Everything Omnia Lab Does

What happens when a lead goes unanswered?

Harvard Business Review studied 2,241 companies. Replying within an hour produced 7 times more qualified leads than replying an hour later, and 60 times more than replying a day later. James Oldroyd at MIT Sloan measured that cutting the wait from 30 minutes to 5 raises lead qualification roughly 21-fold. That is why the platform guards the first reply more closely than anything else.

1. 01

   Escalation ladder at 5 / 10 / 15 / 17 / 20 minutes

   A client writes and gets no answer, so the signal climbs. Minutes 5, 15 and 17 hit the shared rep chat, minute 10 goes to the rep in private, minute 20 reaches the manager's DM and the managers' chat.

2. 02

   One live signal per thread

   A new reminder deletes the previous one, so the chat never turns into a feed. The old message is removed only after the new one has actually been delivered.

3. 03

   Nothing piles up overnight

   Overnight escalation stages are cancelled, not postponed. Nobody wakes up to an avalanche. No reminder lives longer than 48 hours.

4. 04

   Smart acknowledgement filter

   A message that is nothing but "ok" or "thanks" closes the thread. The same word inside a real sentence — "ok, I'll do it a bit later" — means a live client, and the SLA keeps running. "Ok?" counts as live too.

5. 05

   A sticker doesn't start the clock

   A file or photo from the client is a full message. A file from the rep counts as a reply. A sticker is neither.

6. 06

   New lead with a Claim button

   The lead lands in the priority chat with a button. If nobody claims it before the timeout, it is assigned by the shift schedule and gets its own reminder chain.

7. 07

   Honest pickup timing

   The clock starts when the deal is created and stops at the first real contact. Only working seconds count, capped at 3 days.

8. 08

   Hourly unanswered digest

   From 10:00 to 21:00 the chat gets a list of people still waiting, with up to 15 client buttons. The digest never hides a live waiting person because of the deal stage.

9. 09

   "No next step" block at 11:00 and 16:00

   It surfaces deals where the client was answered but not a single open task remains. A separate watchdog tracks inbound flow: 3 hours with no client messages during working hours triggers an alert. That catches a broken integration before the sales floor notices.

How do promises and follow-ups get kept on time?

Velocify looked at roughly 3.5 million leads and found that 93% of deals close by the sixth touch, while half of all leads never get a second call. The platform makes those follow-ups happen so the rep doesn't have to.

1. 01

   A promise becomes a task

   Trigger phrases in the rep's reply — "I'll call you back", "I'll send it over" — create a CRM task with a deadline. The Commitments register shows who promised what, by when, and how it was delivered.

2. 02

   No public shaming before the deadline

   30 minutes before the due time the rep gets a private ping. At 15 minutes overdue, private again, with Done and Postpone buttons. Only after an hour overdue does the shared chat hear about it, and at 20:50 managers get an evening summary.

3. 03

   A check before every ping

   The AI reads the conversation first. If the promise is already kept, the task closes itself and nobody gets nudged.

4. 04

   Client promises are tracked too

   "I'll look at it in an hour" becomes a task called "Client promised to come back", due at the client's time plus 30 minutes. A vague "I'll think about it" creates nothing, and the client returning closes the task on its own.

5. 05

   Follow-ups at 1, 3, 7 and 30 days

   Once a day at 15:00 the rep gets a single private digest with buttons. One message, not ten.

6. 06

   Missed calls don't vanish

   A "Call back" task appears with a deadline 30 minutes out. At night Telegram stays quiet and the deadline moves to the start of the working day plus 30 minutes. A real conversation of 15 seconds or more closes the task; 5 to 8 seconds of ringing does not.

7. 07

   Duplicate guard and reschedule tracking

   If an open task already has a deadline within 45 minutes either way, no second task is created. Pushing a deadline twice or more raises its own signal in Important notifications.

8. 08

   A Done button inside Telegram

   It closes the CRM task and leaves a trace. Only the assignee or a whitelisted admin can press it.

How do you check whether a close was justified?

1. 01

   Every spam and lost close gets reviewed

   The AI reads the whole history: chat, comments, the deal card, the transcript of a recent call. The verdict — justified or not — is written down as structured data, not left as an opinion.

2. 02

   Unjustified closes come back

   A wrongly closed deal is returned to the rep, no more than once per deal, with a decision log. Closes made by whitelisted admins are never reviewed.

