Skip to content

Your theme framework, built from your data and sped up by AI.

Kapiche's engine reads every comment and builds a theme framework from the patterns in your own data, with no code list to set up first. AI then names, summarises and refines it, so your team gets to a framework they trust in minutes, not weeks, with every number traced back to what customers said.

Figure 1.The same question, asked twice: a general AI changes its answer, Kapiche does notDemo data
Project
A general AI
View
Ask 1

Why are customers calling about the app?

Based on the feedback provided, the main reasons appear to be:

  • Login problemsabout a third
  • Slow performance~25%
  • Confusing navigationaround 20%

Source: none

How it works: three layers

  • The Kapiche engine

    Finds the concepts in every comment and how they connect, groups them into themes and theme groups, and counts every comment against them. The same data gives the same framework, every time.

  • AI on top

    Suggests theme names and topics, writes a summary and suggested actions for each theme, and helps you merge, split and tidy. The slow parts of building a framework take minutes.

  • Your team

    Inspects the evidence, edits and verifies the framework once, and the change flows to every report, dashboard and AI answer.

Figure 2.Curating themes: Kapiche builds the framework, your team makes it trueDemo data
Project
Cancellation calls
View
Theme Builder

Framework overview

1,204

total verbatims

3

theme groups

18

themes

+38

new data added

  • Claims (6 themes)47% of all verbatims

    Members waiting on claim decisions and chasing updates by phone.

  • Billing and payments (7 themes)32% of all verbatims

    Delayed refunds, unexpected charges and hardship requests.

  • Digital and app (5 themes)21% of all verbatims

    Log-in failures since the October app update.

Built by Kapiche, named and summarised by AI

A general AI can summarise comments. It can't give you a framework you can run a business on.

Use both: Kapiche connects to Claude and Copilot, so your AI answers from your trusted framework.

Kapiche in your AI

A general AI on its own

Reads the sample that fits in its window, gives different themes each time you ask, estimates the numbers, can invent themes that sound plausible, and can't show you which comments sit behind a figure. It knows nothing about your past frameworks, segments or metrics.

Kapiche

Reads every comment, keeps the same framework until your team changes it, gives exact counts and percentages, opens the verbatims behind every theme, and connects each theme to your segments and to NPS, CSAT or any metric you track.

Starting from your data, not a code list

Traditional coding starts with a list of categories and sorts comments into them, so anything outside the list is missed. Kapiche starts with what customers said. In one employee survey, “new joiners” and “senior staff” surfaced alongside compensation, leadership and career development, two issues a predefined list would have missed.

Figure 3.The same two comments, coded by a fixed list and read by KapicheDemo data
Project
Employee survey
View
Comments
  • “I started three months ago and still have no idea how pay reviews work. The induction skipped half of it.”

  • “The senior team never explains decisions, and nobody has talked to me about where my career goes.”

What you get

  • A theme framework with theme groups and themes, and the share of comments behind each
  • How much of your feedback the framework covers, so gaps are visible
  • An AI summary and suggested actions for every theme, with sample verbatims
  • Each theme's impact on NPS, CSAT or any metric you track
  • New and emerging topics flagged as new data arrives
  • The same themes in Smart Reports, Ask Kapiche, Claude and Copilot
Figure 4.Theme Builder: a theme's AI-written summary and suggested actionsDemo data
Theme Builder: the theme tree with each theme's share of all verbatims, beside the theme Excessive Delays and Poor Service with its AI-generated summary and a suggested action.
Figure 5.Theme Builder: the concept map behind a themeDemo data
Theme Builder: the context network, a map of the concepts behind a theme, such as delays, vulnerability and AFCA.

Results

Kapiche turned unstructured customer feedback into clear, quantified insights, making it easy to understand sentiment and identify key themes—even without predefined scores. It provided the clarity we needed to act on drivers of customer feedback.

Emma WuHead of Customer Experience, PEXA

Questions

Isn't this just ChatGPT on our data?
No. Kapiche's engine builds the framework and does the counting, so results are consistent and every number traces to real comments. AI speeds up naming, summarising and refining on top of that.
Will the themes change every time new data arrives?
No. Your framework stays stable, new comments are mapped to it, and emerging topics are flagged for your team to review.
Can we keep our existing framework?
Yes. Bring your own framework and let the engine and AI fill the gaps, or start fresh from the data.
Does it work on calls and chat, not just surveys?
Yes. Themes work on any text, including call transcripts, chats, emails and complaints.

See it with your own data.