
Workflow 3: Aggressive evaluation with chained brokers
Aggressive evaluation is certainly one of my favourite AI use circumstances as a result of AI is nice at discovering patterns throughout giant datasets. The secret is giving every agent one job.
Give every agent a job
Every agent completes one activity earlier than passing structured output to the following stage:
Agent 1, GBP classes: Appears on the classes for the enterprise I am working with, finds extra classes, and opinions competitor classes.
Agent 2, Location and repair space pages: Appears at opponents’ location or service space pages, compares them to ours, and identifies variations.
Agent 3, Content material gaps: If a device exhibits now we have content material gaps, this agent helps me plan subsequent steps primarily based on search quantity.
Consolidate right into a closing agent: The ultimate agent combines these outputs to establish patterns and something the sooner brokers missed.
Acquire knowledge
Earlier than AI analyzes something, I accumulate structured knowledge. For Google Enterprise Profile evaluation, this contains:
For location web page evaluation, I accumulate:
- Web page sort
- Inside hyperlinks
- Content material construction in comparison with opponents
- Historic natural efficiency
Run aggressive evaluation in 4 steps
The entire aggressive evaluation runs in 4 steps: outline your dataset, pull the information, run patterns, and establish gaps.
This is an instance immediate for GBP classes particularly:
“You might be analyzing native search engine marketing indicators for [business type] in [city]. Use ONLY the information offered under. Don’t draw on exterior information. Determine: 1) the most typical main class, 2) secondary classes in 3+ opponents, 3) any class patterns within the high 3 that differ from others. Return a structured desk. Flag any enterprise with a considerably totally different sample.”
The workflow rapidly surfaces patterns similar to:
- Geo places your opponents deal with that you simply’re lacking
- Evaluate velocity you are not preserving tempo with
- Content material gaps
- Class consensus points, the place your opponents use classes you do not
These insights change into the inspiration for technique, however I nonetheless make the ultimate choices.
