What's New

A running summary of how Aelo has evolved. We ship frequently; this page highlights the customer-visible changes — new capabilities, meaningful improvements, and the occasional important fix. For a full role-by-role guide to everything the platform does today, see the User Guide.

Dates reflect when each change reached production. Internal-only work (infrastructure, tests, refactors) is omitted.

September 2026

Every objection in one place, and one quality score everywhere · 2.47

Objections in full, not only the dropped ones. On the dashboard’s Quality tab, All objections → opens a screen with the objection reasons your customers raised in the period — how often, in their own words, with quotes that open the call, the team’s best replies taken from real conversations and the change against the previous period; objections that fit no reason and older wording that could not be recognized are shown as separate shares. For Clients and Supervisors (a Supervisor sees their projects); in the Bitrix24 app it opens inside the app.
One quality score everywhere. The score in the records list, its filter and the record card now match the dashboard: the mean of the conversation criteria the analysis rated (up to six), with one decimal. Ranges such as 0–79 and 80–100 meet without a gap.
The budget shows what you are charged. “Used” and “remaining” on the dashboard and in Settings → Plan & Billing → Budget are now your usage charges for the budget period — the same amount the threshold notifications use. The budget never pauses analysis.
Supervisors start with their own projects. With no project selected, a Supervisor’s dashboards, leaderboards, reports and digest cover the projects they supervise, and several can be picked at once. These views cover the projects a Supervisor is assigned to — assign them in Settings → Team.
Leaderboards and trends cover the whole team, not the first 20 agents, and Highlights examples open in place.
Bitrix24: search on the Records tab now finds records, as in the web app.
Chats read by name. On the record screen, a Bitrix24 chat reply shows the sender’s name, and quoted text is muted.
Your language, your formats. Dates, the time-zone offset, the currency sign and labels follow the interface language.
Script scoring status is clear. While script adherence scoring is off, the script settings say so instead of showing the rules as active.
Privacy. A phone number, e-mail, card or IBAN split between two speaker turns is now masked, and a deleted project leaves reports and search usually within half an hour instead of days.

A Client always stays in charge, and a quote keeps its author · 2.46

Your organization always keeps a Client. Only a Client can change or remove another Client, and the last Client cannot be demoted or removed by anyone, themselves included. On the Team screen a protected Client has a badge instead of the role menu and an inactive remove button — hover over either to see why.
A quote in an open-line chat is no longer the sender’s words. When a Bitrix24 open-line message quotes an earlier one, the transcript shows the quote as [Quote] «…» — on screen, in exports and in the CRM — and the analysis no longer attributes the quoted words to the sender. A message that is only a quote is not counted as a reply.
Assistant questions and answers follow your retention period. A question and its answer are deleted when the record asked about or any record the answer was built from is deleted — usually together with it, otherwise by a routine cleanup — and in any case at the end of your organization’s retention period, including questions whose answer failed.
Sale probability, clean and on scale. The sale probability sent to your CRM and the API no longer carries rounding noise (55, not 55.00000000000001), and when the analysis returns a value off its scale, the field is now left empty instead of guessing what it meant.

Search that reaches every record you can see · 2.45

Search finds your records however many others match. In a large organisation, a search in the records list of the web app could stop before it reached the records you can see when thousands of others matched first. It now keeps going until it finds them, loading more as you scroll, with a Search further button when a long search needs another step.
You’re told when search runs in a simplified mode. In the web app, if part of the results came from our backup search, a note above the list says some records may be missing and suggests trying again in a minute.

Ask about the whole deal, and find a name however it’s typed · 2.44

Ask the assistant about a whole deal. When a call or chat is linked to a deal, lead, contact or company in your CRM, Ask about the deal (or the lead, contact, company) in the record header answers from its conversations you can see — the ones most relevant to your question in full, and it tells you when a long history had to be shortened. Quotes come with it, and a click on a quote from another conversation opens it. The history of your questions about the deal stays with the deal. Available on Pro and above.
Search finds names whatever the letter case. In the records list of the web app, “олег” now finds “Олег” — search no longer depends on capital letters in Cyrillic.
“Analysed” means analysed. The number of analysed conversations on the dashboard cards now counts only calls and chats that actually have an analysis — missed calls and calls too short to analyse are no longer included, so it may be lower than before.
Fairer analysis of your agents. The analysis now checks who actually said a line before holding it against the agent, so the client’s words — including swearing and mistakes — are attributed to the agent far less often.
Handled objections stay on the card. The “Objections the team drops” card also lists objections your team handled in every conversation, at 0 % in a muted colour, so you see what comes up, not only what goes wrong.

News that finds you, and a competitor counted once · 2.43

See what’s new right after an update. The first time you sign in after a release, a short window shows what changed — in your language, in the web app and inside Bitrix24 alike. It appears once per person, not once per browser.
One competitor, one row, however it was spelled. On the Highlights tab, competitor names are merged through your glossary: “Lemken”, “lemken” and the same name in Cyrillic become one row, and a conversation that mentioned both spellings counts once. Your own brand, the platforms you service, suppliers, partners and colleagues from the glossary are never listed as competitors. Once your glossary lists at least one competitor, a name it doesn’t know yet is marked “not merged”, and under the list you see what share of the names heard that is, with a few examples — send those spellings to support and they merge, past conversations included, without re-analysing anything.
The assistant panel keeps up with you. It names the record it is answering about right under its title, and when you open another record it moves there with that record’s own conversation and draft. An answer that is still being prepared is never lost: the panel waits for it, then offers to move over.
Install Aelo straight from the Bitrix24 Marketplace. Aelo is published in the Marketplace, and the connection wizard, the integration page and the installation guides now open its card directly.

