When I sat down with Jason Smith on The Truth About Social Ads, the part I was most curious about was how he actually uses the product day to day. His answer was Claude.
Jason connected the Wicked Reports MCP to Claude. Now he asks it things like "give me a week-over-week comparison for this client over the last 30 days" or "performance is down, what's going on?" and gets answers built on verified attribution data instead of platform dashboards.
AI wants to agree with you. That's fine for drafting a post. It's a real problem for attribution.
Attribution has rules. After a budget change, there are metrics you should judge it by and metrics you shouldn't touch yet. There are trends that matter and trends that are just noise. A general AI doesn't know which is which, so it answers confidently either way.
Jason gave a perfect example later in the episode. He asked Claude to analyze some Meta data and it told him a frequency of 2 over 30 days was too high for retargeting. He had to correct it. It was repeating Meta's generic best practice, not reasoning from real results.
We built agents that do the attribution science for you. Wherever you're pulling your data, you type "Ask Wicked" and those agents step in.
They work from your goals if you've set them in the app. If you haven't, they ask you first and use your answers as the guiding principles. That matters because the right analysis depends on what you're trying to achieve.
The agents are also prescriptive. Good data should tell you where to move money and where the problem is when returns are off. Sometimes all the spend got pushed onto the first winning creative while ten other good ideas got 80 impressions each. You don't need new ideas. You need to give the ones you have a fair shot.
And we watch the logs. If an answer isn't right, we check which rule the agent should have used and add the instruction. It keeps getting tighter.
Jason does one thing most people skip. When he onboards a client, he gets on a call and fills in the Five Forces settings with them line by line: new customer ROAS, returning customer ROAS, cost per lead and acquisition targets for top of funnel and retargeting, broken down by platform.
He said that made Wicked perform ten times better for his agency. I'd say that's because the AI finally knows what success looks like for that specific client. Without goals, any advice, ours included, is as generic as what Meta and Google tell you inside their own platforms.
It's a connection that lets AI assistants like Claude query your Wicked Reports attribution data directly. You ask questions in plain English and the answers come from verified, click-level data.
It brings in Wicked Reports' attribution agents. They analyze your data against your goals and the rules of attribution, rather than giving a generic AI answer.
It works without them, but it will ask you for them first. Setting your Five Forces benchmarks in the app gives the most accurate, client-specific answers.
Watch Jason describe it here or watch the full episode.