Every one of the 51 updates shipped on the AI side, newest first.
After mentioning a record you can simply continue with "and their invoices?". The chat keeps track of the customer, company, invoice or conversation you named, and carries the date range, sort order and filters into the next question. Three separate topics can run in the same chat.
For questions that want a number, the chat returns a card rather than a paragraph: the figure itself, what was counted, which date range and which filters applied. Asking the same question twice gives the same answer, because the number comes from the query, not from the model.
Define once what "hot lead" or "large customer" means in your company, and the chat uses that definition. Who may change a definition is controlled by permissions, so the meaning of a number doesn't vary from person to person.
When you reopen a past chat, the figure in it doesn't just sit there frozen: the answer states when it was produced and you can re-run it against today's data in one click. You no longer carry a two-week-old number into today's meeting believing it's current. Long record lists now page with a seek cursor, so an answer with thousands of rows isn't cut off at the first page.
Ask the chat about the relationships around a customer or a company: linked companies, related contacts, shared records. Related records you name are listed directly, and only the records you are allowed to see appear.
A question you ask regularly can be put on a schedule: the answer is waiting for you every Monday morning. The scope and recipients are configurable, and past answers are kept, so you can see what changed since last week.
When the chat puts a record first, it says why: which signals, with what weight. If several signals contributed, all of them are listed — instead of "the system said so", you get a reason you can read.
Scoring keywords can have synonyms, edited in bulk — "price", "cost" and "fee" no longer have to be entered separately. Before changing a weight you can save a what-if scenario and test its effect. Each score states how fresh it is, and if model quality drifts over time you are warned.
You can see the effect of an assignment rule before you turn it on: how many records would land on each user. A user capacity calendar carries who is away and who is full into the rules, and workload balance across the team is reported — records piling onto one person don't stay invisible.
Around thirty action forms — company, contact, note, appointment, task, inbox message and more — now open inside the chat itself. Nothing is created until you fill the form in and confirm, and every form says so on its header: “0 credits · nothing is created before you confirm.” You can record what you are looking at without leaving the conversation for another screen.
The surface of questions you can ask for free has grown: invoice breakdowns by company and by person, open-invoice summaries, invoice line insights, imported invoice drafts, current accounts, promissory notes, stock movement type and invoice type (with mixed-currency totals), status and type dictionaries, and the KVKK destruction log. “Overdue” is now treated as a predicate rather than a column, so it is answered correctly regardless of where it is recorded. When two customers share a name, the assistant asks which one you meant; and it now gives both the count and the total of a company's invoices in one sentence.
The catalogue now suggests not only “how many” but “which way”: a family of trend cards was added. Two-dimensional cross-breakdown cards arrived and behave correctly in all three languages. A second filter can be added to a question, operational cards state the exact date they used, and operational alert thresholds are now yours to manage.
Scoring categories specific to your industry — real estate, healthcare, finance — are now actually applied in every analysis. The setting existed, but analyses were always scored against fixed generic categories. The same goes for your mandatory compliance statements (a data-protection notice, say): they are now genuinely checked, and results are shown in your own wording. Required and optional items are also distinguished. Coaching heat maps and trends deliberately stay on the five core categories so comparisons with past analyses still hold.
Revenue, ciro and выручка can be queried directly — "revenue this month", "revenue vs. last year", "revenue of lost opportunities". Time-bucket phrasing now works in all three languages: "by month leads", "aylara göre lead", "помесячно" return a real breakdown instead of a flat total. Three new cross-table views were added: calls by agent, overdue instalments by project, late deliveries by supplier.
Questions that resolve to a single number — "how many leads this month?" — now come back with the comparison attached: "Previous period 120 (↑19.2%)". No second question needed. Month-over-month and year-over-year comparison cards for leads, opportunities, revenue, customers, quotes, tasks, appointments, calls and sales invoices are now discoverable in the catalogue.
