NICE CXone

TURNING DATA INTO A CLEAR NEXT STEP

This was the tool I wished existed back when I worked in operations.

CXone Agent Observability: a deployed AI agent's page with an AI-generated summary, recommended actions and results charts
The problem

How do you find what matters in a sea of data?

CXone's reporting suite was powerful on paper, packed with dashboards, drill-downs and configuration. But honestly, contact center admins didn't have time to explore it. Reporting is only one slice of their day, squeezed in between everything else.

When we talked to them, almost every admin said some version of the same thing: “I don't know what I don't know.” They wanted to use the platform more and understand their contact center better, but the depth of the tools kept getting in the way.

What they really needed were three answers: what's happening right now, what should I care about most, and what should I do about it?

I'm reading as a…
NewcomerDeveloperSecurityProductResearcher
Start here: KERIpy demo scripts
The Latest for developers

Ten readers, ten front doors

Pick a persona and the home page rewrites itself for your level.

🔍vLEI
vLEI ecosystemGLEIF
ACDC specToIP
IIW session notesIIW

Search across 96 sources

One search bar on every page, pulling from all of KERI.

SOCIAL PASS🔐
Stadt Dresden
Valid · 2026

A wallet people already know

Credentials look and feel like the cards in your phone's wallet.

🏛️City of Dresden✓
🔏Issued by Kaprion✓
🙂Your Social Pass

A chain of trust you can read

Every credential shows exactly where it came from, in plain language.

Recent activity
Transit discountToday
Museum entryMon
Pass shared w/ librarySun

Activity like a banking app

A familiar log shows when and where your pass was used.

This week ✦
Calls are up 18%, mostly billing questions.

A summary before the charts

Admins get the story in a sentence, then dig in only if they want to.

✦ AI insight3 sources
Long holds on refund calls
📞 “I've been waiting 20 min…”
📞 “Where's my refund?”

Evidence behind every insight

Each AI finding links to the real calls behind it, so admins can trust it.

When a refund call waits 5 min
Then offer a callback
Create automation

Preview the automation first

Admins see exactly what will happen before anything is switched on.

ShippingRefundPassword
Draft · Refund policy

Drafts from a topic

Pick the topic and a reply drafts itself from the knowledge base.

↻ RegenerateSimplify✎ Prompt

Agents stay in charge

Regenerate, simplify or steer the draft before anything is sent.

Custom prompt
Make it warmer and shorter|
More friendlyAdd next steps

Guidance for prompting

Suggested prompts help agents ask for what they need.

🏥 Clinic 🛡️ Insurer 💊 Pharmacy 🔑

Map trust before screens

I mapped who authenticates with whom before designing a single screen.

Transactions
Lab results → InsurerVerified ✓
Rx → Pharmacy
Referral → ClinicVerified ✓

Every transaction at a glance

Admins can see in real time that sensitive data arrived safely.

Set up a data pathway
1 Connect2 Verify3 Share
✓Pathway secured

Setup, one step at a time

A guided flow turns a complex protocol into three clear steps.

🩷
My Clara Card
KindnessAdventureFamily

A Clara Card to reflect

Daters define what they actually want before they swipe.

Prospects
SamChatting
Alex1st date
Jordan2nd date ♥

Prospects, organized

A simple log keeps track of who's who and how it's going.

After the date
How did it feel?😊🤔😍

Journal after each date

Quick reflections help daters spot the patterns that keep repeating.

🎧 “What's the return window?” ✦ 30 days, receipt needed

Listens, then answers

The copilot hears the question and surfaces the answer, hands-free.

Channel
🎙 Voice💬 Chat✉️ Email

Adapts to the channel

Help changes shape for voice, chat or email conversations.

✦ Suggestion Offer a 10% loyalty credit?
AcceptDismiss
✓ Agent approved

Agents keep control

Nothing happens without the agent's okay.

What I heard
If I show my pass at the museum, what else can they see?
Resident
Only what you choose to share. Want to see who issued it?
Me
Yes. I just need to know it's really from the city.
Resident
A composite of what residents told us
I have 40 dashboards and no time to open any of them.
Admin
What would you want to know first thing Monday?
Me
What changed, why, and what I should do about it.
Admin
A composite of what admins told us
Half my replies are the same refund answer, rewritten.
Agent
What would make an AI draft feel safe to send?
Me
It has to come from our policies, and I need to be able to tweak it.
Agent
A composite of what agents told us
I send records to the insurer and just… hope they got them.
Admin
How do you check today?
Me
I call them. Sometimes twice.
Admin
A composite of what administrators told us
I keep dating the same person with a different name.
Dater
What do you wish you could see about your dating life?
Me
What's actually working, and what I keep repeating.
Dater
A composite of what daters told us
By the time I find the article, the customer's already annoyed.
Agent
What if the answer came to you while you talk?
Me
As long as I decide what gets said.
Agent
An imagined conversation, based on what we knew about agents
I keep hearing about KERI, but where do I even start?
Newcomer
What would help you most on your first visit?
Me
Something written for my level, and a way to see what changed lately.
Newcomer
An imagined conversation, based on what the personas need

This was the tool I wished existed back when I worked in operations, so I was thrilled to lead the product strategy and design Automation Opportunities end to end!