An answer you can check, and a way back to a plan · 2.42

Ask a question about a call and get an answer you can verify (beta, Pro and above). Open an analysed conversation and ask what you want to know about it — what the customer objected to, what was agreed, whether the agent followed the script. Where the assistant can back a thesis with this record’s transcript, the quote is shown under it: pinned to one place in the recording, it carries a timecode that moves the player to that moment; unconfirmed, or found in several places, it is marked as such instead of being passed off as verified. A thesis it cannot back stands without a quote rather than with one that only looks like evidence. Nothing has to be taken on trust. For now the assistant answers about the one record you have open. Included from Pro upward and on there by default; tell us if you would rather it stayed off. Conversations with the assistant are kept for up to 365 days and are deleted together with the record.
A Highlights tab that sums the period in plain words. On the client and supervisor dashboards: conversations from received to analysed, the three most frequent issues by scoring criterion, the conversations with no high or critical attention flag beside the list of the flagged ones, the average score with the best and the lagging agent, what the conversations were about, serious product complaints and competitor mentions. A supervisor sees only their own projects.
Objections ranked by the reason behind them. Every objection is now sorted into one reason from a closed list of 19 — price, not now, delivery and availability, needs approval, and so on — so the same reason is counted together however differently customers phrase it. Objections analysed before the list existed are sorted through a stored mapping of past wordings; the share that could not be recognised is shown on the card rather than folded into the rest.
A cancelled subscription no longer stands between you and a new plan. After a cancellation every plan card offered “Manage subscription”, the button opened a Stripe portal with nothing left in it, and buying a plan again was unreachable from the app. Plans are purchasable again, and the page no longer promises a cancellation for a period that has already ended.
The delivery webhook says whether the fact check actually ran. Since 2.41 that pass is skipped when an analysis has no extracted fields and no agreements — there is nothing for it to guard. The verification block now carries fact_check beside the status, so an automation can tell a skipped pass from a completed one. status itself is unchanged: the new field is additive and safe to ignore.
A call transcript names two speakers, and “erm” is no longer counted as a word. Hesitation sounds are hidden in the text you read, and the speech score now goes by how often they occur per minute rather than by the raw total — a long conversation is no longer marked down for the same manner of speaking that a short one scores well for.

Verification where it protects something · 2.41

Fact-checking now runs where there is a claim to check. This pass exists to guard the fields and the agreements an analysis sends to your CRM. With those switched off it had nothing to verify and reviewed the wording of the summary instead, marking two records in three as unverified over objections to how a sentence was phrased. When there are claims, the summary is reviewed alongside them; when there are none, it now reaches your CRM without that review. The safety review that runs on every analysis is unchanged.
An objection whose quote had to be hidden for privacy stays visible. When a supporting quote contained something like an order number, privacy minimisation removed the quote — and the objection disappeared from the record and the printed report with it. The objection now remains, with its section and its reason; only the quote is withheld.
A telephony source connected for audio only stays audio only. Where telephony supplies recordings for calls your CRM owns, its call history is no longer turned into a second set of records beside them.
Speech recognition runs on a single provider. The original provider has been removed from the product, the sub-processor register and the credential store. Transcription is unaffected: the provider added in 2.6 has been the platform default since July.

Verification you can see, and calls that are not written off · 2.40

An analysis that could not be fully verified now shows you what it could not confirm. Where the panel used to report only that something was unconfirmed, it now names the status, the reason, and each specific claim: which section it came from, what kind of problem it is, and — whenever an exact quote can be found in the recording — the words that support it. The same appears in the printed report, so the reason travels with the record.
The quote under “In the conversation” is the conversation’s own text. It used to be the wording the model produced; it is now a slice of the transcript, matched word for word. A quote that does not match exactly is not shown at all — the objection stays, with its section and reason, but nothing is put in someone’s mouth.
The analysis report now always reaches your CRM. An analysis with an unconfirmed claim used to produce no CRM report at all, which left the salesperson with nothing instead of something qualified. The report is now published with a notice of what was verified and what was not.
The script-adherence summary matches the breakdown beneath it. A step’s score follows the criteria that actually matched and the weights your script gives them; an optional criterion that did not happen no longer costs the step anything; a step with no criteria is left unevaluated instead of being marked failed; and skipped steps are named rather than left to inference.
“Turn evaluation off” and “change the script” are now separate actions. The “Pin script” toggle promised behaviour the analysis never had, so it is gone, and the guides that described it have been corrected.
No scores reads as no scores. An employee with nothing scored yet was shown “My Score 0/100” — a real evaluation of nothing. An absence now looks like an absence, on the headline, the trend, and the progress card.
The analysis preview is computed for the project you are looking at, and shows the prompt the server actually built for it.
Signing up and setting up your first project is one continuous screen instead of a sequence of separate pages.
Fewer calls are written off as having no recording. A call now reaches the audio index before anything decides that no recording will come, and the waiting period is counted from the last time that index was written rather than from the call itself.
Call duration is reported as a whole number of seconds, rounded once when the record is created — including in the public API, where the documentation now says so.

Numbers you can trust, and calls that stop going missing · 2.39

A record halted because its source was disconnected can be sent back into processing. When an integration is disconnected, a call still waiting for its audio stops with that reason — and it used to be a dead end: you saw the reason and could do nothing about it. Reconnect the source and you can now return the record to processing from the record itself. The action appears only where it can actually help, and a failed attempt leaves the record exactly as it was.
A quality sub-metric score is the score the rubric gave. Scores of 10 or below were being multiplied by ten, on the assumption they came from an older 0-10 rubric — so a genuine 5 out of 100 appeared as 50. We measured every scored analysis and found no trace of that old rubric, and 494 real low scores being inflated by it. Those analyses now show a lower and correct number. If you keep our scores in your own reporting, the values you pulled earlier for those calls will no longer match.
A colleague is one person, on every screen that ranks them. An employee who exists in two connected systems — or under two ids in the same CRM — was counted twice in the ranking denominator, could have another person’s calls pulled into their card, and could appear at one position in the daily digest and another on the page that email linked to. All three now follow one rule, and coach notes are found by the person rather than by whichever id the request carried.
Calls with no recording are no longer lost. A call whose provider mints the audio link later, or that a phone system never flagged as recorded, was being discarded on the way in without appearing anywhere. It is now ingested as a record. A message the platform does refuse now shows up in the Activity Log with a reason, instead of being visible only to us.
A long or difficult call still produces an analysis. When the first attempt does not fit, the retry now keeps the model’s reasoning and asks for a more compact answer, rather than returning nothing.

