AI Workspace Features
Problems we solve
Case studies
Frameworks & solutions

AI That Thinks Out Loud: Introducing Configurable Reasoning in AI Workspace

Nhi Ha

Friday, May 23, 2025

4

min read

The Problem

In today's workplace, AI is already handling substantial workloads – summarising documents, generating content, and responding to queries. But for many users, the process still feels opaque. You ask a question. You get an answer. But you're often left wondering how that answer was formed.

Our latest upgrade in AI Workspace solves that.

Why Transparency in AI Matters Now

Imagine you're a customer support agent drafting a troubleshooting email. As the AI generates the content, you see its reasoning unfold in real time – from retrieving past ticket history to referencing product documentation. You’re not just receiving an output – you're witnessing the AI's logic.

That's the power of reasoning-capable AI – now transparent and configurable in AI Workspace.

Previously, this lack of visibility created a trust gap. Now, with configurable reasoning steps, AI Workspace allows organisations to control how much of the AI's thought process is revealed – from high-level summaries to detailed, step-by-step reasoning. The result? AI that's more explainable, adaptable, and aligned to your team's context.

What’s New: Key Capabilities

  • Admin Control Over Reasoning Depth – AI admins can customise reasoning visibility across teams and scenarios.
  • Live Reasoning Display – Real-time updates during AI content generation (e.g. “Analysing customer issue”, “Referencing documentation”).
  • Reasoning Traceability – Every output includes a transparent logic chain, aiding audits, training, and debugging.
  • Future-Ready Architecture – Lays the foundation for further explainability and alignment tooling.

Real-World Scenario: Smarter Support Emails

Use case: A customer reports their fridge isn’t cooling. The agent asks AI Workspace to draft a response.

Here’s what unfolds:

  • “Getting customer issue” – Pulls context from historical tickets.
  • “Analysing customer issue” – Identifies root causes (e.g. compressor failure).
  • “Looking for product documentation” – Finds relevant manuals.
  • “Thinking about a response” – Structures a clear message.
  • “Drafting email content” – Outputs a send-ready response.

The result: A message the agent understands – and trusts.

Who It’s Built For

Why It Matters

This upgrade directly addresses three common pain points:

  • Limited Explainability – Previously, users couldn’t see how AI made decisions. Now, each step is visible.
  • Lack of Configurability – Teams need different levels of insight. Now, admins control that.
  • Underused Model Power – Advanced models already supported reasoning – now the platform surfaces it.

In short: you gain clarity, control, and confidence.

Still trusting black-box AI? Turn on the lights with AI Reasoning in AI Workspace.
See it in action – check out the demo below and discover how transparent AI can transform your workflow.

Contact us to enable this feature or request a personalised walkthrough.

Don't miss future insights like this.

Continue reading

Thought Pieces
Case studies
Frameworks & solutions

Why Workforce Supply Is Constrained by Training Throughput, Not Skills Shortages

Why Workforce Supply Is Constrained by Training Throughput, Not Skills Shortages

Button Text
Studio for VET
Case studies
Frameworks & solutions

Why Reviewing Training Content Shouldn’t Require a Download

Still downloading AI-generated training materials to check formatting and structure? That outdated workflow is costing your team serious time - and quality. Here’s how to fix it.

Button Text
Studio for VET
Case studies
Frameworks & solutions

Are Your Training Standards Up to Date? Why Expiry Blind Spots Are a Growing Compliance Risk

If you're a training manager or programme owner working with NZQA-aligned unit standards, here's a question worth asking: Can you quickly identify which of your current standards are about to expire? For many teams, the answer is “not really.” And that’s a problem.

Button Text