# A weekly AI summary of customer feedback, set up from your assistant

Source: https://clickruntime.com/use-cases/summarise-customer-feedback

> Your assistant builds an automation that gathers the week's new feedback from your sheet, survey tool or help desk, groups it into themes with the built-in AI, and posts the summary with real quotes.

## Can ChatGPT or Claude summarise customer feedback every week?

If you paste feedback into a chat, an assistant will summarise it. Doing that every week without pasting means something has to collect the new feedback and run the summary on a schedule. With ClickRuntime, the assistant sets up an automation that reads only the feedback added since the last summary, uses the built-in AI to group it into themes, and posts the result to Slack or email.

### Why the assistant can't do this on its own

- No schedule and no collection: a chat can't fetch last week's survey answers on Monday morning by itself.
- Only new feedback should be summarised. The automation remembers the last row it read.
- The built-in AI needs no API key, so the summary works including on the free plan.

## Set it up from your AI assistant

Prompt to copy:

> Create a ClickRuntime automation: every Monday at 9:00, read the rows added to my Feedback sheet since last time (columns: Date, Source, Text), group them into up to 6 themes with a count and one short quote each, and post it to #product in Slack.

1. **Connect ClickRuntime to your assistant.** Add ClickRuntime as a connector (MCP server) in ChatGPT, Claude, Cursor or another MCP client, using https://clickruntime.com/api/mcp. You sign in with Google or email; there is no API key to copy. This is done once per assistant.
2. **Paste the prompt.** Start a new chat and paste: “Create a ClickRuntime automation: every Monday at 9:00, read the rows added to my Feedback sheet since last time (columns: Date, Source, Text), group them into up to 6 themes with a count and one short quote each, and post it to #product in Slack.” Change the details (times, channels, addresses) to yours.
3. **Read the draft.** The assistant restates what will happen, asks for anything missing, such as your timezone or which channel to post to, and saves a draft automation. A draft never runs on its own, so nothing happens until you agree.
4. **Connect accounts and save keys.** The assistant asks you to connect Google and pick the Feedback sheet, and to connect Slack for notifications. For Typeform, Zendesk, Intercom or app-store reviews it asks for that service's API key, saved in the ClickRuntime dashboard.
5. **Test it.** The assistant runs it now on the existing rows and shows you the summary before turning on the schedule.
6. **Turn it on.** The assistant tells you what will run, when, and which services it will contact, then asks your permission to turn it on. From then on it runs on ClickRuntime's servers, and you can close the chat.
7. **Check it in the ClickRuntime dashboard.** Each Monday's run shows how many new items it read and the summary it posted. State keeps the last row read, and the run result keeps the themes so you can compare weeks.

Connector setup for each assistant: https://clickruntime.com/assistants

## What the assistant writes

`weekly_feedback_themes.py`

```python
from google.oauth2.credentials import Credentials
from googleapiclient.discovery import build
from clickruntime import on_trigger, credentials, state, ai, notify

SHEET_ID = "your-spreadsheet-id"
SCHEMA = {"type": "object", "properties": {"themes": {"type": "array", "items": {"type": "object", "properties": {
    "theme": {"type": "string"}, "count": {"type": "integer"}, "quote": {"type": "string"}},
    "required": ["theme", "count", "quote"]}}}, "required": ["themes"]}

@on_trigger
def run(event):
    sheets = build("sheets", "v4", credentials=Credentials(credentials.connection("cr_google_sheets_picked", "work")),
                   cache_discovery=False).spreadsheets()
    rows = sheets.values().get(spreadsheetId=SHEET_ID, range="Feedback!A2:C").execute().get("values", [])
    start = state.get("next_row") or 0
    new = [r for r in rows[start:] if len(r) >= 3 and r[2].strip()]
    if not new:
        notify.slack("No new customer feedback this week.", channel="#product")
        return {"summary": "No new feedback"}

    lines = "\n".join(f"{i + 1}. [{r[1]}] {r[2][:400]}" for i, r in enumerate(new[:300]))
    try:
        themes = ai.generate(lines, system="Group this customer feedback into at most 6 themes. "
                             "Quote one short, real sentence from the input per theme.",
                             json_schema=SCHEMA, max_output_tokens=800).json["themes"]
        body = "\n".join(f"• *{t['theme']}* ({t['count']}): “{t['quote']}”" for t in themes)
    except ai.AIError:
        body = f"{len(new)} new items. AI summary unavailable this week; see the Feedback sheet."

    notify.slack(f"Customer feedback, last week ({len(new)} items)\n{body}", channel="#product")
    state.set("next_row", len(rows))
    return {"items": len(new), "summary": body[:2000]}
```

- Items are numbered and trimmed to 400 characters each, then sent in one AI call, which keeps token use low.
- The JSON schema makes the AI return themes, counts and quotes in a fixed shape the script can format.
- The row position is saved only after posting, so a failed run doesn't skip feedback.

