← Back to all work

Unified AI Agents

One experience framework for many AI agents

Our AI capabilities each delivered value on their own, but together they felt fragmented. As Product Manager and UX Lead, I defined shared interaction principles so agents and admins could understand the whole system.

Role
Product Manager and UX Lead
Scope
Agent Helper, Case Quality Auditor, Knowledge Agent, Escalation Manager
Try the working prototype (opens in a new tab)Read my full case study (opens in a new tab)

The problem

As our enterprise AI portfolio expanded, capabilities were built to solve different support challenges. Each delivered value independently, but the overall experience became fragmented. The challenge was no longer improving individual features. It was creating a coherent experience across the ecosystem.

  • Fragmented mental models

    Each AI capability had its own interaction pattern, so users had to learn how each tool worked instead of focusing on the customer's issue.

  • Inconsistent AI experiences

    Recommendations, evaluation results, evidence, feedback and human review looked different in every capability.

  • Admin and platform complexity

    Separate configuration surfaces and Salesforce deployment considerations made administration, governance and scale harder.

  • Hard to tell one story

    Demos, onboarding and customer conversations meant moving between separate product experiences, so the platform didn't feel connected.

“Why should agents have to understand four tools to complete one support workflow?”

My role

As Product Manager and UX Lead, I connected the platform vision with the realities of delivery, while feature work continued across several AI products.

Product and platform strategy

  • Identified fragmentation across agent-facing and admin experiences
  • Defined the vision and priorities for a unified AI experience
  • Weighed product, UX and platform trade-offs
  • Aligned stakeholders on direction and scope

UX architecture and hands-on design

  • Defined shared interaction patterns across AI capabilities
  • Designed critical agent and admin workflows
  • Explored AI states, evidence visibility, feedback and human review
  • Validated directions with interactive prototypes

Leadership and execution

  • Delegated capability-level design to designers
  • Set design principles and interaction guardrails
  • Reviewed work to resolve inconsistencies and edge cases
  • Worked with engineering, QA, security, product, GTM and customer teams

The solution

FromFeature-centric
ToSystem-centric

I defined shared interaction principles that made the platform understandable across AI capabilities, without making their workflows identical. Case context, AI output, supporting evidence, feedback, human review and action status follow one pattern. Capability-specific controls stay flexible.

Before Four tools, four patterns
  • Agent Helperown patterns
  • Case Quality Auditorown patterns
  • Knowledge Agentown patterns
  • Escalation Managerown patterns

One support workflow, four tools to learn

After One experience framework
Case 61204Webhook deliveries failing · fictional
Risk rising

Summary, timeline and a drafted reply for this case.

AI outputEvidenceFeedbackHuman reviewAction status

Illustrative recreation with fictional case data. Every capability keeps its own content but shares the same pattern for output, evidence, feedback, review and status.

Try the working prototypeOpen a support ticket with every AI capability in one unified view. Built in Lovable. (opens in a new tab)
“This wasn't about making everything look the same. It was about making the system understandable, trustworthy and scalable.”

Admin, agent and analytics views

Problem
Separate configuration surfaces create administrative friction.
Decision
Shared configuration and governance patterns.
Why it matters
Admins get a more predictable management experience.

Decisions that shaped the platform

01One experience or several improved products?

Direction

A shared experience architecture, keeping capability-specific workflows where needed.

Trade-off

More coordination up front, in exchange for consistency and long-term scale.

Unify how users understand and interact with AI, without forcing every capability into an identical workflow.
02Standardization or flexibility?

Direction

Standardize case context, AI output, evidence, feedback, human review and action status. Keep capability-specific controls flexible.

Trade-off

A predictable mental model means keeping some intentional differences.

Consistency should reduce cognitive load, not restrict functionality.
03Experience unification or technical consolidation?

Direction

Separate the shared UX model from backend and Salesforce packaging decisions.

Trade-off

Shared experiences can span technical boundaries, but need coordination on navigation, admin, permissions and compatibility.

A unified experience shouldn't depend on presenting every capability as one monolithic system.

How the work ran

  1. 01Discovery and experience audit
  2. 02Strategy and alignment
  3. 03UX architecture and prototyping
  4. 04Design and engineering collaboration
  5. 05Integration and validation
  6. 06Rollout and optimization

Impact

  • Simpler experience

    Shared patterns reduce the need to learn separate interaction conventions.

  • More AI transparency

    Evidence, explanations and human controls make AI-assisted work easier to understand.

  • Operational consistency

    Common administration and interaction patterns make the platform easier to manage.

  • Room to grow

    A shared framework is the foundation for adding future AI capabilities.

These are strategic benefits of the design direction, not measured production outcomes.

Customer stories for products in this system

Watch and read