Eating your own dog food

Hello again everyone and welcome back to AI Corner. Here at SyncHub we believe in eating our own dogfood, and that means we connect our own cloud services and use SyncHub to build our reports.

This lets us both use a fantastic tool (if I do say so myself) and experience the same pain points as our customers, which in turn lets us make the tool even better.

Like many of you we have discovered for ourselves how effective AI can be at building dashboards and reports. There is nothing quite like the feeling of asking a simple question and seeing a beautiful suite of charts roll out before your eyes. But in the process of building these for ourselves we have confirmed the importance of one of our maxims: A report is only as good as the data behind it.

SyncHub gives us access to all the raw data we could want but it’s still up to us (and the AI Assistant) to tease out the precise details we want. So here are my tips for getting a quality data source for your reports with AI.

Work out a clear question

"Show me sales" invites ambiguity, but "Show me monthly recurring revenue by region for the last 12 months, excluding cancelled accounts" gives the AI a much better shot at getting it right the first time. It’s fine to start with a vague question and iterate on it with your AI to remove all that pesky ambiguity. Ask your AI to critique your question before trying to generate data from it.

Sanity-check the first output

Before trusting an output, spot-check a few numbers against a source you already trust. If you don’t have easy access to a trusted comparison ask the AI to tell you everything that might be wrong with the data it produced. AI can often identify issues with it’s own output that it ignored on a the first pass.

Document your definitions

This ties directly into the first point - a clear question is a well defined one. If "active customer" or "closed deal" has a specific meaning in your business, say so explicitly in your prompt or in your Insight setup.

Iterate in small steps

This applies at every level, from working out your definitions to finalizing the layout of an entire dashboard. The narrower the task, the more reliable the result.

Use existing reports as inspiration

If you are trying to replicate or modify a report from your cloud service, describe that report to the AI and ask it how it would create that report from the data available in SyncHub.

Split the report into multiple queries

You don’t have to get everything from a single query or Insight. Create several that independently get some part of the date you want and pull them all into the same dashboard.