How to connect AI to Mango Office telephony: missed calls, SL, and callbacks

Four ways to "connect AI to Mango Office" — and why the result differs

When a head of sales searches for "a neural network for telephony," four classes of solution hide behind one query. The difference decides what actually comes off your plate and your supervisor's.

1. Connector layer. A no-code service links Mango to a spreadsheet or CRM on a rigid rule: a call event arrived → create a row. This is not AI, it is data transfer. It cannot measure SL, catch missed calls with no callback, or analyze recordings.
2. Mango's built-in dashboard statistics. The "Call statistics" section honestly shows how many calls were answered, how many were missed, what percentage was answered within the target (SL), and the load on groups. Useful for seeing the picture, but it is a storefront: it shows "+1 to missed" and leaves the callback, the task, and the analysis to a human.
3. Single-purpose voice bot / auto-dialer. Performs one scripted action: dial through a base, read out a script, take a keypress in the IVR. Fast, but blind to the meaning of the conversation and disconnected from reporting and quality.
4. Ready-made AI agent. The same agent power, but configuration is in plain language. It connects to Mango over the official VPBX API and runs call operations itself: compiles daily call and SL reports (incoming, answered, missed), catches missed calls in a group with no callback and sets a task or makes the callback, analyzes recordings and their transcripts, finds lost clients, and runs quality control on a sample. You set the target in words — it works to it.

From here on, "agent" = only the fourth class. A connector and a bot are tools; the built-in statistics are a storefront, not an autonomous performer.

Configuration in words, not development

This is the main thing that sets a ready-made agent apart from everything else. A connector forces you to assemble a no-code scenario. The dashboard statistics live in a separate window and only display. A bot has to be programmed with a scenario. With Samreshuuu you simply explain the task, as you would to a supervisor: "every morning, send a report on yesterday's calls and SL; if the sales group has a missed call with no callback within 15 minutes, set a task for the manager or make a callback; once a week, take a sample of conversations and check them against the quality checklist." Need something more complex — the agent assembles the required query for the task itself. No programmer and no manual integration setup.

How to connect AI to Mango Office: step by step

The connection works over Mango Office's official VPBX API — with a key from your account dashboard, without handing over a login and password. The agent forms the signature of each request (sha256(vpbx_api_key + body + secret key)) itself — you do not need to know about it.

1. Get the API credentials in your Mango dashboard. In the VPBX dashboard, enable API access and copy the vpbx_api_key and the signing key. This is the platform's standard mechanism; the secret key is never sent in a request.
2. Connect the agent. In Samreshuuu → "Settings → Integrations → Mango Office," paste the key and the signature. The agent immediately sees call history and statistics, groups, employees, and recordings. No dashboard login and password needed.
3. Describe the rule in words. For example: "compile the daily SL for the sales group; missed calls with no callback go into tasks; for each negative conversation in the sample, send a summary and escalate."
4. Choose the control mode. Routine (reports, the missed-calls digest, callback tasks) — on autopilot; outward-facing actions (a callback to a client, SMS) — in "draft → confirmation" mode.

What the agent actually does with Mango telephony

The key difference: the built-in statistics show the numbers, the bot runs one scenario blindly, and the agent uses Mango as the source of truth about calls — and runs the reporting, callbacks, and quality control itself instead of hinting them to a human.

Why the "missed call → callback → SL → revenue" chain decides everything

Most "AI for telephony" tools work piecemeal: dial a base here, show a chart there. But a sales team's money is lost in one quiet gap — between "the call was missed" and "the client was called back in time." Picture it: at lunchtime a new client calls, the manager is busy on another line, the call drops. The dashboard statistics neatly add "+1 to missed" — and that is it. Nobody calls back, because the callback is nobody's job, and by evening that missed call dissolves among a hundred rows of the report. Meanwhile the client has already reached a competitor, and you will never even know you lost the deal: a number in a report does not scream, it stays silent.

The agent closes exactly this gap. It sees the missed-call event (by group, from the call history), checks whether a callback happened, and if not — makes the callback itself through Mango or sets the manager a task with the number and the time of the call. And it measures SL right away, so the dip does not repeat: the agent sees that the sales group consistently loses calls between 1 and 2 pm, and says so. Then it takes the call recordings, produces summaries, and checks on a sample how managers handle the script. This turns "AI for calls" from a storefront into a performer that closes the hole between a lost call and money.