3. 03

   Storm detection

   More than 30 closes in 2 hours means a human is deliberately cleaning the pipeline. Returns go silent — the system does not argue with a conscious decision.

4. 04

   Junk moves to spam

   At 75% confidence or higher, an off-target deal is moved from lost to spam. That way your loss numbers cover real losses only.

5. 05

   Prepayment is a hard stop

   A deal with a prepayment is never filtered out and never dragged into the cold pipeline. Not even when everything else about it looks like junk.

6. 06

   Cold quarantine with automatic return

   Confirmed spam moves to quarantine at 20:59. A deal comes back only on a real client action: if that client writes or calls, their own card returns on its own. The Filtering tab shows how many deals someone tried to dump and how many came back.

Why don't marketplace auto-replies skew response time?

1. 01

   Service banners don't open an SLA

   "The user viewed the phone number" and similar platform notices start no timer and enter no digest. The trace still stays in the history.

2. 02

   An empty lead never reaches the floor

   If the entire content is just the marketplace's service card, the lead goes to a delayed quarantine instead of pulling a rep away from real work.

3. 03

   Any live text wins

   One sentence from the client cancels the quarantine. "I looked up the number and couldn't get through" reads as a live human.

4. 04

   Errors resolve in the client's favor

   If the check fails, the lead gets distributed anyway. Better one extra message than a lost buyer because of a filter.

How are calls reviewed and scored?

1. 01

   Every call is reviewed

   Transcription, on-target or not, new conversation or continuation, a 0–100 score and recommendations. The result lands in the call log and as a review task in the CRM.

2. 02

   The task is written from the rep's point of view

   What to do now, the score and the logic behind it, what was strong and what was weak, a recommendation and a link to the methodology. Not one mention of AI.

3. 03

   Not a task for every call

   New calls always get one; continuations only if the score is below 65. Otherwise the review itself becomes spam.

4. 04

   Code does the scoring, not the AI

   The AI finds facts and quotes them. The outcome sets the band — 90-100, 80-89, 65-79, 40-64, 1-39 — and process quality decides the position inside that band.

5. 05

   A critical error caps the score, it doesn't subtract

   And it only counts with a word-for-word quote from the call. No quote, no detection. A separate layer holds your company's own rules: banned phrasings, weak closings.

6. 06

   Credit for compensation

   A step skipped early but covered later, or covered by the outcome itself, gets full credit. Asking the name at the end while still closing the sale is not a mistake.

7. 07

   The score follows whoever spoke

   If someone other than the deal owner picked up, the score, the report and the task go to the person who actually talked.

8. 08

   Cooling off after a bad call

   A rating of 50 or below takes the rep off new leads for 2 working hours, and night does not eat the clock: a call at 20:00 means a return at 11:00. If the rep is alone on shift, leads keep coming. Off by default.

What does the daily deal review and coaching look like?

1. 01

   Active deals reviewed at 18:00

   Across four axes: rep, client, deal, process. Managers get a report, reps get tasks, the platform keeps the score history. Optionally the whole run moves to night — batch at 00:05, top-up by 09:15, delivery at 10:00 — which halves the AI cost.

2. 02

   "Needs attention today" comes first

   Priority 0–100, threshold 70, at most 10 deals. The rep score is 0.6 × conversation plus 0.4 × process.

3. 03

   A private daily debrief for the rep

   One thing done well, one growth area with a concrete angle. Not a list of twenty complaints nobody finishes reading.

4. 04

   The "Next step" recommendation

   A CRM task with "why now" and "how to start", deadline required. A new one closes the previous one, and the deal moving forward closes it by itself.

5. 05

   A comment in the deal feed

   It appears when the rep axis drops below 50 or a red flag fires. At most one per deal per day.

6. 06

   "Reviewed by the head of sales"

   Five reasons in two classes. "Work was done" holds until the first new risk; "disagreement with the system" holds until the risk gets worse, up to 14 days. The mark can be lifted and everything is in the audit trail.

7. 07

   A morning plan for each rep

   One private message with four blocks: deals needing attention, promises due to clients today, tasks for the day, and recommendations with buttons.

8. 08

   Weekly analysis on Monday at 10:10

   Personal reviews where code does the math and every conclusion rests on a quote. Time off from the shift schedule is taken into account, and each growth area is folded into the next week's analysis.

When is a lost deal not the rep's fault, and who counts as VIP?

