirmation or cancellation Any change is first shown as a command card — executed only after your "yes"; read requests (reports, search) are executed immediately

Why it matters. A conversational interface should feel like talking to an assistant, not dictating a rigid command. Copilot carries on a conversation rather than executing steps according to a protocol.


Dish Photo History and Restoring a Previous Version#

Every replacement or deletion of a dish photo is saved — old images are not lost. Using your voice, you can both view this history and restore any of the previous versions.

Command What it does Examples
Show photo history Feed of previous dish photos directly in the chat (executed immediately, changes nothing) “Show photo history of Khachapuri”, “Show photo history of Caesar”
Restore previous photo Restores the previous version, the version before it, or the very first image “Restore previous photo of Khachapuri”, “Restore original photo”, “Roll back photo to the previous one”

If multiple dishes match the description, the copilot will first ask to clarify which one you mean, and restoring a photo — like any change — will be executed only after confirmation.

What is not available yet. There is no general log of "what I changed today" or bulk rollback of edits via voice: only the specific dish photo is rolled back. For changes across different sections, check those sections directly — for example, articles and website pages have version history, and counterparty cards have their own event feed.


Copilot Suggestions#

The copilot offers suggestions in two scenarios.

1. You asked — it checked. A single phrase triggers an analysis that concludes with a ready-to-use action plan:

Request What you get in response
"What's missing in the menu?" A prioritized summary of gaps: descriptions, photos, allergens, translations, prices, calorie counts
"Translation report" Where and in which languages translations are missing — across dishes, categories, add-ons, and variants
"Find duplicates" A list of similar items with a merge plan that you approve before applying
"Fix categorization" Dishes that are "in the wrong category" and a proposed new menu structure — also reviewed before execution

2. You asked about a feature you don't have. If you ask about a capability, but the corresponding section or extension is disabled, the copilot provides a direct answer and suggests enabling it — via a card with a brief "why you need this" rationale. A suggestion appears only in response to your question: the copilot never offers anything proactively without a prompt. If you decline, it remembers this — and the next time you ask about the same feature, it will transparently remind you: "You previously declined this suggestion; if it's relevant now, enable it here." Once you enable something, it stops suggesting it altogether. A maximum of one suggestion is provided per response.

The suggestion card appears only for text-typed commands: during voice input, the transcribed phrase does not reach this mechanism. It works exclusively in the admin panel copilot — the kitchen display, courier app, and waiter app do not have these cards.

The copilot does not initiate conversations on its own — it will not message you first about dropping sales or rising meat prices. A screen that collects such alerts automatically already exists in the product: Needs Attention — which, since late August 2026, is no longer a separate sidebar item, but Work Chat → "Action Required" section → "Monitored" tab (the old link /attention automatically redirects there; the general Work Chat layout is described in Guest Chat). This tab displays overdue B2B invoices, pending proposals, CRM tasks for today, recent negative reviews, POS risks, telephony issues, documents (expiring and unread), and significant guest debts. Urgent matters — returning an inquiry call, confirming a reservation, accepting an order — live in the adjacent "Urgent" tab within the same section and can be resolved directly there, with a single button per item. This screen cannot be opened by voice — access it through the Work Chat.


Privacy and Your Data#

  • Voice is transmitted for recognition over a secure channel and is not stored: only the command text remains after recognition.
  • The request text, model response, and selected action are recorded in your account's AI request log — this is needed for analyzing edge cases and improving response quality.
  • In the Russian perimeter, speech is recognized and processed by Yandex services (SpeechKit and YandexGPT) — neither audio nor text leaves this ecosystem.
  • Suggestions are built on your venue's data — data is never shared between different restaurants.
  • Chat history is stored locally in your browser and can be cleared with a button in the assistant settings.
  • Recognition and responses are billed using account AI credits; if the microphone is unavailable, the copilot accepts the same commands as text.

See also#