One score with one meaning, and integration keys under lock · 2.38

An agent’s quality score means the same thing everywhere it appears. The number on the agent’s own page, the one a supervisor sees for that same person, the metric columns underneath, the coaching summary and the digest email were each worked out slightly differently — different rounding, a different set of records, a different way of combining the individual criteria. They now follow one rule, so a metric marked as a growth area always sits below the heading it belongs to, and a record with no rubric is shown as unscored instead of as a zero.
A supervisor and an agent see the same six figures. Opening a person’s performance from the team view and opening it as that person gave different answers for the same period — a different record set, a different rank denominator, no team markers on the supervisor’s view. Both views now read the same source.
The report delivered to your CRM shows script adherence. If your organization has a script, the analysis already scored how well it was followed — but the manager reading the deal in the CRM could not see it. The timeline entry now carries the score, the steps, and the required phrases that were missed.
The starter scripts a new organization receives now apply on their own. Four of the six are bound to the call categories they belong to; the other two are labelled as templates to adapt, both in the interface and in the documentation. Organizations created earlier receive the same bindings.
Having the AI read your script no longer costs you anything. Parsing a script in settings runs on our account, with a fair-use limit per organization, and is excluded from your cost reports.
Attachments in a chat are described in your language instead of as a service marker.
Highlights and notes on a transcript stay where you put them. A highlight and a selection now cover the same unit of text, overlapping highlights no longer print a marker twice, and a note is stored against the original wording — so it survives a re-analysis of the conversation.
“Total Transcripts” on the dashboard counts conversations. It counted rows, and a single call can have several — a cleaned-up version and a re-marked one both point back at the original. The breakdowns underneath the card now cover exactly the same set.
A record that refused to be re-analysed can be re-analysed. Where the analysis version had drifted between the two stores behind the product, the record is repaired automatically instead of reporting an error.
Telephony and Bitrix24 credentials are encrypted at rest. In 2.37 the keys for speech recognition, analysis and email delivery were sealed by the application itself. The same now applies to the credentials stored for your telephony provider and your Bitrix24 portal.
Connecting a source recovers from a bad configuration. A malformed setting no longer breaks the connection, a tab left open explains itself instead of acting on stale state, and disconnecting affects exactly the installation you chose.
Removing the Bitrix24 app finishes cleanly. An ordinary removal from the portal no longer reports an error, and the notification binding that tells us about it is restored where it had been lost.

August 2026

Chats tell your team from the customer, and a price covers exactly what runs · 2.37

Chat analysis now knows which side sent each message. Who said what in an Open Channel conversation was decided by comparing the sender with the activity’s responsible employee — so a chat answered by anyone else could come out with the roles swapped, or be judged as though your side had never replied. The side now comes from the conversation itself.
The price you are shown before starting pending records covers exactly the records that will run. A single run processes up to 200 records, and the card now says how many of the accumulated ones that price was calculated for; the confirmation step adds how many are left for the next run.
Recognition confidence on the dashboard reflects what was actually measured. A call that had been cleaned up or re-marked counted several times over, and “not measured” was shown as 0.0%.
What the AI parser produced is visible before you save it.
Provider API keys are encrypted at rest. The keys stored for speech recognition, AI analysis and email delivery are additionally sealed by the application itself, so reading the database alone does not yield a usable key. The privacy policy and the DPA state it.

Scripts show what they apply to, and your minutes have one answer · 2.36

Every script now shows its projects. The scripts page lists what each script applies to. A project where this script would never actually run — because the project is paused, or because another script is the one chosen there — is shown dimmed and moved to the end, and a script where every project is like that is marked Not assigned instead of looking like any other.
The AI script parser keeps your script a script. Speech-quality notes (“confident tone”, “no filler words”) and instructions written for the analyst are recognized as such, left out of the steps, and listed with a short reason before you save — instead of quietly becoming criteria that duplicate what call analysis already measures.
“How many minutes are left” is one number everywhere. The remaining included minutes on your subscription screen, the price quoted for pending records and for an import, and the limit analysis itself applies are now worked out the same way. Organizations that had moved between plans could previously see a figure calculated over a period that had already ended.
The pause notice tells you the current state. Once you top up, or when a new cycle renews your included minutes, the banner stops saying funds ran out. If analysis was switched off automatically, that — and when it happened — is now visible under Settings → Organisation, not only above the records list.
A record stopped for lack of funds says so. Previously such a record could come back as “not started yet”, hiding the actual reason.

Analysis becomes a decision you make, and an empty balance stops being a dead end · 2.35

Analysis now has two named modes instead of an on/off switch. In Automatic, a new record goes straight into processing, as it always has. In Manual, a record waits and spends nothing until you start it. Set the mode in the banner above the records list or under Settings → Organisation — in the portal and inside Bitrix24 alike.
Start what has accumulated: one record, a selection, or a batch of the backlog. The button is on the record card; the bulk start sits above the list. A single run takes up to 200 records and tells you how many are left for the next one. Before anything is charged you are shown what that run will cost, calculated record by record rather than from a total — and where a call’s length is not yet known, you are shown a “from” amount together with how many such calls there are, because their duration is only measured once the recording is transcribed.
An empty balance is now one honest pause instead of a wall. Analysis moves your organisation to manual mode once, naming the reason and the moment, rather than marking every record “no funds”. Records held back this way can now be started like any other — previously they could not be started at all. Free organisations are not paused: their monthly allowance renews on its own.
Topping up does not restart analysis by itself. What accumulated waits for you to start it, so a top-up never turns into an unexpected charge. The pause email, the notification and the FAQ all say this in the same words.
The billing screen answers “do I have minutes?” first. Remaining minutes are now their own card and read as of today, independent of the period you are browsing. The plan’s allowance shows as a fraction — “0 of 100” rather than a bare “0” — what your spending was covered by is listed under covered, and the banner says which pool the current analysis is drawing on.
Answered incoming calls reach Aelo sooner. Bitrix24 marks a call completed about a minute and a half after it stops reporting on it, and calls that arrived in that gap used to wait for the hourly catch-up. They are now picked up as soon as the call closes.
A metric nobody measured no longer shows as zero. On an agent’s dashboard, a metric that does not apply to the channel — speech clarity in chats, for instance — now shows a dash instead of “0 out of 100”. A genuine zero, in turn, now appears in growth areas and coaching, where it used to be filtered out along with the placeholders.
Photos and links sent in a chat now reach the transcript. A photo sent without a caption, and any message containing a formatted link, used to be dropped before analysis.
Calls from Binotel now reach Aelo. Their webhooks are sent from the provider’s own network ranges, which the connection’s IP allowlist now recognises.