1. 01

   A client profile for every conversation

   Retail buyer, budget that doesn't match the request, reseller, broker, kid or prank, scammer, service-only, unclear — each with a confidence level and quotes.

2. 02

   Not pushing an unqualified buyer is not a mistake

   If the person genuinely isn't a buyer, the rep isn't blamed for the lost sale. Politeness, though, is scored for everyone, every time.

3. 03

   The distrust rule

   A client profile that excuses the rep counts only at 75% confidence or higher and only with a direct quote. Haggling, "too expensive", asking for a discount, short replies and silence are never proof that someone can't pay.

4. 04

   Scammers stay visible to management

   They are excluded from closing statistics but always escalated. The system won't let that conversation quietly disappear.

5. 05

   A budget-order threshold

   Set as an amount, with its own policy for anything under it. The threshold is yours and so is the policy.

6. 06

   VIP detection

   Based on your company's criteria: a flag in the register and in the CRM card field, plus a separate analysis stream. On by default.

Who fills in the CRM instead of the rep?

1. 01

   Deal details from the call transcript

   The name and fields are pulled from the conversation. The name changes only if it is still the default one; anything a human typed stays untouched.

2. 02

   Tasks create themselves

   Call review, rep commitment, client promise, callback, recommendation. Nobody has to type them in.

3. 03

   Reminders matched to the task type

   Client tasks and callbacks: 15 minutes before and at the deadline. Rep commitments: 60, 20 and 5 minutes before, plus at the deadline.

4. 04

   Comments are only appended

   Verdicts, prepayments and handovers are stored as structured records. Nothing a human wrote is ever overwritten. The "Deal essence" and "AI analysis status" fields fill in automatically.

5. 05

   One communication history

   Every message, call, owner change and system event in one place. A new rep can pick up a deal without asking around.

How do reps work inside Telegram?

1. 01

   Buttons instead of typing

   Replied, Claim, No reply needed, Done, Postpone, Handled, Not relevant, and a per-client button. The tap is confirmed instantly, even when the CRM is slow to respond.

2. 02

   Every message names its owner

   So in a shared chat it is obvious who it is for. Button permissions are limited to the right people.

3. 03

   Chats with clear jobs

   Shared rep chat, managers' chat, Important notifications, priority chat for new leads, and a technical chat. Each one does one thing.

4. 04

   Evening cleanup at 20:05

   The shared chat and Important notifications are cleared; messages people wrote by hand are left alone. The evening runs 20:00 last digest, 20:05 cleanup, 20:50 manager summary, 20:59 quarantine report.

5. 05

   Overnight leads in a single digest at 10:05

   One list instead of a burst of separate messages. The morning starts with deals, not with clearing the chat.

What does the manager see in analytics?

1. 01

   What's in the panel

   Dashboard, analytics, reps, modules, errors, settings. Inside analytics: conversion by fact and by cohort with drill-down to individual deals, response time, weekly quality, funnel, filtering, At risk, Win/Loss and Commitments.

2. 02

   The panel reads its own copy of the data

   Data refreshes hourly at :25, a full reconciliation runs once a day, and history goes back 90 days. The panel puts no load on your CRM portal.

3. 03

   A prepayment belongs to its own month forever

   Numbers don't quietly move after the fact, and a fallen-through prepayment shows as its own metric. Anti-gaming: a deal must sit at the stage for at least 20 minutes.

4. 04

   False-start filter

   "Ok" at 12:00, a question at 14:00, the answer at 14:05 — that is 5 minutes, not 2 hours. Otherwise the metric makes the team look worse than it is, and people stop trusting it.

5. 05

   An Exclude button

   A disputed response-time measurement can be dropped, but with an audit record of who dropped it. Nobody edits a number in silence.

6. 06

   Win/Loss on every close

   The full history is gathered and broken down into factors, key moments and the deciding driver. Empty deals never dilute the win rate.

7. 07

   Junk stays out of conversion

   Spam stages and deals with "spam" in the name are excluded. Rep score and deal score are separate metrics in separate columns.

How are leads distributed across shifts?

1. 01

   Shift schedule from a Google Sheet

   Synced every hour. Matrix format: people down the rows, days across the columns, a shift code in each cell. The codes live in their own reference table.

2. 02

   Leads go only to people on shift

   Time-outs after a bad call are taken into account too. A lead never lands with someone who isn't working today.

