Problem

Salesforce’s Global Search handles over 600M queries monthly, with 20M+ already in natural language (e.g., "my open cases"). With the arrival of Einstein Copilot, a conversational AI assistant, the opportunity was clear: bridge Search and Copilot into a single, intuitive experience that respects user expectations and existing habits.

A user summarized it best:
“I like to start with what I’m familiar with—like the search box—and I want it to take me where I need to go, without making me figure out which tool to use.…”

Yet, merging two paradigms—traditional search and generative chat—introduced friction:

• Where should users go: Search or Copilot?
• Would users trust AI-generated results over traditional lists?
• How do we avoid confusion while offering powerful new functionality?

Product Vision & Business Goal

Create a seamless, intuitive bridge between Einstein Search and Copilot that enables users to naturally transition from traditional keyword search to conversational AI assistance. By leveraging familiar search entry points, this approach reduces friction, increases user engagement, and accelerates adoption of Salesforce Copilot—maximizing the business impact and ROI of AI-powered workflows..

UX Strategy

Maintain a familiar and trusted global search experience while gradually integrating generative AI capabilities through seamless coexistence of Search and Copilot. Recognize that users are accustomed to traditional keyword search and expect relevant, contextual results, so avoid disrupting established workflows abruptly.

Enable flexible user journeys that allow starting in Search and switching to Copilot for multi-turn conversations, or vice versa, without friction or confusion. Avoid duplicating conversational patterns in both interfaces to reduce complexity and user cognitive load.

Leverage existing Einstein Search functionalities within Copilot to provide consistent results and filtering options, preserving usability at scale (1.1 billion monthly searches). Adopt an iterative, user-driven evolution—monitoring search patterns and AI trust to inform a gradual transformation from keyword-driven search toward conversational AI as users and industry mature.

Discover

To understand how users interacted with both Einstein Search and Copilot, identifying key patterns in natural language queries ranging from straightforward record lookups to complex summaries and aggregated data requests. User interviews and competitive research revealed a growing expectation for a unified experience that supports both traditional keyword searches and conversational AI interactions. However, users faced confusion about where to begin their queries—whether in the familiar global search or the newer Copilot assistant.

This research highlighted a critical gap: while search provided fast, relevant results, it lacked seamless integration with Copilot’s multi-turn conversational capabilities. Users desired a fluid journey that allowed them to start with natural language queries in search and effortlessly transition into richer, context-aware conversations without losing continuity or control. .

Define

From our discovery work, it became clear that users were unsure when to use Search versus Copilot. While both tools had their strengths, the lack of integration created confusion. Users often started with Search because it was familiar, but found themselves needing Copilot for tasks that required more context or guidance. This disconnect led to friction—users had to decide which tool to use, switch back and forth, and sometimes restart their task entirely.

The problem we needed to solve was not about choosing one tool over the other, but about bringing them together in a way that felt natural. Users wanted a single starting point where they could quickly find information and, when needed, get intelligent support through AI without switching contexts. The design challenge became: how do we unify the structured efficiency of CRM Search with the flexibility and intelligence of Copilot, so users can focus on getting things done instead of navigating tools?

This definition aligned the team around a clear goal: create a seamless, intelligent experience that supports both quick lookups and deeper, AI-powered assistance—all within the same flow.

Ideate

During the ideation phase, we explored ways to unify Copilot and Einstein Search without disrupting existing user behaviors. With over 1.1 billion searches per month, it was clear we couldn't radically change the search experience just to introduce generative AI. Search is a core behavior—users depend on it for speed, precision, and structure. Any changes needed to be additive, not disruptive. One of the first strategic decisions was to avoid introducing two separate conversational patterns. If Copilot already handled natural language interaction, we didn't want to replicate that inside the search results. Similarly, we had learned from earlier iterations—like our Generative Search Answer component—not to support follow-up conversations within components where they weren't naturally expected. Doing so created ambiguity and clutter.

Instead, we looked at how we could leverage existing Einstein Search functionality within the Copilot experience. For example, when a user asked Copilot to “show me the top 10 opportunities,” the assistant could return a direct answer with a link to the filtered results page from Einstein Search. This approach kept the conversation fluid while grounding it in familiar, high-utility patterns.

Throughout ideation, our guiding principle was strategic evolution over forced change. We aimed to keep global search intact while gradually introducing AI-powered enhancements based on evolving user behaviors and industry trends. This allowed us to test, learn, and scale without breaking trust or usability.

Design

Once the team aligned on supporting natural language queries across the system, we defined a shared decision logic: all queries would follow the same flow, and if the system detected that Copilot could answer a query, it would surface a one-shot AI-generated reply within the search results page. I designed the one-shot Copilot reply card to align with existing Copilot design patterns and visual guidelines. The reply appears at the top of the Einstein Search results—providing quick, contextual help without disrupting the core search experience. This ensured consistency in both function and form across Salesforce’s evolving AI interface.

During this phase, we also addressed an important platform-level question: What happens when both Copilot and Generative Search Answer are enabled? Rather than forcing users to choose, we designed the experience so both answers can coexist when relevant. Admins can configure their org to display just one, the other, or both—based on their use case and preferences.

This approach maintained flexibility and control for enterprise customers, while honoring the integrity of both AI systems. It also ensured that the Copilot reply was a lightweight, non-intrusive enhancement—enhancing discovery without redefining what “Search” means to the user.

Collaboration

This initiative required high-level collaboration across multiple teams. We worked closely with product managers from both the Einstein Search and Copilot teams, as well as designers and technical leads from each domain. Given the shared surfaces and overlapping capabilities, cross-functional alignment was essential to ensure a cohesive user experience.

Regular working sessions, async design reviews, and shared strategy docs helped us stay coordinated across orgs. Our collaborative approach ensured that Search retained its strengths while Copilot was introduced in a way that felt purposeful and seamless for users.

Impact

Integrating Copilot replies into Einstein Search improved user efficiency by delivering AI insights without disrupting familiar workflows. This seamless blend reduced confusion about which tool to use, lowered barriers to AI adoption, and increased AI Copilot adoption.

Giving admins control to enable or disable Copilot and Generative Search Answer ensured flexibility and customer satisfaction. The solution preserved billions of monthly searches while enabling Salesforce’s AI strategy to evolve thoughtfully, balancing innovation with user trust