Free analysis tool

How many of your support tickets
are "how-to" questions?

We ran this analysis on 7,970 real tickets. 39% were users asking how to do something a human agent explained in words — not bugs, not billing, just "how do I do X?" Upload your export and find out your number.

39% of SaaS tickets are "how-to" questions — our analysis of 7,970 real tickets
14% of issues fully resolved by AI today — despite vendor claims of 40–80%
45% of AI self-service failures: the system didn't understand what the user was trying to do

Gartner survey of 5,728 customers, August 2024. Source ↗

The gap AI and docs leaves open

AI chatbots are good at finding answers. The problem is users who have the answer and still can't complete the task. They know what DNS is. They can't figure out where to paste it. Your chatbot sends them a doc. They open a ticket anyway.

Show Me How is building for that gap — it reads the actual screen and walks users through each click in real time, on the page they're already on. No pre-built flows to maintain. No generic instructions.

Learn more about Show Me How ↗

What this tool gives you

  • Your % of "how-to" tickets — the category Show Me How is built to deflect
  • Full breakdown across 9 issue types with AI confidence scores and reasoning
  • Downloadable CSV — every ticket classified, ready to share with your team

What you'll need

An OpenAI API key

Used to classify your tickets with GPT-4o mini. Costs a few cents for most ticket volumes. Get one here ↗

A ticket export from Zendesk, Intercom, or Freshdesk

We'll walk you through exactly how to export. Works best with at least a few hundred tickets; a few thousand gives the most reliable picture.

About 15-30 minutes

Export and upload take a few minutes. Classification runs in the browser — a few thousand tickets finishes in under 30 minutes.

Your tickets never touch our servers. Ticket text is sent directly to OpenAI using your own API key. Emails, phone numbers, and URLs are scrubbed before each request. Per OpenAI's privacy policy ↗, no data sent is shared or used for training.

Step 1 of 6

Enter your OpenAI API key

The key is used to classify your tickets with GPT-4o mini. It stays in your browser and is never sent to us.

Don't have one? Get one at platform.openai.com ↗

Step 2 of 6

Which platform are you using?

Select your helpdesk to get export instructions.

Step 3 of 6

Upload your export file

Your file is read locally — it never leaves your browser.

Drag & drop your export file, or browse

JSON for Zendesk · CSV for Intercom & Freshdesk

Step 4 of 6

Configure the analysis

Filter by group / team

Uncheck groups that aren't relevant (e.g. internal or sales queues).

Your product

Help the AI understand what your product does so it classifies accurately.

Custom classification notes

Tell the AI about anything in your queue that needs special handling — things a new analyst would need to know. If you see the AI putting things in the wrong category after testing, add notes about that here.

Truncation

Controls how much of each ticket is sent to the AI. More context = more accurate classification. Less = faster and cheaper. The stats below are calculated from your actual tickets.

Agent replies upload tickets to see stats
2,000 chars

The agent reply is the strongest classification signal — resolutions land at the end, so when truncated we keep the beginning and end and cut the middle.

User messages upload tickets to see stats
600 chars

User messages are a secondary signal — the agent reply usually tells us more about what category the ticket belongs to.

Estimated API cost for full run: calculating… · GPT-4o mini, OpenAI's most affordable model

Step 5 of 6

Test run — 25 tickets

Review the classifications below. If they look off, go back and adjust your custom notes.

Starting…

Running analysis

Classifying your tickets…

This may take a few minutes depending on your ticket count.

Preparing batches…