Admins didn't want more data. They wanted someone to tell them what mattered and how to fix it, kind of like how a good fitness app turns a pile of numbers into a clear goal and a next step. Bite-sized insights, not another wall of charts.

The other big finding: people hesitate to trust AI they can't see into. If an insight couldn't show its work, admins wouldn't act on it, and that shaped everything that came next.

I designed around three goals:

  • Make complex analytics instantly understandable.
  • Make every insight actionable.
  • Help admins feel confident, not overwhelmed, using AI.
OCIOracle certified
AI Foundations Associate
OCIOracle certified
Generative AI Professional
Constraints & trade-offs

Earn attention in seconds, not minutes.

  • Time-poor users. Reporting is one slice of an admin's day, so every insight had to make sense at a glance.
  • Trust in AI. Recommendations couldn't act like a black box. Each one had to show its evidence before anyone would act on it.
  • An existing product. This had to live inside CXone's reporting, alongside the dashboards people already used, not replace them.
  • Speak the business's language. Insights had to connect to things admins could take to leadership, like handle time, containment and savings.
  • Power users still matter. The full dashboards stay put, so nobody loses the detail they rely on.
What I designed

Insights that tell you what to do next.

An AI-generated summary in plain language above three recommended next steps, each tagged with its expected impact

Lead with a plain-language summary, not a chart

Every view now opens with an AI-generated summary: a few sentences on what's happening and why it matters. On a deployed AI agent, that's how much password reset requests have dropped since launch, when the change started and what to expect next.

Right underneath sit the recommended next steps, each tagged with its expected impact, like more contained calls or shorter handle times, with the button to act right there on the card.

Charts in CXone comparing password reset inquiries with and without the AI agent, by week and by category

The charts are still there, just below, comparing results with and without the AI agent for anyone who wants to dig in.

The trade-off: a summary means the AI decides what to say first, and that's a lot of trust to ask for. We kept every chart on the page instead of hiding them, so admins can always check the summary against the numbers.

We also showed just the top three recommendations, with a link to see the rest. Showing everything would have been more complete, but it would've turned right back into the dashboard overload we were trying to fix.

An expanded AI insight showing projected savings, a chart of automatable work and a link to the interactions behind it

Attach evidence to every AI insight

Trust was the real barrier, so every insight comes with its receipts. In Recommended Actions, each row leads with what's happening, like billing inquiries spiking after statement delays, then why, the suggested fix, its priority and its projected annual savings.

Expand a row and you see the proof: a chart of how much of that work could be automated, a plain-language explanation, and a link to the real interactions behind it. Admins can see exactly where a conclusion came from without leaving the page.

Recommended Actions in CXone: AI insights ranked by priority and projected savings, one expanded with its evidence, next to the AI Assistant

And if they'd rather just ask, the AI Assistant sits right alongside, with prompts like See what has changed since yesterday.

The trade-off: evidence takes up space. A tighter list would fit more insights on screen, but an insight without receipts is one nobody acts on. So the evidence lives inside the row, one click away, instead of on a separate page.

Leading with dollars was a deliberate choice too. Projected savings are an estimate, but they connect each fix to something admins (and their bosses) already care about.

Animation: choosing Create AI Agent on a recommendation opens a preview of what will happen, its impact and projected ROI

Show exactly what an automation will do before it's created

When an admin acts on a recommendation, like creating a billing inquiry tool, they get a preview first: what will happen in plain words, the intent it affects, the expected impact on CSAT, how many escalations it touched today and the projected ROI. Only then do they hit Create AI Agent. No black boxes.

Once it's live, Deployed Actions keeps score: what's running, its status and the impact it's having. That closes the loop from insight, to action, to results.

Deployed Actions in CXone: summary stats for automations that are live, and a list of each one with its type, status and impact

The trade-off: a preview adds a click between insight and action. We kept it anyway, because that pause is what gives admins the confidence to say yes, and an automation they trust is one they'll actually turn on.

Before & after
Before

Powerful dashboards, and no time to dig through them.

After

A plain-language snapshot, the evidence behind it, and a clear next step.

Big Picture Data and Automation Opportunities shipped in CXone, and the response has been really positive! For the first time, reporting feels less like a chore and more like a partner: a clear snapshot, the evidence behind every AI insight, and a straight line from insight to action to results.

Getting to design the tool I wished I'd had back in operations was such a full-circle moment.

Looking back

I'd love to track how often admins act on a recommendation, and how many of those automations are still running a month later. That's the clearest signal the insights are earning their trust.

I've learned that trust in AI is a design problem, not just a data problem.

HAVE AN IDEA?
LET’S JAM!
HELLO
CONTACT ME
HOLA