OI Predicted Resolution

How it works:

Predicted Resolution is powered by a custom-trained model built specifically for your environment. It analyses transcript patterns from historical interactions to predict whether an interaction was Resolved or Unresolved.
 

What it checks:

The model looks for signals near the end of the transcript to identify if any actions remain outstanding. For example, it checks whether the customer or agent needs to follow up, wait for an outcome, send additional information, or was transferred elsewhere.
 

Speed & scale:

Because it runs on pre-trained models, Predicted Resolution is fast and scalable, capable of running across 100% of interactions.
 

Output:

The model always returns one of two possible scores:

• Resolved

• Unresolved
  It assigns whichever label it is most confident about.
   

Consistency:

Predicted Resolution is deterministic — the same transcript will always produce the same result if re-run. This makes it reliable for tracking performance trends across teams, queues, and time periods, provided the interaction types are consistent.

OI-Q Resolution

How it works:

OI-Q Resolution uses a large language model (LLM) and a predefined prompt to interpret the full transcript. Instead of relying on pre-trained classification patterns, it uses contextual understanding to assess whether the issue was resolved within that specific interaction.
 

What it checks:

It determines resolution status based on the conversation itself, without reference to a trained dataset. The LLM considers tone, phrasing, and context to make its judgment.
 

Speed & scale:

Because it relies on generative AI, OI-Q Resolution is slower and more memory-intensive. It’s not suitable for running across very large datasets but works well for analysing individual interactions or small samples.
 

Output:

Responses are free-form and may include:

• Resolved

• Unresolved

• Unclear

• Unable to determine
   

Consistency:

OI-Q Resolution is non-deterministic, meaning the same transcript can return slightly different results if queried again.
It should not be used for performance benchmarking, but it can provide deep contextual insights into individual interactions.

When to use

Example

An agent does everything correctly, but the customer must still wait (e.g. for a process to complete or for another team’s action):

• Predicted Resolution: Unresolved (because the customer still has an outstanding action).

• OI-Q Resolution: 

  • Resolved (if the agent fulfilled their responsibility)

  • Unresolved (if the customer expresses dissatisfaction), 

  • Unclear (if context is ambiguous).

Final Takeaway

Use OI Predicted Resolution for consistent, large-scale performance measurement and OI-Q Resolution for deeper, context-based insight into individual interactions.