Anvayric

IMFIT / Interactive implementation demo

See what happens
after the question.

Change the knowledge. Hand work to a person. Catch a reply that is already out of date.

Real implementation patterns.
Fictional demonstration data.

This browser-only demo reconstructs selected IMFIT behaviors. Replies are deterministic, not generated by a live AI model. Nothing is sent or saved to IMFIT.

Read the implementation case study

An unpublished change must stay out of the answer.

Save different hours as a draft, then prepare a reply. It still uses the approved hours. Publish the draft and prepare another reply to see the difference.

Customer conversation

Fictional lead
Customer / message 1

What are your opening hours?

Response preview

Prepare a reply to see the approved information it uses.

Starting captures the current message and knowledge version. Finishing checks whether the message is still current.

Business context

Approved v1
demo-hours / fictional source

Opening hours

06:00–21:00

These are sample values, not IMFIT’s actual operating hours.

No draft saved. Approved knowledge is unchanged.

Choose an experiment and try the controls. All changes stay in this page.

Inspect the decision

A response should have a receipt.

State
No reply prepared
Source
Waiting for a reply
Reason
The decision will appear here.

Event history

Newest first

    What this means
    for a business.

    The model’s response is one component. Knowledge ownership, workflow state and failure handling make it part of an operation.

    Change information once.

    Review and version business facts so the next interaction can use approved information. A saved draft alone should not change a customer answer.

    Keep responsibility visible.

    Unknown payment terms need an accountable person. An AI response should not invent an agreement to keep a conversation moving.

    Respect the latest context.

    A delayed response should not overwrite a newer customer message. Check state before applying the result.

    What is reconstructed here?

    The knowledge-version snapshot and message-ID comparison model observed IMFIT patterns. This simplified interface, sample facts, event history and staff task are built specifically for the public demo. It does not run the Go backend, Firestore transactions, WhatsApp delivery or the production language model.

    Explore the engineering evidence

    Does your application
    need controls like these?

    The demo isolates knowledge updates, an illustrative staff-review task and stale replies. A client project starts by defining the equivalent cases in your application.

    Discuss your application