A chat costs less, and telephony can supply audio your CRM lacks · 2.34

A chat now draws half an included minute instead of a whole one. Twice as many conversations fit into the same allowance, and the point where analysis runs into an exhausted plan comes later. The price of a paid chat is unchanged: a chat beyond the included minutes still costs the flat per-chat rate.
Telephony can now be connected as an audio source, from the wizard. Some CRMs register a call but never attach its recording. A telephony installation can now be switched into that role during setup: it creates no records of its own and instead supplies the recording the CRM record is missing. It appears under Settings → Data Sources → Connection as its own card, with its webhook URL and the prepared email for the provider’s support. Available today for Binotel alongside Bitrix24.
The billing screen says what your minutes were covered by. One card for the period shows how the minutes split between your plan’s grant, your bonus and overage, with the amount charged next to the balance it came from. The transaction history filters across the whole period rather than the page in front of you, and loads more on request instead of stopping silently.
A conversation with no recording now tells you why. Records whose audio never arrived used to say the recording had been received and stored. They now name the reason — a missed call, or a CRM that logged the call without a file.
A self-served webhook secret can be read again after setup. Previously it appeared once, while connecting, and could not be recovered afterwards.

The transcript goes where you asked it to · 2.32

The transcript can now be written into a CRM field you choose. The setting existed in field mapping, but delivery never filled the field. It works now — and it arrives switched off, because the field carries the full text of the conversation and that is your decision to make, together with the field it goes into. Nothing changes for Smart Process delivery: there the transcript already appears in the item’s timeline next to the analysis report.
The company name you enter during Bitrix24 setup is now the name of your organisation. It was asked, checked, and then dropped, leaving the portal domain in its place. If you have already renamed your organisation yourself, nothing is touched.
A chat with a transcript is analysed even when speakers cannot be told apart. Previously such a conversation could be closed as having nothing to analyse.
A phone-system error is no longer read as “there is no recording”. Those are different answers, and treating them alike left calls waiting a day for audio that had already been offered.
Shortening your retention period now applies to calls already on file, not only to those that arrive afterwards.
Reconnecting a phone system keeps it as a source of audio if that is the role you gave it. Running the wizard again used to reset it quietly, and the effect showed up a day later.

Recordings can now come from your phone system, not only your CRM · 2.31

When your CRM registers a call but never stores the audio, Aelo can take it from the phone system instead. Some telephony integrations write the call into Bitrix24 but keep the recording on their own side — the card exists, the file does not, and until now there was nothing to analyse. A telephony connection can now be added purely as a source of audio: it stops creating its own records and simply supplies the conversation for the calls your CRM already has. Aelo matches the two by the call identifier your CRM carries, and failing that by number, time and length. Binotel is the first provider supported in this role.
Processing your accumulated conversations now works from one button. Calls that arrived before analysis was switched on can be sent through in a batch, including those whose recording had not yet reached us. Before starting them, Aelo pulls your operator’s call log for the same period, so older calls can still be matched to their audio.
A conversation that will not be analysed says so, and says why. Instead of a record that silently stays empty, you see whether the recording is confirmed absent at the provider, whether several calls matched and none could be chosen, or whether the audio never arrived.
Chats now use your plan’s included minutes. A chat spends one minute of the monthly grant, and your wallet is charged only for what the grant does not cover. You can also set the minimum number of messages worth analysing — shorter exchanges are neither analysed nor billed.
Every conversation now shows both times, each labelled. When the conversation happened, and when it reached Aelo — so the gap between them is readable at a glance instead of guessed.

You can now subscribe to the newsletter, not only unsubscribe. Until now the only place to say yes was a checkbox during web sign-up — which nobody who arrives from a Bitrix24 portal ever sees. It is now offered in three places: at sign-up, once when a portal account is linked to a web account, and in Settings › Notifications, where the account owner or an administrator can turn it on and off at any time. It is separate from the alerts and digests about your own account, which keep arriving either way.
Fonts are now served by Aelo itself. The interface, the website and the PDF report used to load their typefaces from Google’s font service, which meant your browser opened a connection to Google directly — before any consent banner had a say — and handed over your IP address and browser details. The fonts now ship with the product, so that connection no longer happens at all.
The sub-processor registry describes telephony exactly as it works. It previously covered only recordings submitted through the public API. It now also spells out the two other ways a recording reaches us: providers Aelo polls using the credentials you configure, and providers that simply send us a link to the recording.

Calls arrive in minutes, and a portal account is yours to keep · 2.29

A call that arrives without a webhook is picked up within minutes. When your CRM does not notify us — which, measured on live traffic, is how a third of Bitrix24 activities actually reach Aelo — the reconciliation pass used to run twice an hour, so those calls could sit for up to an hour before analysis even began. It now runs every five minutes, and the safety window before an activity is treated as final is two minutes instead of ten.
Signing in to the web app from a Bitrix24 portal keeps you in your own organization. Bitrix24 does not share a user’s e-mail address with an embedded application, so opening the full web version used to create a new, empty organization around you. There is now a link from the portal that binds the account you sign in with to the profile you already have.
A source whose authorization has expired is shown as disconnected, with the reason. It used to keep appearing as a working connection, so a portal that had stopped delivering calls looked healthy.
Your team list matches employees by their portal identity. Under Bitrix24’s minimal permissions the e-mail address is not shared, so matching by address found nobody even when the connections were working correctly.
Withdrawing consent for marketing e-mail now unsubscribes you. Previously the withdrawal was recorded, but delivery stopped only through the unsubscribe link in the message itself.

