Intelligence & Reports

Conversation Intelligence

Built-in tools that read and learn from every conversation automatically, no setup or manual tagging required. There are three parts, running across all your channels:

• Sentiment detection reads the emotional tone of each message, so you can see who is happy, frustrated, or needs urgent attention, and flags a customer the moment their tone shifts.
• Conversation memory automatically summarises each conversation when it ends and builds a growing profile of each customer, so any agent can pick up mid-thread fully prepared.
• AI quality scoring rates every AI reply on accuracy, tone, and helpfulness, flags weak replies for review, and shows quality trends over time.

What Reports is & who can use it

Reports is the analytics workspace for your organization. It collects activity from every connected inbox and channel and presents it as dashboards, trends, and tables, so you can answer questions like "Is the AI handling enough on its own?", "Are customers happy?", "Where is my team falling behind?", and "What is this costing me?"

Who can open Reports

• ✓  System Owners
• ✓  Organization Admins
• ✓  Individual users (single-user accounts)
• ×  Regular (non-admin) members, they won't see the Reports menu

If a teammate can't find Reports, it's almost always because their role doesn't include reporting access. An Organization Admin can adjust their role under Team Management.

Getting around

Opening Reports. Reports lives at the /reporting page. Open it from the main app and you'll land on the Overview.

The six sections

A left-hand menu groups every report into six sections, click a section to expand its reports:

• Overview, the single-screen business summary
• Conversations, messages, channels, quality, satisfaction, and your team
• AI Assistant, how the AI is performing, its confidence, and its cost
• Knowledge Base, your uploaded documents and how the AI uses them
• Marketing, WhatsApp campaigns, contacts, templates, and messaging cost
• System, account growth, system health, billing, and team activity

The date-range filter

The top bar has a date picker with quick presets: Today, Last 7 days, Last 28 days, Last 30 days, This month, Last month, Last 90 days, This year, All time. Changing it re-runs the visible report for that window. The view opens on a recent window (about the last month) by default.

Live data, sharing & more

• Live data & refresh. A Live badge and Refresh button sit in the top bar. Numbers are calculated fresh each time you open or refresh, there's no overnight delay (except long-term growth trends).
• Sharing & exporting. Reports can be shared by email, WhatsApp, or copy-link, and exported as a screenshot/PDF or CSV from the report screen.
• Big tables page in batches. Long lists (conversations, escalations, recipients, contacts, snapshots) are paginated, load a page at a time and click through.
• On mobile, the section menu becomes a dropdown at the top of the screen; the reports themselves are the same.

How to read a report (shared concepts)

A handful of ideas appear again and again. Learn them once here.

AI resolution %

The share of your outgoing replies that the AI sent without a human editing them. High = the AI is doing the work on its own; low = your team is rewriting a lot of its drafts.

CSAT vs AI self-rating

Two different "quality" scores, don't confuse them. CSAT is your customers' satisfaction on a 1–5 scale, from ratings they submit. AI self-rating (also called AI confidence) is the AI grading its own answer, also 1–5, an early-warning signal, not customer opinion.

Escalations & the 24-hour SLA

An escalation is a conversation that left AI handling and needs a human. Any escalation still open beyond 24 hours is flagged as breaching SLA, so you can chase it.

Reply-time SLA & percentiles

Reply time is how long a customer waited for an answer. Because a few very slow replies distort an average, reports use the median (the middle value) and percentiles (e.g. p95 = "95% of replies were at least this fast").

SLA funnel, heatmaps & funnels

• SLA funnel, three bars showing the % of replies sent within 1 hour, 4 hours, and 24 hours.
• Heatmaps, a 7-days × 24-hours grid where darker cells mean more activity; great for spotting busy hours to staff or schedule campaigns.
• Funnels, a series of shrinking stages (e.g. Opened → AI replied → Resolved → Escalated) that show where conversations drop off.

Freshness buckets

Documents grouped by age: Current (updated in the last 7 days), Aging (8–30 days), Stale (older than 30 days). Stale documents may give the AI out-of-date answers.

Cost figures are estimates

AI/token costs are calculated from a fixed per-model price table, so treat them as close estimates for budgeting, not an invoice. Subscription billing uses your plan's currency (typically KES); WhatsApp messaging costs are in USD.

The report sections

Every report is grouped into sections. Each section has its own page, with each report explained in plain business terms and a short How it's calculated note. Jump to the area you need: