The problem

Most CCaaS platforms today are still built around a traditional, manual workflow. Agents are expected to juggle customer information, CRM records, past interaction history, and next steps entirely on their own — with no real assistance during the conversation itself.

This creates heavy cognitive load. Agents constantly switch between screens and tools just to piece together context, which slows down resolutions and makes the job harder than it needs to be, especially for newer agents.

Research & competitive analysis

My role: I led this research and competitive analysis.

I studied how leading CCaaS and support platforms — Genesys Cloud CX, Zendesk, Five9, and Talkdesk — handle the agent experience. The pattern across all of them was consistent: the problem was never a lack of features. It was complexity.

That insight became the foundation for how we approached the whole product.

My role

I worked across the full product rather than a single module — owning UX research and competitor analysis, and contributing to the layout and interaction design across every major module of the CCaaS platform, working closely with the team throughout the 3-month build.

The solution

We rebuilt the product around one core principle: make the AI a working partner throughout the agent's process, not a bolt-on feature. Here's how that played out across the key modules.

Outcome

The result was a fully automated, conversation-first workflow — a meaningful departure from how traditional CCaaS products operate. Where most competitors offer AI as an isolated feature in one part of the product, we built AI assistance into every major touchpoint of the agent's process.

Early feedback from users was positive, particularly around how much manual work the conversational automation removed from day-to-day tasks.

Reflection

Working across the entire product rather than a single module pushed me to think beyond individual screens and consider how modules connect as one system — the home dashboard, agent workspace, and campaign flow all needed to feel like one consistent conversational experience, not three separate features stitched together.

The hardest part of the project was resisting the instinct to just add more panels and information the way most competitor products do. The research made it clear that the real opportunity wasn't more features — it was reducing what the agent had to manage manually. Committing to a conversational, AI-first approach meant rethinking familiar patterns, like tasks, reminders, and interaction history, instead of just copying how existing tools handle them.

If I were to take this further, I'd want to validate the AI Assist responses more rigorously with real agents over a longer period, to see how it holds up across edge cases and high-pressure conversations — not just typical ones.

The problem

Across the company's product line, account and plan administration followed the pattern most enterprise SaaS tools still rely on: admins clicking through tab after tab to get anything done. Managing users, assigning permissions, buying add-ons, or updating a subscription each meant a separate detour through its own screen and its own set of forms — and because every product in the lineup handled this differently, the experience wasn't even consistent from one tool to the next.

Research & discovery

My role: solo research and analysis.

I looked at how leading enterprise platforms structure their admin experiences — specifically how each handled user management, billing, and permissions.

A few patterns showed up consistently across all four: administration was fragmented across separate sections for users, billing, and permissions, adding navigation overhead for admins trying to see the full picture. Workflows leaned heavily on forms — creating a user or updating a subscription meant working through multiple configuration screens. Permission systems were flexible but complex, often taking several steps to configure for a large org. And automation was minimal almost everywhere; even routine, repetitive tasks relied on traditional click-through interfaces rather than any real assistance.

My role

I designed the end-to-end administration experience for MyPlans, solo, from scratch — defining the overall information architecture for the admin center, structuring account management, user management, and billing into one unified experience, designing the workflows for roles, users, teams, and subscription management, and building the conversational AI flows for common administrative tasks. I also built the reusable UI patterns and components that kept the experience consistent across modules, and worked closely with product managers and engineers to validate workflows against real implementation constraints.

The solution

MyPlans brings everything an administrator needs — account, users, billing, and AI-assisted actions — into one consistent workspace, shared across every product in the company's lineup.

Outcome

MyPlans replaced a fragmented, tab-heavy admin experience with one unified workspace — consolidating account, user, and billing management, with conversational AI handling the most repetitive actions directly instead of routing admins through multi-step forms.

Reflection

Owning MyPlans end-to-end, solo, was a different kind of challenge than working within a larger team. Every architectural decision — how account, users, billing, and AI actions related to one another — was mine to make and defend, which meant the research phase carried more weight than usual. Getting the information architecture right early mattered, since a wrong structural call would have been expensive to unwind later inside a one-month timeline.

The hardest decision was how far to push conversational AI as the primary interaction model rather than a secondary shortcut. It would have been safer to keep traditional forms as the main path and add AI on the side, the way most competitors do. Committing to AI-first for core admin actions meant designing fallback paths carefully, so admins who preferred traditional controls weren't penalized for it.

If I extended this further, I'd want to pressure-test the conversational flows against more ambiguous, real-world admin requests — not just clean, well-formed examples — to see where the natural language model needed more guardrails.

The task

Write the concept, script, and storyboard for a 30-second product video introducing an AI-powered Customer Data Platform with Customer Value Management capabilities — aimed squarely at CMOs in retail and hospitality, an audience with very little patience for another generic "AI-powered" pitch.

Process

The idea the video needed to land in twenty seconds: customer churn almost never happens all at once. It builds quietly, through signals that get missed — an unopened email, a skipped store visit, a complaint that never got resolved. By the time a business notices, the customer is already gone.

The product's answer is to unify those fragmented signals — email opens, store visits, complaint history — into a single connected customer profile, and use that profile to catch churn risk before it happens instead of explaining it after the fact. Once a customer is flagged as at-risk, the platform triggers a re-engagement offer automatically, turning a passive warning into a direct action.

Storyboard & graphic design

The video opens on a customer rendered as light and scattered data points — present, but not yet whole. No shape has formed. It's a quiet visual metaphor for a customer slipping away before anyone notices.

Beat 01 — signals with nowhere to land. The customer hasn't taken shape yet.

As the product enters the story, that absence starts to fill in. Scattered signals — chat, social, calls, email — pull together into a single connected shape, mirroring how the platform gathers fragmented touchpoints into one unified customer view.

Beat 02 — the same signals now converge into one unified customer profile.

Animation clip — signal-to-silhouette formation

When the re-engagement offer triggers, the silhouette resolves from a data-point outline into a labeled, at-risk customer record — the moment the platform turns detection into action.

Beat 03 — risk is identified and named, ready for a re-engagement offer to fire.

The video closes on a deliberately understated line, positioning the platform against generic "AI-powered" claims in the category before landing on its real promise: intelligence built specifically around the client's own data, not a generic model.

Outcome

The result on screen ties the story directly to a number: a 20% reduction in churn, alongside a supporting 28% boost in reactivation reported by a past client — turning an abstract "AI catches churn early" pitch into something a CMO can act on.