Riff: build a customer-insights workflow you can inspect

Define the customer-insights job before connecting services

The 24 October 2025 launch video builds a Customer Insights Analyzer with Google Sheets, OpenAI and Slack. Riff turns a natural-language idea into an application, while the demonstration provides a concrete connected workflow to examine. Its claim to be the first platform for “real work” is promotional framing, not independently established here.

1. Use the exact product resource: https://riff.new/?utm_source=youtube&utm_medium=video&utm_campaign=RiffLaunch&utm_content=AILUKE.
2. Follow the app overview at 0:35 and integrations at 2:10, distinguishing the demonstrated Sheets/OpenAI/Slack route from other named integrations such as HubSpot, Notion and QuickBooks.
3. Write which customer records should be analysed, the question to answer and where the result should go.

Inspect the app from source rows to returned insight

Riff Cards and templates appear at 4:00, followed by the live build at 5:45. A template can organise the work, but the analysis still depends on the actual rows and the instructions supplied to the model. Keep the original data available so a persuasive insight can be checked rather than accepted because it arrived in a polished app.

1. Begin with a small set of recognisable test records and confirm which sheet the application reads.
2. Compare the analysis with those records, noting unsupported conclusions, omissions and assumptions that need a clearer brief.
3. Inspect the Slack destination and actual message, checking that the output belongs to the intended dataset and audience.
4. Revise a specific weak stage and rerun the same example before expanding the input.

Check scheduling independently of the first successful build

Scheduling and automation begin at 9:00. A configured trigger and a useful delivered report are separate outcomes. The historical launch giveaway required an app before 1 November 2025 and offered credits for 2025; that deadline and period have passed, so it is not an active offer in this companion.

1. Define the recurrence, source data window and expected output before enabling the recurring run.
2. Inspect an actual scheduled result and compare its source rows, analysis and Slack delivery with the manual example.
3. Keep missing data, failed integrations and unresolved questions visible instead of silently treating a partial report as complete coverage.

The takeaway

The useful business app has a traceable path from the intended records to a checked insight and delivered message, including later scheduled runs.