The problem

AI Conversation Expert (ACE) is RingCentral's conversation intelligence platform. It transcribes calls, surfaces coaching tips, tracks customer sentiment, updates CRMs automatically, and gives business leaders a real-time view of what's happening across every customer interaction.

But ACE doesn't have one kind of user. It has two.

On one side: frontline agents. They're on live calls, moving fast, and have no time to read anything longer than a sentence. On the other: sales managers and executives. They need context, trend data, and the confidence that ACE is giving them an accurate, trustworthy picture of their business.

The challenge was that both audiences lived inside the same product. The UX writing had to serve a skeptical agent who wasn't sure they wanted AI coaching them and a manager hungry for every insight ACE could surface.

The goal

Write the microcopy, onboarding flows, empty states, tooltips, error messages, and in-app coaching notifications for ACE in a way that respected the fundamentally different needs and headspaces of two distinct user types.

User archetypes

Because ACE serves two meaningfully different users, I developed two archetypes to guide the work.

Archetype 1: The frontline agent

Informational needs

• What does this notification mean?

• Is AI listening to my call right now?

• What am I supposed to do with this?

Jobs to be done

• Get through my calls efficiently

• Don't let anything slow me down or distract me mid-conversation

Psychological profile

• Reactance: Agents who feel surveilled by AI push back. Copy that frames ACE as a tool that helps them, rather than monitors them, is more likely to be accepted

• Cognitive load: Mid-call is the worst time to read anything. Coaching tips and notifications need to be scannable in under two seconds.

• Autonomy bias: People are more receptive to suggestions they feel they chose. Framing coaching tips as options rather than mandates keeps agents receptive.

Archetype 2: The sales manager / executive

Informational needs

• How is my team performing?

• What are customers saying about us?

• Where do I need to focus my attention?

Jobs to be done

• Identify coaching opportunities before they become patterns

• Spot trends early and act on them

• Show leadership that we have clear visibility into the business

Psychological profile

• Authority bias: Managers trust data. Copy needs to feel precise and grounded. Vague language undermines confidence in the insights.

• Progressive disclosure: Executives want the headline first. Supporting detail should be a click away, not the first thing they read.

• Zeigarnik effect: Dashboards tease data and prompt questions, motivating managers to interact with the AI and dig into insights.

Ideation and development

The two archetypes pointed to a clear writing strategy: the same product needed two different voices depending on where the user was in their workflow.

For agents, I kept coaching notifications terse, action-oriented, and framed as helpful nudges rather than performance critiques. The goal was to surface one clear, useful thing, not a paragraph of AI-generated feedback no one has time to read.

I followed a simple rule of thumb: If it wouldn't work in a single breath, it was too long.

For managers, I had more room to work with, but I still applied progressive disclosure. Dashboard labels, insight summaries, and tooltip copy were written to lead with the finding and let managers choose whether to drill down. Empty states on the analytics screens were written to explain what data would appear and why it matters.