The meaning of the conversations themselves — why the client left, which objections keep repeating, where the manager said the wrong thing — is a separate layer: how a transcript becomes an action is covered in detail in the piece on AI speech analytics for calls. This article is about operations: who was not called back, what the SL is, whom to put under quality control. And so that the callback task lands straight in the manager's pipeline, the agent links Mango to amoCRM or Bitrix24 — the deal and the call end up in one place.

Honestly about the downsides

If all you need is to see call and SL reports, Mango's built-in dashboard statistics are enough — a separate agent is overkill here. If the task is simply to dial through a base on one script, a voice bot will do. The agent is justified where you need to tie together missed calls, callbacks, SL, and conversation quality and act on a rule, rather than execute a single scenario. And it needs the rules described in words once — a bit longer than switching on a ready-made auto-dialer, but it works on more than one script.

Checklist: how to choose AI for Mango Office telephony

1. Do you need to see reports, dial a base, or run it all together? Seeing — the built-in statistics. Scripted dialing — a bot. Tying together missed calls, callbacks, SL, and quality — the agent.
2. Does the solution catch missed calls with no callback and carry them through to an action? Without this, the client who did not get through is lost silently.
3. Does it analyze the meaning of the conversation, not just its duration? A summary and quality control from the recording is the difference between "the call happened" and "the call was handled."
4. Does it measure SL by group rather than in one big pile? A dip is visible only in the cut of group and hour.
5. Connection over the official VPBX API, and is there a confirmation mode? Do not hand over a login and password; start outward-facing actions (callback, SMS) with "draft → confirmation."

Frequently asked questions

What is AI for Mango Office telephony and what can it do? AI for Mango Office telephony is an agent that connects to the cloud PBX over the VPBX API and takes over call operations: it compiles call history and statistics (CDR), measures daily SL across incoming, answered, and missed calls, catches missed calls with no callback and sets a task or makes the callback, transcribes recordings and sends summaries, finds lost clients, and runs quality control on a sample. The difference between the classes is large: the built-in statistics only display, a bot dials along a script, while an AI agent like Samreshuuu runs all of it itself on a rule you set in plain language.

How do I connect a neural network to Mango Office? Get the vpbx_api_key and the signing key from your VPBX dashboard, then in Samreshuuu open "Settings → Integrations," pick Mango Office, and paste them in — no login and password needed, API requests are signed automatically. Then describe the rules in plain language: which call and SL report to send, when to react to a missed call, how to sample conversations for quality control. The agent does it itself and sends outward-facing actions — callback, SMS — for confirmation. The connection takes a couple of steps and needs no developer.

I need an AI assistant for a sales team's calls — where do I start? An AI assistant for a sales team's calls starts with the most expensive gap — missed calls with no callback. Connect the agent to Mango over the API and set the target: a missed call in the sales group = a task for the manager to call back within 15 minutes, an overdue one — escalation to the head of sales, daily — a digest of SL and average callback time. Then add quality control (a sample of conversations, summaries from recordings) and a link to your CRM, so the call and the deal are in one place. Samreshuuu runs this from the target, rather than showing numbers after the fact.

Is it safe to give AI access to telephony? Yes, if the connection runs over the official VPBX API with a key from the dashboard (rather than a login and password), requests are signed with a secret key that is never transmitted, and the service has an "on confirmation" mode for outward-facing actions. Samreshuuu connects over the API and by default sends contentious steps — a callback to a client, sending an SMS — for confirmation, while it runs reports and recording analysis automatically. The key can be revoked in the Mango dashboard at any time.

Will an AI agent replace Mango's built-in statistics and a voice bot? Often yes. The built-in statistics and a bot each solve one piece: one shows numbers, the other dials along a script. The agent does both, connected — it compiles reports and SL, catches missed calls and makes the callbacks itself, analyzes recordings — so it usually replaces the "reports + auto-dialer" combination. But if all you need is to view statistics without acting, the built-in dashboard reports remain simpler.

---

Last updated: July 2026.

Sources: MANGO OFFICE VPBX API documentation (authentication with vpbx_api_key and the sha256(vpbx_api_key + json + vpbx_api_salt) signature; two-step statistics export /vpbx/stats/request → /vpbx/stats/result and the extended /vpbx/stats/calls/..., with a window of no more than a month; the commands/callback callback command, call recordings, transcripts and summaries, employee and group directories); MANGO OFFICE help on the "Call statistics" section (answered/missed/service level), callbacks, and SMS sending; public descriptions of Albato, Make, ApiX-Drive, and auto-dialer voice bots.