What the analysis knows, and what it only guessed · 2.28

An analysis that fails its automatic check now says so. The report e-mail carries a note, and the delivery webhook a verification block with the status and the reasons. CRM automation was already held back in that case; now the other channels tell you the same thing instead of delivering the report as if nothing happened.
The data retention window is editable by the organization owner. 1 to 365 days, with a preview of how many records a shorter window would remove and an explicit confirmation before it does.
A phone number that arrives in a contact’s name is masked like a phone number. CRMs create a contact from an inbound call with the number itself in the name field; those now show the last three digits in e-mail and personal webhooks, as everywhere else.
A disconnected CRM can be reconnected, and you are warned first. If your organization has two sources connected, both stay visible after a disconnection and each comes back on its own button. The warning that the portal tile goes dark now appears in the confirmation dialog, before the disconnection rather than after it.
A chat that failed before it was analyzed can be re-analyzed. The button now replays from the recording itself rather than expecting an analysis that was never saved.
Voice metrics say “not enough data” instead of showing zeros. When pitch or pause detection degrades on a recording, the card and the PDF report both say so, rather than presenting the zeroed values as a measurement.

Three languages instead of one, chats on equal footing with calls, and every analysis double-checked · 2.27

Report language and recognition language are now two separate settings — and both default to Auto. The report is written in the language of the conversation, and the recognizer detects the spoken language on each recording, so a project no longer inherits a language from the screen you happened to be using. Set a fixed report language per project from the Analysis Setup hub when you need one; pick the recognition language from 90+ supported languages with search. Quotes from the conversation are never translated.
Chats are analyzed on the same footing as calls. A chat used to land in the catch-all category with no scores and no alerts. It is now categorized into the same ten categories as a call — from the whole conversation, not the first few messages — and scored with chat-appropriate criteria, so analysis profiles, script rules, scores and alerts apply to chats too.
Your own list of forbidden words. Each project can name the words an agent must not say; every hit shows who said it, in what form, with the quote — and there is an alert rule for it. The metric formerly labelled “Stop words” is now called Agent uncertainty, which is what it always measured: hedging about price, deadlines and promises.
Every analysis is independently checked before it reaches you. Statements about safety, agreements and CRM fields must be backed by the transcript; anything that cannot be is withheld and the record is marked for review, and CRM automation acts only on a verified analysis. Organizations that want a human in the loop can ask us to enable the Review Queue (beta), where withheld actions are approved or rejected. Re-analysis no longer overwrites the previous version, so history is kept.
Metrics stopped grading what they cannot measure. CRM completeness stays silent when there is no CRM to compare against; qualification no longer expects a private buyer to name a budget and a decision-maker; a promising sale is no longer an “attention” reason next to threats and data leaks; and voice metrics — pitch variability, speaking ratio, pauses — are real measurements again, kept on re-analysis. A zero result is shown as zero, not hidden.
Free plan: 100 analyzed minutes a month instead of 30 — for every free organization at once, no action needed.
Data retention is yours to set — from 1 to 365 days. Shorten the window and Aelo shows how many records that would remove before it changes anything.
Notification settings in one place. Rules, recipients and channels are now one section, in the order they apply, with a summary of what you actually receive today; digests can be switched off for the whole organization. Save works the same way on every settings screen — bottom left — and leaving with unsaved edits asks first, including when you switch organization.
Privacy in alerts. Telegram alerts — team channels and personal ones alike — no longer include the other party’s name, number or quotes: only the rule, the severity, the category and a link to Aelo. Where a contact is still shown, in the app, in e-mail and in webhooks, it is masked to a first name and a surname initial, or to the last digits of the number. Marketing e-mail is sent only to people who separately opted in.
CRM connections you can switch off — and back on. Disconnecting a Bitrix24 or Uspacy installation now really stops the flow, and a Reconnect button brings it back without reinstalling. Signing in from Bitrix24 no longer needs an e-mail address: you are recognized by your CRM identity.
The Analysis Setup hub explains each step — what it decides, what happens if you skip it, and where the result shows up — and no longer scrolls the editor out of view. Renamed for clarity: “Script” is now Script rules, “prompt profile” is analysis profile. The “Competitions” page has been removed, and the team quality table lists people without ranks or medals.
Smaller fixes you may notice: the projects list filters correctly and “All” includes paused projects; the quality figure on project cards is a 30-day window; “Open in CRM” leads to the record’s own portal; call statistics respect the selected project; a dashboard alert link opens that alert; counts and rule-builder labels read naturally in every language.

Compliance settings that fit your jurisdiction, and Aelo speaks your language from the first screen · 2.26

Set up recording compliance the way your organization actually works. The approach and the legal basis are no longer limited to consent and to GDPR article numbers: if your organization does not rely on the consent of call participants, choose Other lawful basis and then select the basis that applies — including Other basis under applicable law for one defined outside the listed categories. Every option now explains in a line what it means in practice.
The compliance screen is available in your language. A language switcher sits on the screen itself, so the declaration you sign is one you can read. Nothing is preselected for you either — the approach, the jurisdiction and the legal basis are yours to state, and the form waits until you have.
Installing from Bitrix24 sets up Aelo in your portal’s language. The organization, its project and the sample data on the first dashboard now all follow the language of the portal you installed from, instead of starting in English and waiting to be corrected by hand.
Times and durations read naturally in all four languages. Minutes, hours and seconds are now written the way each language writes them — “1 хв”, “8 год 30 хв”, “hace 5 min” — across dashboards, widgets and record cards.

Filter by several projects at once, and a paused project stops taking calls · 2.25

Dashboard filters accept more than one project now. Select two or more and every number reflects exactly that set — previously only the first project you picked was applied, silently narrowing the results to one.
The selection carries through when you drill into the records list, so the records you land on match the projects you just filtered by, not just one of them.
Archived projects no longer show up in the project filter. Only active and paused projects are listed, so closed-out projects don’t clutter the dropdown.
A paused project stops accepting new records everywhere — CRM sync, the public API, and file upload now all decline new calls and chats for a paused project, the same way an archived project already did. Records already in the project are unaffected.

The report you send onward says everything the panel does · 2.24

The printed report now carries every section of the analysis panel. Agreements reached, voice metrics, script adherence, business intelligence, the analysis profile and the deal’s stage all travel with the document — so a report forwarded to a manager or handed to a client shows the same picture you see on screen, not a shorter one.
It reads as a document, not a screenshot. The report is laid out in Aelo’s own paper design, the one this site and our emails use, so it prints cleanly and looks like something worth attaching to an email.
Headings and metric names are in your language, in all four we support.
A check that did not run says so. “Passed”, “failed” and “not applicable” are now three different answers instead of two, so a blank line no longer has to be guessed at.

Analysis starts when you decide · 2.23

A new organization no longer starts analysing on its own. Calls flow in and appear in the list from the first minute, so you can see the integration is working, but nothing is transcribed and no minutes are spent until you press Start analysis. You get the time to set up projects, criteria and prompts first.
What arrived while you were setting up is offered separately, with the price shown up front — how many conversations, how many minutes, and how much of it goes beyond your plan. You decide whether to process all of it or only today’s.
An empty balance now reads as “top up”, not as an error. Conversations wait in an “Awaiting top-up” state and resume once the balance is topped up, instead of appearing as failed.
Compliance is confirmed right after installation, before the setup questions — so calls stop being turned away during those first minutes.

Aelo comes to Uspacy, and your org chart keeps everyone · 2.20

Aelo now works inside Uspacy. Install it from the Uspacy Marketplace and your calls flow into Aelo the same way they do from Bitrix24 — analysed, scored and delivered back to the CRM card. Bitrix24 is no longer the only way in.
Importing your department tree keeps every employee, not only the ones already using Aelo. People without an account are listed as pending and become full members automatically once they sign in — you no longer have to re-import after onboarding your team.
Large org charts stay workable. Filter by name, collapse branches, or switch to a “problems only” view to see just the departments that need attention.
Set your compliance settings without leaving the CRM. The organization owner chooses consent mode, jurisdiction and legal basis right inside the embedded app.

July 2026

Department heads see their team’s calls · 2.19

Import your company’s department tree from Bitrix24 under Settings → Team → Org Structure. Aelo matches employees to their Aelo accounts by email address and shows you by name anyone it could not match.
A department head sees the calls of everyone below them in the tree — their own department and every department under it — without needing a project assignment. This works alongside project supervision, not instead of it: whoever oversees a project keeps seeing that project.
Assign a head by hand for any department, including ones where Bitrix24 names nobody. A manual assignment takes precedence over the portal and survives later imports; when the portal names someone else, the row says so.
The same boundary applies everywhere — the call list, the record card, transcripts, analyses and search all agree on what a person may open.

Supervision, staffing and alerts that follow your projects · 2.18

Projects are staffed from either side. Assign projects to a person under Settings → Team, or open a project’s own Team tab and choose who supervises it. Inviting a supervisor now asks which projects they will oversee, and the team list shows at a glance who is waiting for an assignment.
A supervisor sees the calls in the projects they oversee — transcripts, recordings and analyses — and the alerts and daily digest cover the same projects, so what they are told matches what they are responsible for. (Since 2.19 a supervisor can also see calls through the department tree — see above.)
Agents reach their own calls directly, without needing a project assignment, and only their own.
Rating an analysis is open to everyone who can see it, agents included — the quickest way to tell us where the AI read a conversation wrong.
Alert rules can require a specific phrase and report which mandatory phrases a conversation missed, so a script commitment becomes something you can measure.
Alerts reach your team chat, with sensible per-role defaults out of the box and one-click Telegram — no setup needed before the first alert arrives.
A rule tells you on its card when its metric is no longer available, so a rule that has stopped matching is visible instead of quiet.
The support chat knows who it is talking to inside the Bitrix24 app, so you can ask a question without introducing yourself first.

Telephony records carry the numbers behind the call · 2.17

Call direction is read from the call’s own numbers. Phone systems vary in how reliably they state direction, so Aelo determines it from the record itself — which keeps the client’s number and the agent’s on the right side of every call.
An inbound call says which line it came in on, so you can see which of your numbers the customer dialled.
The customer’s number is always on the card, even when no contact name is known — on an inbound call it is the one fact that identifies the caller.
A number where there is no name. Some phone systems do not attach an operator to a call at all. Aelo shows the extension that handled it, so calls from different seats are told apart even when no names are available. This is separate from the line: that is the public number the call went through, this is the seat that answered it.

Alerts now show their evidence — and let you act on it · 2.15

Every priority alert carries the line that triggered it. The card shows the quote from the conversation and where it was said, with the most serious alerts first — an older critical one stays above a newer minor one. You can tell at a glance whether a card deserves your morning, instead of opening five records to find out.
Opening a record from an alert plays the relevant moment. Instead of starting at the beginning, playback jumps to a twenty-second window around the quote, with one control to keep listening to the rest. It works the same inside Bitrix24.
Two actions moved into the alert itself. When someone promised to call a client back, the alert can create a Bitrix24 task or a planned call without leaving the page — and repeating the request will not create a second one. When a client’s tone dropped during a conversation, the alert can schedule a coaching session for that agent, tied to that recording.
Aelo can check whether promised callbacks were actually logged. If a conversation contains a commitment to call back and there is no matching open call in the CRM, you get an alert. This is off until you switch it on for a project — turning it on also enables extracting agreements from conversations, and the app says so before you confirm.
Client accounts see the same team picture as supervisors. Team overview, quality trends, calls to review and coaching now live on the client dashboard too, computed from the same data rather than a parallel version of it.
Recordings show which line the call came in on, for telephony that distinguishes several numbers.
Billing follows the conversation. The balance check and the charge are based on how much was actually spoken.

Personal data leaves your transcripts before they are saved · 2.14

For conversations processed through the new write paths, phone numbers, emails, payment cards and valid IBANs are removed from a transcript on write. For those paths, only the redacted text is kept; this was not retroactive and original audio remains unchanged. What gets removed is confirmed rather than guessed: card numbers by their checksum, IBANs by their country’s format, phone numbers by whether the number exists in a real numbering plan. Business figures stay intact — a sum, a date, an order number is not a phone number.
A new page explains exactly what happens to your data. Personal Data in Transcripts lists what is removed, what is kept, and — just as plainly — what the redaction cannot catch: your call recordings are stored unchanged, records created before this release are not rewritten, and a number dictated in words is not detectable.
CRM auto-fill note. Fields whose value is itself a phone number or an email can no longer be filled from what was said in the call. The contact’s own number is unaffected — it comes from the record, not from the transcript.

Cancel or downgrade with a clear conversation, and find integrations faster · 2.13

Canceling or downgrading now asks why, and confirms the change. When you start a cancellation, Aelo asks what prompted it and offers a fitting option before anything takes effect; downgrading a plan asks you to confirm first, so neither happens on an accidental click.
Connecting a data source is easier to navigate. Integrations are grouped by type — telephony and CRM — in a clean two-column layout, so the one you need is easier to spot as the list grows.
The business overview shows what you have left. A new card puts your remaining balance and included minutes alongside the existing metrics.
Your Bitrix24 portal language carries into Aelo. The interface language set in the portal now follows through to the app and your project.

Plans and billing, stated plainly · 2.12

What separates the plans is included minutes, the per-minute rate beyond those minutes, and access to the public API. Everything else — Bitrix24 and smart process, CRM import, routing rules, annotations and reviews, direct telephony, email digests — is part of every plan, Free included.
Changing plan shows what you will be charged. The confirmation names the amount that comes off your card at that moment, calculated from the same proration as the charge itself.
Your balance is always visible — including on plans where we credit the account under an agreement rather than you topping it up by card.
Emails from Aelo keep their layout in every major mail client, Outlook included.

The API is easy to find, and plan cards say what you get · 2.11

The public API now appears where you connect your calls. Adding a data source shows it as a fourth option alongside Bitrix24 and the telephony providers, and takes you to Settings → API Keys, where the plan requirement is explained in place. That section is visible on every plan now — including Free — so you can see what the API does before deciding to upgrade.
Plan cards read the same source that bills you. Each plan shows its price, included minutes and the rate beyond them, and — for paid plans — what becomes available for the first time at that level, so the card cannot drift from your invoice.
Unitalk joins the source settings, and Ringostat is listed with the cloud PBX providers rather than call analytics.
The data-sources guide covers connecting via the API, in all four languages.

Upload calls to Aelo from your own systems · 2.10

A public upload API. You can now send calls and chats into Aelo straight from your own code — no Bitrix24 app required. Create a secret key under Settings → API Keys, then post recordings to the API and they flow into the same analysis pipeline as everything else. Available on Starter and above.
Built for integrations. Each key is scoped to what it may do, requests are rate-limited and idempotent, and the API tells you up front — before anything is stored — when a record can’t be processed on your current balance. Full details are in the API reference and upload guide.

A faster start, with everything included · 2.9

Installing from Bitrix24 takes seconds. Demo data is written in bulk, so you land in a ready dashboard almost immediately.
Every capability is on from day one. Alert rules, call scripts and data export are now included on every plan, and upcoming features appear as previews you can explore — no hidden switches to hunt for.
The app speaks your portal’s language. Installing from Bitrix24 picks up the language of your portal profile everywhere: the interface, organization emails and the default analysis language of your project.
A supervisor view for owners. The dashboard gains Quality and Agents tabs, so owners get the same per-agent quality breakdown supervisors use.
Progress at a glance. The agent dashboard shows the score trend as daily bars, and agent names in “Needs Attention” and “Top Improvers” now link straight to that agent’s dashboard.
Budget got its own tab in Plan & Billing, so spending controls no longer hide behind the plan details.

Analysis setup in one place, scores in the records list · 2.8

A guided “Analysis Setup” hub. Everything that shapes how calls are analyzed — evaluation criteria, scripts, category assignment, project overrides and delivery to the CRM — now lives in one place, laid out as a pipeline with a live readiness status on every step. The new “Configuration” section in the sidebar takes you straight there.
One records list instead of two. The “Analyses” page merged into “Records”: analysis quality and sentiment are now columns in the records table, with filters to match (including a QA-score range) and column visibility that remembers your choice. Old links redirect automatically.
A tidier records list. In the narrow layout the deal outcome shows as a colored dot with a tooltip, long call series group sensibly in the depth funnel (1–5, 6–10, 11–50, 51+), and short-call detection respects each project’s own threshold.

Steadier categorization · 2.7.5

Analysis runs on our AI provider’s newer recommended interface, which measured as more reliable end to end. Your transcripts are still not retained by the provider.

Sharper voice metrics and interface polish · 2.7.1

Speaking pace measured against the transcript. Pace is counted from the words actually said rather than estimated from the audio signal alone, so the words-per-minute figure and its slow/normal/fast verdict describe the conversation. When the measurement cannot be made reliably, Aelo says so instead of showing a confident number.
Polish across the interface. Empty states name the next step, speaker colours follow the speaker’s role, and Ukrainian, Russian and Spanish wording is more consistent throughout.

Analysis that fits the call · 2.7

Different scripts for different calls. Bind a script to a call category and pin the analysis config a project uses, so a support call is no longer graded against a sales script.
A fairer quality score. The score now averages six skills instead of seven. CRM completeness left the rating: it applies to a minority of calls, so averaging it in penalized agents for data collection that was never their job on that call. It stays as its own compliance indicator. Expect scores to shift slightly upward for teams that were carrying that penalty.
CRM completeness knows when it does not apply. If the customer is already in your CRM — a linked deal, contact or company — the metric reports “not applicable” and drops out of the rating, instead of scoring low for a task the call never had.
Scorecards that follow the deal stage (beta). Aelo can read a deal’s stage from Bitrix24 and grade the call against the scorecard bound to that stage, so a first call on a fresh lead and a contract-signing call are no longer judged by the same criteria. Opt-in per organization — ask us to switch it on.
Deal outcomes on the call. Calls now carry the outcome of the deal they belong to and how that deal moved between stages, with a conversion funnel on the Conversion tab.
Dashboards that explain their numbers. Quality trends mark the day you changed the AI configuration, KPI tooltips say where each number comes from, and supervisors get “objections the team drops”, “skipped script steps”, and a real compliance pass rate.
A coaching plan that writes itself. Each agent’s plan is built from the improvements their own calls suggested.
How it sounded, not just what was said. The analysis panel now shows pace, pauses and loudness measured from the call audio.
Every quote comes from the transcript, checked against it word for word, and categorization moved to a newer model picked by a head-to-head evaluation.
Tell the AI about your business. Organizations and projects can describe what they do and what they want out of analysis; that context goes into every call we analyze.
Recover calls that lost their recording. Admins can spot calls whose audio never arrived and re-fetch it from the CRM in batches.