## A schedule, not a trigger link

Use a schedule for the summary: weekly or daily.

If feedback arrives by webhook (a survey tool), add a trigger link that stores each answer as it comes in, and keep the weekly schedule for the summary.

## Limits

- One AI call takes about 32,000 input tokens. Larger weeks are split into batches, or summarised per source first.
- AI tokens count toward the workspace's monthly allowance, including on the free plan.
- The AI reads text only. Screenshots and attachments are not sent.

## Troubleshooting

### The summary is too generic.

Ask your assistant to name your product areas so the AI groups by those, or to include two quotes per theme.

### It summarised old rows again.

State was cleared or rows were deleted above the saved position. Ask the assistant to track a Date column instead of the row number.

## Related

- [Google Sheets automations](https://clickruntime.com/apps/google-sheets)
- [Slack automations](https://clickruntime.com/apps/slack)
- [Typeform automations](https://clickruntime.com/apps/typeform)
- [Zendesk automations](https://clickruntime.com/apps/zendesk)
- [Intercom automations](https://clickruntime.com/apps/intercom)
- [app-review-digest](https://clickruntime.com/templates/app-review-digest)
- [get-an-ai-summary-of-new-feedback-rows](https://clickruntime.com/templates/get-an-ai-summary-of-new-feedback-rows)
- [ai-sentiment-and-category-tagging-of-feedback-rows](https://clickruntime.com/templates/ai-sentiment-and-category-tagging-of-feedback-rows)
- [AI triage and summaries](https://clickruntime.com/use-cases/ai-classify-and-summarise)
- [Form submissions to Google Sheets](https://clickruntime.com/use-cases/form-to-google-sheets)
- [Slack channel summary](https://clickruntime.com/use-cases/summarise-slack-channel)
- [Assistant scheduled tasks alternative](https://clickruntime.com/alternatives/chatgpt-claude-scheduled-tasks)
- [Make alternative](https://clickruntime.com/alternatives/make)
- [Docs: state](https://clickruntime.com/docs/state)
- [Docs: reference](https://clickruntime.com/docs/reference)

## Extend it: what you can add next

The first version is rarely the last. Because the automation is a Python script with memory, AI and any API available, you grow it by asking your assistant for the change. Each change is a new version, and a failed deploy never replaces the working one.

- **Pull feedback straight from Typeform and Zendesk instead of a sheet** (Any app or API)
  Prompt: "Change my weekly feedback themes automation so it also reads last week's Typeform survey answers and Zendesk satisfaction comments."
- **Tag each item with sentiment and theme as it arrives** (Built-in AI)
  Prompt: "Add to my weekly feedback themes automation: every hour, add sentiment and a theme label to new rows in the Feedback sheet."
- **Alert #support right away when very negative feedback comes in** (Trigger links)
  Prompt: "Add a trigger link to my weekly feedback themes automation for new survey answers, and post answers scoring 1 or 2 to #support straight away."
- **Compare themes with last week and flag what's rising** (Memory between runs)
  Prompt: "Change my weekly feedback themes automation so it remembers last week's themes and marks themes that grew or are new."
- **Create a Linear issue for any theme mentioned five or more times** (Any app or API)
  Prompt: "Add to my weekly feedback themes automation: for each theme with 5 or more mentions, create a Linear issue in the Product team with the quotes."
- **Separate summaries for free and paying customers** (Multi-step logic)
  Prompt: "Change my weekly feedback themes automation so it posts two summaries, one for paying customers and one for free users, using the Plan column."

## Questions

### Which feedback sources can it read?

Any source with an API or export: a support tool such as Intercom or Zendesk, a survey tool, app store reviews, or a Google Sheet where feedback is collected. Save each key as a secret in the ClickRuntime dashboard.

### Can ChatGPT summarise survey responses automatically?

With ClickRuntime connected, ChatGPT can set up an automation that collects new responses and posts an AI summary on a schedule, without you pasting anything.

### Do I need an OpenAI or Anthropic key?

No. ClickRuntime's built-in AI needs no key and is included, including on the free plan.

### Will it make up quotes?

It's told to quote sentences from the input, and you can ask the assistant to check each quote appears in the source before posting.

### Where can the feedback come from?

A sheet, a form or survey tool, a help desk such as Zendesk or Intercom, app-store reviews, or any API.