3. 03

   Reassigning deals from people who left

   Daily at 10:00, up to 25 deals per run, early stages first. Two sources are covered: chat-born deals and deals created by a call. Manual assignments by a manager are left alone.

4. 04

   A check every minute

   New deals stuck on the service account are found and handed out. Nothing sits ownerless for half a day.

5. 05

   The rep directory is the source of truth

   A deal with an unknown owner triggers no pings and never enters paid analysis. Nobody gets reminders about someone else's work.

How reliable is the platform and what does AI cost?

1. 01

   Lost messages are recovered

   Every 30 minutes, plus two-way syncing of the "replied" status. The dashboard shows a counter of how many messages were recovered.

2. 02

   Two transcription systems

   They back each other up. A call is only lost if both die at once.

3. 03

   A backup AI provider

   Failover is automatic and the reason shows in the Errors section. Your reviews don't stop because of someone else's outage.

4. 04

   A daily AI budget

   $15 a day by default, with alerts at 50% and 100%. Only deferrable jobs are throttled — client-facing flows never are.

5. 05

   A cap on close reviews

   20 per hour. Anything above that is postponed, never dropped.

6. 06

   Fail-open by design

   Any failure resolves in the client's favor: the lead gets distributed, the conversation stays visible, a filter switches off rather than hiding data.

7. 07

   Monitoring and memory

   Platform uptime, AI provider status every 3 minutes, an error log. Planned maintenance doesn't count as downtime. On Sundays at 23:30 long deals are consolidated — history older than 14 days is compressed while the latest 20 records stay intact.

What does Omnia Lab integrate with?

• Bitrix24 — the primary CRM. Deal events, stage robots, tasks, timeline comments, custom fields, pipeline and stage changes.
• amoCRM — part of the integration list: we implement and configure it.
• Messengers and open channels. Conversations arrive in full; a file or photo from a client counts as a real message.
• Telephony — call completion events. For phone systems that send no events, the platform polls telephony itself every 2 minutes.
• Telegram — all chats, buttons, digests and summaries.
• Google Sheets — the call log and the shift schedule.
• Classifieds and marketplaces, with recognition of their service banners.
• Two transcription providers and an AI stack that fails over to a backup model automatically.

What does the system deliberately not do?

Half the value sits in where the platform stops. These limits are on purpose.

• It doesn't replace your CRM. It runs on top of your Bitrix24, and reps stay in the interface they know.
• It doesn't call anyone out publicly before a deadline. Until the time is up, the conversation stays private.
• It doesn't argue with a conscious human decision: more than 30 closes in 2 hours and returns go silent.
• It doesn't punish a rep for not pushing a genuinely unqualified buyer. Politeness is still expected from everyone.
• It doesn't treat haggling, "too expensive", short replies or silence as proof of low budget.
• It doesn't accept a critical call error without a word-for-word quote.
• It never filters out a deal with a prepayment and never moves it to the cold pipeline.
• It doesn't reach into everything: terminal and won stages, out-of-scope pipelines and reps marked "no reminders" are all skipped. The list of pipelines and stages with no SLA is configurable.
• It doesn't mention AI to reps. Tasks and reviews are written in plain human language.
• It doesn't turn chats into feeds: one live reminder per thread, and the chat is cleared in the evening.

What results do existing clients see?

Deals with a prepayment

from 4–7% to 13%

Median first response time

from 32 minutes to 6.5 minutes

Revenue

+33% in 4 months with no extra headcount

Implementation

3–5 days

Free audit

3–5 days

Measured before and after rollout at existing clients, including a premium car dealership. Implementation and audit are our standard project timelines, not measured results.

How long does rollout and configuration take?

1. 01

   Deployment in 3–4 hours

   The Quick Launch wizard sets up a new client. Until you switch to live mode every module sleeps and writes nothing into your CRM.

2. 02

   33 modules, each with its own switch

   Turn off anything you don't need and turn it on later. Starting with two or three is perfectly fine.

3. 03

   85 settings in the panel

   They apply in under 60 seconds, with no restart and no developer. Stages, trigger phrases, VIP criteria and conversation rules are tuned to your company.

4. 04

   Your brand, not ours

   Your company name is the big one; "powered by Omnia Lab" sits small. Reps see their own system.

5. 05

   Full implementation in 3–5 days

   With a free 3–5 day audit before it. We look at your funnel, response times and closes. Then we show you where money leaks out.