Dashboards you can trust · 2.6

Real trends on every metric. Every KPI on every role dashboard shows its change against the previous period, including per-agent trends in the leaderboard.
Conversion funnel for call series. The owner dashboard shows how call series progress stage by stage, filterable by project, plus a breakdown of calls by direction, project, and hour of day.
Call category mix everywhere. Every dashboard now includes a breakdown of analyzed calls by category — for the whole team, per organization, and personally for each agent.
A real toolkit for supervisors. Hourly load and per-agent workload charts, a “calls to review” queue that surfaces flagged or low-scoring calls with one-click drill-down, script adherence by agent, and a working coaching tab.
Agent growth view. Agents can jump straight from a weak focus area to the exact calls that lowered it, and track their weekly streak and personal bests across their full history.
Owner finances at a glance. A revenue chart, cost breakdown by type, and balance with estimated runway — all computed from your actual usage charges.
Operations for admins. Integration health at a glance, processing errors grouped by source, and one-click retry right from the errors list.
Clear empty states. When a period simply has no data, dashboards say so in words.
Dashboards speak your language. Relative times, chart labels, categories, and coaching insights are now fully localized in all interface languages.

Sharper analysis & guided Bitrix24 onboarding · 2.3

More accurate call categorization. Internal and personal conversations are told apart from customer ones, and a call that mixes small talk with a real business request — prices, availability, an order — is routed to the customer category it belongs to.
“Other” is a described call, not a scored one. Personal and internal calls get a summary, topic, sentiment and key moments, and the analysis card explains why a sales score would mean nothing for them.
Call category at a glance. The record card header now shows the call’s category tag — the same one as in the records list — so you always know what kind of call you’re looking at without going back to the list.
Smarter re-analysis. Re-running an analysis now defaults to “Auto — category profile”: Aelo picks the evaluation profile bound to the call’s category, exactly like the normal pipeline. Selecting a specific profile remains available as an explicit override when you want particular criteria.
Speaker-level flags rest on evidence. Profanity and attention flags name a speaker only when the transcript shows it, and critical moments — churn risk, competitor mentions, sharp sentiment drops — are flagged more reliably.
Guided start inside Bitrix24. New embedded users get a welcome narrative with sample data for their industry, and a “Full portal” button opens the complete Aelo workspace in one click — no separate sign-in.
Category filters on the analysis card make it easier to jump to the score, script, speech, or insights sections.
Fuller owner dashboard. Organization owners now see the complete administrative picture — team leaderboard, sources, alerts, and delivery management included.

June 2026

Deeper Bitrix24 integration · 2.1

Smarter follow-up routing. Commitments that Aelo detects in a call now turn into the right Bitrix24 action automatically — an open promise becomes a record comment, while a “call you back” becomes a scheduled call activity. Each one carries a concrete due date and links back to the parent deal.
Call-chain tracking. Calls tied to the same CRM entity are grouped into a series, with each call numbered in sequence — so you can see how a deal progressed conversation by conversation.
Analytics into Smart Processes. Every analyzed call writes its full metric set (quality score, sale probability, sentiment, component scores, and more) into a dedicated Bitrix24 Smart Process item, kept up to date automatically.
Auto-fill empty fields. Delivery can populate blank CRM fields from the analysis without overwriting values your team already entered.
Demo analyses are clearly marked. Analyses generated from sample data now show an explicit notice, with a nudge to connect a real source.
Clearer short and missed calls. Very short calls (under the ingest minimum) and missed calls are now labelled everywhere, so it’s obvious why they don’t carry a quality score.

Prepaid billing, compliance, and a new look · 2.0

Transparent prepaid billing. Aelo moved to a prepaid model with live Stripe payments and minutes-based pricing. Top up your balance, see what each analysis costs, and track spend from the dashboard.
EU AI Act and GDPR documentation pass. A full review of how data is handled, with refreshed Privacy Policy, Terms, DPA, and sub-processor documentation.
Redesigned website and pricing, plus a platform-wide pass on usability, accessibility, and translations.

Category-aware analysis · 1.18

Category-bound prompt profiles. Analysis now adapts to the type of conversation (sales, support, and more), so each call is evaluated against the criteria that actually matter for it.

More telephony, email digests · 1.17

New telephony connectors: Unitalk, Binotel, and StreamTelcom join the supported integrations.
Email digest reports — recurring summaries delivered straight to your inbox.
Faster, more reliable transcription with an upgraded speech-to-text pipeline.

May 2026

Alerts you can act on in bulk · 1.16

Bulk-resolve for alerts — clear all matching alerts in one action, with an accurate live count shown in the sidebar.

Aelo arrives · 1.13–1.15

The platform is now Aelo (rebranded from AICA), with a refreshed design system across the app and website.
Demo Data v2 — explore the full product with realistic sample calls before connecting your own.
Public documentation site — the guides you’re reading now.

See where points are won and lost · 1.12

Component-level scores in the Analyses grid — drill into the individual criteria behind each conversation’s overall score.

April 2026

Coaching feedback and faster screens · 1.9–1.11

Agent rating & feedback — rate analyses and leave coaching comments, with a dedicated stars column in the records view.
Filter bars on Analyses and Alerts, plus a faster, smoother dashboard.

March 2026

Billing, search, and guided onboarding · 1.4

Stripe billing for subscriptions and usage.
Semantic search across your conversations — find calls by meaning, not just keywords.
In-app product tours that walk new users through each area.
Consent gate to support compliant recording.

Bitrix24 embedded mode · 1.2–1.3

Embedded Bitrix24 app — run Aelo directly inside your CRM, with controls over how results are delivered back to deals.
Guided onboarding and demo data for a faster first-run experience.
GDPR groundwork across data handling and consent.

1.0 — General Availability · 1.0–1.1

Aelo reaches 1.0: a redesigned interface, public website, documentation, and feature flags for controlled rollouts.
Category-aware analysis and lower AI costs.

January – March 2026 · the foundation

The early releases established the core platform:

AI Script Parser — turn your sales script into automatic adherence checks.
Call categorization powered by multiple AI providers.
Embedded mode and a rebuilt, role-based permissions model.
Multi-language interface with full localization.
A major performance and UX overhaul of the analytics experience.

Looking for the technical, version-by-version history? That lives in the project’s internal changelog. This page is the curated, customer-facing view.