Every operation running in your project (plan, patch, staging release) now appears in an activity list, and you can open any job to inspect its steps. A running job can be cancelled with "Stop" and you can watch it wind down — cancelling never touches your source, preview or staging. Build, test and preview runs state their credit cost up front, and nothing starts before credit is reserved.
Start a project from scratch in your own isolated workspace, or copy one of the verified starter templates. Each starter comes with its own setup checklist, and the plan that summarises your goal and page structure is now saved to the project instead of disappearing on refresh. A separate plan provider suggests layouts — review-only: the model cannot write code, grant permissions or publish. You decide.
The chart palette has been reselected and a pattern channel added alongside colour, so two series are never distinguished by hue alone. You can drill from a chart into the detail behind it and export what you see — the export carries the filters it was taken with.
Summaries and cards generated from the record you are viewing no longer consume credits. Questions can now follow a relationship in two steps (from a customer to their opportunities, for example), search inside the body text of notes and descriptions, and search call transcripts straight from the chat. The assistant also tells you which actions depend on your plan or an add-on.
For supported CRM and reporting questions, you can see a visual summary alongside the answer. Drill into the relevant record or report breakdown; answers are generated only from data and permissions available to you.
Topics the AI could not answer now turn into a draft: an FAQ suggestion filtered by subject, stripped of personal data and free of links. You can also see which conversations a single answer affects — every chat that draws on the same piece of knowledge, in one list.
A coaching profile you share can now be protected with a password and capped to a number of views. You can also see whether the link was opened and when. On top of that, items in an improvement plan turn into real tasks with one click.
A lead who called five times and hung up angry every time no longer scores the same as one who spoke warmly. Scoring takes into account the customer's sentiment in analysed calls and the objections they raised. Organisations that do not use call analysis are unaffected — nothing is deducted, the signal is simply not counted.
The copy button on an opportunity card gives you a link that opens the same card for whoever receives it. For cards that are done or dismissed the view is carried in the link too, otherwise the other person would land on an empty list. And when a run could not be measured, the verdict card now takes you to the measurement history with your period and engine filters intact.
You can export the coaching comparison as a CSV that respects your permission boundaries. On the Voice of Customer side, sentiment breakdowns are now clickable: pick a sentiment and drill into the conversations behind it.
Review your AI agents' real conversations from one screen, rate them and mark them for training. Close the loop with review progress, pending review and average rating; move fast even at high volume with search, filters and saved views.
A new module that measures your brand's visibility in generative-AI answers. Define the set of prompts to track, watch each engine's answer and the sources it cites; whether you were mentioned, with what wording and from where — all on one dashboard.
Incoming conversations route to the most suitable AI agent, and a handoff map finally answers “why did it go to this agent?”. With load balancing, threshold simulation and handoff to a human when needed — your rules, a traceable decision.
Search the full transcript across every meeting, reassign speakers and jump to the right moment from sentiment and critical-moment timelines. Follow live recording progress, export the result as Markdown and import calendar events from an .ics file.
Beyond an assistant you ask one question at a time: agents that carry out work in the background. Define a playbook and the agent runs the steps in order — if the previous step succeeded do this, if it failed do that. Pause an agent without deleting it, see when and how recently it last ran, and stop a running job mid-flight.
When an agent proposes an action — a drafted email, a record update — you can now change the details before approving. No more rejecting the whole suggestion and starting over just because one field is off; fix it and approve.
A Resolution Analysis tab shows which conversations your agents closed and how. See the breakdown of outcomes by status; the queue card now surfaces average and maximum wait time plus timeouts. Where your agents get stuck is finally measurable.
“How long is my sales cycle?”, “How long does lead-to-customer take?”, “What’s our first response time?”, “Who are my 10 most valuable customers?” — customer lifetime value (LTV), conversion times and first response time can now be asked directly. Chat also guides you through multi-step analysis and doesn’t just hand you a number: it suggests what to do next.
