Skip to main content
🔥 90 days freeRequest access
Konectus
Experimental

AI Assistant

An AI assistant for your CRM: ask it about your chats, customers and sales

The AI Assistant is your copilot inside Konectus. You use it like ChatGPT, but it knows your chats, your customers and your sales, and it can make changes for you with your permission. It is experimental: it works, it may change and it can get things wrong.

90 days free · No card

AI Assistant demo, with its “Experimental” label and its model picker: Karla asks what a customer requested last time; the assistant answers with a summary and its sources, and proposes scheduling a reminder message, which waits for approval. Karla approves it and the action is logged, with an option to undo.

Karla asks the AI Assistant about a customer

What the AI Assistant is

It is a chat panel inside Konectus. You write to it as you would to any AI assistant, with one difference: this one looks up your organization's data. Your chats, your customers, your sales and your tickets. It answers with text, tables and charts, and shows you the sources behind each answer.

It also acts. It can tag, assign, move a stage or schedule a message, and create reports and dashboards. Before it changes anything it asks you to confirm, and an admin decides what it may do without asking.

It is not the AI agent: that one talks to your customers on your channels. The AI Assistant talks to you and your team. This is how you use it:

  • It opens as a panel from any module, or in full screen.
  • It accepts attachments and reads images.
  • It keeps the history of your conversations with it.
  • It has a “Knowledge base” mode for questions about how to use Konectus.
  • You choose the AI model it answers with.

What you can ask

Ask it the way you would ask someone on your team

The examples in quotes are questions the app itself suggests.

  • About a chat

    “Summarize this conversation”
    “What is the customer asking for?”
    “Is there any pending commitment?”
    “Write me a reply for this chat”
  • About a customer

    “What has happened with this customer?”
    “What next step do you recommend with this customer?”
    “Are there duplicate customers?”
  • About the inbox

    “Which unread chats are the most urgent?”
    “Which chats have been waiting longest for a reply?”
    “How many open chats have nobody assigned?”
  • About your sales

    “Which sales have gone longest without moving?”
    “How many sales did we win this month?”
  • About your operation

    “How many new customers came in this week?”
    “Create a dashboard with this week's chats and sales”
    “Make me a presentation with the month's results”
  • About how to use Konectus

    Switch to “Knowledge base” mode and ask it how to do something in the product.

What it can do for you

It makes changes for you, and asks you to confirm first

Tag, assign, move a stage or schedule messages. You ask in your own words and the AI Assistant prepares the change.

  1. 1

    You ask

    For example, to tag a chat, assign it to someone on your team or schedule a reminder for a customer.
  2. 2

    It shows you what it will do

    The proposal arrives as a card with the details of the change and its risk: low, medium, high or critical. If nobody answers it, it can expire.
  3. 3

    You decide

    Approve, approve with changes or reject. If an admin set that action to automatic, it runs without a card.
  4. 4

    It is logged and can be undone

    What it ran shows up in the AI activity log, with Undo. Some actions ask for approval to be undone as well.

What it can create

Reports in Excel, PDF or slides, and dashboards with indicators

Ask for the report the way you would ask a person. The AI Assistant gathers the data and hands you the file, or builds a dashboard with the indicators you name.
  • “Make me an Excel report of this month's chats”
  • “Make me a presentation with the month's results”
  • “Create a dashboard with this week's chats and sales”

Demo of the AI Assistant in full screen, with its “Experimental” label and its model picker: Karla asks which chats have been waiting longest for a reply; the assistant answers with a table of three chats and its source. Then she asks for an Excel report of the month's chats and the assistant generates the file, ready to download.

Karla asks the AI Assistant about the inbox

Several AI models

Five models: you choose which one answers

The picker is in the panel itself. A more capable model reasons better and uses more AI credits per question.

Five models: you choose which one answers
DeepSeek V4.1 FlashTierFastWhen to use itThe fastest and cheapest. Ideal for simple questions and instant answers.
GLM 5.3 FlashTierBudgetWhen to use itFast and inexpensive, a bit more accurate for everyday work.
Qwen 3.8 FlashTierBalancedWhen to use itBetter reasoning at a low cost; it answers a little slower.
Claude Sonnet 5.5TierSmartWhen to use itVery capable at analysis and writing. The best value for money.
Claude Opus 5.5TierMaxWhen to use itThe smartest, for complex tasks. The most expensive.

Tiers and descriptions as shown in the app's model picker, translated, in November 2026. The list is updated when a better model comes out, so it may change. Model names are trademarks of their respective owners.

