Skip to main content
Gezora.ai

Answer from your knowledge base, consistently

Custom Business AI Chatbots answer from your own knowledge base to improve response times and keep answers consistent across every conversation with your customers.

Inside the workflow

From a customer question to a resolved conversation

Step 1: Your knowledge base is connected

What is this chatbot allowed to answer from?

Connected sources

Your documentation, FAQs, and knowledge base, indexed and kept current

  • Product documentation

    Indexed again as your content changes

    Indexed
  • Support FAQs

    The answers your team already gives

    Indexed
  • Internal policy pages

    Used for context, never shown to customers

    Restricted
  • Out of date release notes

    Removed from the answer scope

    Removed

Ready when

The scope you approve before anything goes live

  • Every source has an owner
  • Out of date pages are removed
  • Restricted topics are listed

Step 2: A customer opens the chat

What is the customer asking?

Opening exchange

The first turns of a live conversation

  1. CustomerMy order has not arrived yet
  2. ChatbotI can check that. Do you have the order number?
  3. CustomerYes, it is on my confirmation email
  4. ChatbotThanks. Reading your delivery status now

Conversation record

What the console fills in as the chat opens

Channel
Your customer chat
Customer
Existing account
Answer scope
Approved sources only
Availability
Around the clock
Handoff
Ready if confidence drops

Step 3: Intent is detected and routed

What kind of request is this?

Intent match

How the message was classified before answering

  • Delivery status question

    94 of 100

    Matches your order tracking answer

  • Refund request

    38 of 100

    No refund wording in the message

  • Complaint

    17 of 100

    No escalation signals detected

Routing path

Where this request goes next

  1. Message read, done
  2. Intent detected, done
  3. Queue chosen, in progress
  4. Answer drafted, waiting

Step 4: The answer is written

Where does the answer come from?

Answer draft

Written from your sources, cited, in your own tone

Question
Where is my order
Source
Your delivery policy page
Citation
Linked beside the answer
Tone
Your own support voice
First response
Typically 8 seconds

Source match

How your own sources ranked against this question

  • Delivery policy page

    96 of 100

    Answers the question directly

  • Shipping FAQ entry

    61 of 100

    Supporting detail only

  • General terms page

    12 of 100

    No delivery detail found

Step 5: Guardrails check the reply

Is this safe to send?

Checks before sending

Every reply is checked before the customer sees it

  • Blocked topics

    Never answered, closed politely

    Automated
  • Source check

    Every answer carries its source

    Automated
  • Low confidence answers

    Held back and passed to a person

    Needs a person

Who clears each check

Two of these three checks clear without a person

2/3
  • Clears on its own2
  • Needs a person1

Step 6: The requested action runs

Can it resolve this, not just explain it?

Action queue

Routine actions running through the integrations you approved

Requested

  • Reschedule the delivery

    Approved integration

  • Resend the confirmation

    Approved integration

Running

  • New delivery date set

    In your order system

Written back

  • Confirmed in the chat

    Customer told straight away

Held

  • Request outside policy

    Sent to your team

Action controls

What runs alone and what waits for your team

  • Approved integrations

    Only the systems you connected

    Automated
  • Action confirmation

    Written back into the conversation

    Automated
  • Anything outside policy

    Sent to your team first

    Needs a person

Step 7: A person takes over

What happens when it cannot help?

Handoff sequence

Context carried across so nothing is repeated

  1. in chat

    Confidence drops, done

    The draft answer is held back

  2. at handoff

    Queue chosen, done

    Routed to a real support queue

  3. first reply

    A person picks up, in progress

    Full conversation context preserved

  4. after close

    Outcome recorded, waiting

    Reviewed with the other escalations

Handoff holds when

Conditions for a clean escalation

  • Handoff reaches a real support queue
  • The customer never repeats themselves
  • Escalations are reviewed regularly

Step 8: Support leaders read the results

Is this actually working?

Support results

The figures behind deflection and satisfaction

Routine queries handled automatically
About 85%Routine queries handled automatically
Typical first response
8 secondsTypical first response
Target satisfaction rating
4.8 out of 5Target satisfaction rating
Availability
Around the clockAvailability

Before and after

How the same queue reads once this is live

CriteriaSupport beforeSupport now
Routine queriesAnswered one by oneAbout 85% handled automatically
First responseWhenever someone is freeTypically 8 seconds
AvailabilityWorking hours onlyAround the clock
SatisfactionReviewed after the factTracked against 4.8 out of 5

Illustration of how Custom Business AI Chatbots runs, with sample records rather than live account data.

What it does

Every answer comes from a source you approved

Custom Business AI Chatbots answer only from the documentation, FAQs, and knowledge base your team already maintains. They read what each customer is asking, route it to the right queue, reply in your support voice with the source shown beside the answer, check every reply before it sends, and run routine actions through the integrations you connected. When they cannot help, the conversation goes to a person with the full context attached.

This is not a wall between customers and your team. It clears the repetitive questions off the queue so your people take the ones that need them.

Who it is for

  • Support teams handling high volumes of repetitive questions
  • Internal IT and HR help desks fielding routine requests
  • Companies whose support demand does not stop at closing time
  • Support leaders who want every answer traced to a source

How it works

The support conversation, phase by phase

Nothing here is a generic support script. The phases run from the sources you approve through to the results your leaders review, which is the order a real conversation resolves in.

  1. Phase 1, Knowledge base agent

    Source grounding

    • Indexes your documentation and support FAQs
    • Removes out of date pages from scope
    • Confirms an owner for every source
  2. Phase 2, Chat intake agent

    Customer intake

    • Opens the chat around the clock
    • Captures the question as it is asked
    • Records channel, customer, and answer scope
  3. Phase 3, Intent routing engine

    Request routing

    • Classifies what the customer is asking
    • Ranks the likely request types
    • Sends the request to the right queue
  4. Phase 4, Answer drafting agent

    Grounded response drafting

    • Ranks your sources against the question
    • Writes the reply in your support voice
    • Attaches the citation beside every answer
  5. Phase 5, Answer safety agent

    Safety review

    • Checks every reply before it sends
    • Declines topics you marked restricted
    • Passes low confidence answers to a person
  6. Phase 6, Support action agent

    Request execution

    • Runs actions through approved integrations only
    • Writes the outcome back into the chat
    • Holds anything outside policy for your team
  7. Phase 7, Support handoff agent

    Handoff to a person

    • Escalates the chat into your support queue
    • Carries the full conversation context across
    • Marks who the chat now belongs to
  8. Phase 8, Support insights agent

    Performance reporting

    • Tracks how many routine queries resolve automatically
    • Measures first response and satisfaction trends
    • Compares resolution rates before and after

FAQ about Custom Business AI Chatbots

Straight answers on scope, timelines, and what running Custom Business AI Chatbots asks of your team.

Get started

Stop paying people to do what an agent can

Tell us what you want to automate. We will map the workflow, deploy the right agents, and train your team to run them.

  • Every agent is trained on your own workflows, never a generic template
  • Most deployments are live within two to four weeks
  • SOC 2 compliant, with a complete audit trail on every deployment
Free demo