You can now upload an Excel, CSV, PDF or Word file to AI Chat and analyze it right in the conversation. Summarize your tables, ask questions and get insights — without leaving the chat.
Lists and tables produced in the chat can now be downloaded as a real Excel or CSV file with one click. No copy-paste — the report is ready to use.
You can edit a message and regenerate the reply, get formatted answers (headings, lists, code blocks) and copy code blocks with one click. Plus, many combination and free-text questions are now credit-free.
Ask AI Chat about a customer and see invoices, collections and tasks at a glance. On top of that, anomalies in your data surface automatically as “insight cards” in the chat — brought to your attention before you even ask.
Copilot now covers every action, can build nested conditions/branches (then/else), and reveals the steps it creates on the canvas one by one. Changed nodes are highlighted live.
The AI hub, Rocketly Apps and Automation now live under a single “AI & Automation” section. Fewer clicks to reach the right tool, less getting lost.
Describe what you want in plain language and Copilot builds the workflow on the canvas for you: “When a new lead arrives, send a WhatsApp message, then assign it to sales.” Follow-up messages don’t start over — they edit the EXISTING flow (“wait 5 minutes in between” → a Delay step is inserted) and changed steps are highlighted.
Ask the AI assistant directly about your expenses, invoices, collections and stock levels. It can also take actions like drafting an expense/invoice or adjusting stock — with your approval (HITL). The assistant only sees the modules you have access to.
A 'typing' indicator while the agent replies, per-message credit transparency, a clear 'out of credits' notice and a 'Buy credits' button are here. Open an agent's details from its header in one click; scheduled summaries arrive as a repliable message in the agent's DM.
Ask the AI for the most sensible next step on a lead with one click. A credit is charged only on the first calculation; the same lead isn't re-charged again within a short cooldown.
All your conversations turn into a single picture: sentiment breakdown, the most-raised needs, critical moments and trends over time. Click any theme to drill into the real conversations behind it.
Ask the assistant in plain language and it no longer just answers — it acts. Create an opportunity, appointment, quote, task or call note, update a customer's status, or assign an owner, all from chat. Now on mobile too: streaming replies, inline charts, saved conversation history and message rating.
Connect your own OpenAI-compatible endpoint (Ollama, vLLM and similar). On provider outages the system gracefully falls back, so AI features keep working.
Pick an agent from the template gallery or build your own. Give it a knowledge base, schedule automatic (cron) runs, assign it tasks, or summon it with @ in team chat to run instantly. Agents remember preferences, distill lasting insights from their runs (Beta), and consumption is always shown transparently in credits.
For your upcoming meeting, a brief is prepared automatically on your Workspace home: a summary pulled from the contact’s email, WhatsApp and website conversation history — instantly, with no AI needed.
Questions you ask the CRM assistant now come to life visually: funnels, KPI cards, gauges, waterfalls, heatmaps, a geographic map and rich data tables. Drill down on pie and funnel charts, compare periods and forecast, switch between day/week/month granularity, catch anomalies automatically, see which data the answer is based on, and export any chart to Excel in one click.
Your notes now do the work: get an AI summary in one click, break a note into task items, or ask for a suggested next step or sentiment read. Speak into the mic and your voice note turns into text automatically. Convert a note straight into a task, follow-up or appointment. And with “ask these notes” you can query all the notes on a customer or deal with AI — while the AI assistant and Super Agents can now search inside your notes too.
AI Chat can now drill into opportunities and teams: ask "Which team closed the most this month?" or "What's the breakdown by loss reason?" and get answers across lifecycle, loss reason, priority and channel dimensions. Answers are scoped to your role and team permissions.
No updates match these filters.
51 results
Sports & Wellness
Personal Care
Real Estate
Automotive
E-Commerce & Retail
Food & Beverage
Tourism & Events
Finance & Insurance
Professional Services
Technology
Media & Advertising
Home Services
Discover the CRM built for your industry.