You decide how far it goes

The AI Assistant acts only as far as an admin allows

All of this is set in Settings → AI.

  • A level for every action

    Off, Suggest only, Ask for approval or Automatic. Each action also shows its risk: low, medium, high or critical.
  • Three starting policies

    Careful: everything waits for approval, except low-risk actions. Balanced: low and medium risk run on their own; high and critical ask for approval. Agile: it also sends and schedules the messages you ask for, without a card.
  • Approval cards

    Every proposal comes with three options: Approve, Approve with changes or Reject. Proposals can expire.
  • Undo

    Actions it has run have Undo. Some ask for approval to be undone as well.
  • Pause all actions

    One button stops execution. The AI Assistant keeps answering questions, but runs nothing.
  • AI activity log

    What AI did is logged, so you can review it later.

How it is billed: with AI credits

Every question to the AI Assistant uses AI credits, which are paid for separately from the plan price. There is no fixed price per question: it depends on the model you choose and how long it is. More capable models use more. When the balance runs out, the assistant stops answering until you top up.

Three figures to do the math (details on the pricing page):

  • Included credits, granted once when you subscribe: 100 on Starter, 500 on Professional and 1,000 on Business. Early Access includes 10.
  • Top-ups: 10 credits per dollar, from 5 USD. The account owner buys them and they expire after a year.
  • It is the same balance that AI agents, reply suggestions and summaries use.

So you don't mix them up

AI Assistant and AI agent: they are not the same

AI Assistant and AI agent: they are not the same
Who it talks toAI AssistantYou and your teamAI agentYour customers
Where it livesAI AssistantIn a panel inside Konectus that opens from any moduleAI agentOn your channels: WhatsApp, Instagram and Telegram
What it is forAI AssistantAsking, analyzing, making changes and creating reportsAI agentAnswering your customers from your documents and covering for your team
Who controls itAI AssistantAn admin sets what it does without asking; everything else waits for approval on a cardAI agentAn admin writes its rules, uploads its documents and decides when it replies
How it is billedAI AssistantWith AI credits. You choose the modelAI agentWith AI credits. The model is chosen automatically, based on the question
StatusAI AssistantExperimentalAI agentIn beta

Questions about the AI Assistant

How is it different from ChatGPT?

You use it the same way: you write and it answers. What changes is what it knows and what it can do. ChatGPT is a general assistant; the AI Assistant works inside Konectus with your organization's chats, customers, sales and tickets, shows you where each answer came from and can make changes for you, once you confirm.

Is the AI Assistant the same as the AI agent?

No. The AI Assistant talks to you and your team inside Konectus: it answers questions, creates reports and makes changes with your permission. The AI agent talks to your customers on WhatsApp, Instagram and Telegram, with your documents and your rules.

Can the AI Assistant get things wrong?

Yes. It is an AI and it is experimental. That is why it shows the sources of its answers, asks you to confirm before making a change and leaves what it ran in a log, with Undo. Verify important information before you use it.

Can it send messages without my knowing?

It depends on the policy an admin chooses in Settings → AI. With Careful or Balanced, sending or scheduling a message waits for approval on a card. With Agile, it sends and schedules the messages you ask for without a card. What it runs stays in the AI activity log, and the “Pause all actions” button stops it.

Which model should I pick?

For simple questions, a fast one: DeepSeek V4.1 Flash or GLM 5.3 Flash. For analysis and writing, Claude Sonnet 5.5. For complex tasks, Claude Opus 5.5, which is the most expensive. You pick it in the panel's model picker. The list is updated when a better model comes out.

How much does the AI Assistant cost?

It has no separate price: it uses AI credits. Plans include 100, 500 or 1,000 credits, granted once when you subscribe, and you can top up at 10 credits per dollar from 5 USD. Each question uses credits depending on the model and how long it is. Early Access is 90 days free, no card, with 10 AI credits. Details are on the pricing page.

Does the AI Assistant see all my data?

To answer you it looks up your organization's data in Konectus: chats, customers, sales and tickets. What it may change in that data is decided by an admin in Settings → AI, action by action.

What does it mean that it is experimental?

That it works and you can use it, but it may change and it can get things wrong. We use the same word the app uses. While that is the case, actions, reports and dashboards are switched on per organization.

Ask the AI Assistant about your own chats

Early Access: 90 days free, no card, every feature. Subscribe within that window and you pay half for life.

90 days free · No card