Wingman reviews every interaction your team handles using specialized agents that work together as a single pipeline. Each agent handles one part of the review and adds its own layer of information to the record.

Every interaction passes through all configured agents as one pipeline, and Wingman runs this on 100% of interactions. Each agent produces one output, and between them the agents produce a score, a summary, a next action, extracted entities, a translation, and a record update for every interaction.

The depth of the review is configurable by interaction type, so a one-minute reminder call and a 30-minute claims call don't have to receive the same level of attention.

Use this guide to understand what each agent does and what it produces.

Reviewer

The Reviewer scores each interaction against your quality criteria and stores the reasoning behind the score. Because that reasoning is saved with the result, every score is explainable — when you review a score, you can see what led to it.

Produces the Score.

Summarizer

The Summarizer writes a short, readable summary of what the interaction was about. Anyone can understand the interaction from the summary without reading the full thread or listening to the call.

Produces the Summary.

Next Action

The Next Action agent identifies the follow-up the interaction calls for and records it, so commitments and next steps aren't lost when the conversation ends.

Produces the Next action.

Extractor

The Extractor pulls key entities out of the conversation — names, dates, amounts, and other details that matter to your business.

Produces the Entities.

Translator

The Translator translates the interaction so your team can read it in their own language, across 50+ languages. Customers write in theirs; your team reads in yours.

Produces the Translation.

Tip: To learn how Wingman fits with the rest of Spinoco, including coverage, privacy, and setup, see our Get started with Wingman guide.