What users disliked was surprisingly consistent.
The current experience felt visually outdated, too layered, and overly dashboard-like. Users also needed clearer loading, error, and accessibility states to trust the assistant more.
Project / 02
Now Assist Redesign
A two-week ServiceNow internship project focused on making Now Assist feel clearer, calmer, and more trustworthy inside enterprise product workflows.
This project started by looking closely at how the current Now Assist panel felt in use and how top AI products were shaping user expectations around clarity, motion, entry points, and trust.
I translated that research into two altitude directions: a near-term redesign for how the assistant should feel inside the platform today, and a future-facing exploration imagining how these AI patterns could evolve over the next three years.
Discuss this project ↗Context
Researching what made the assistant feel noisy and untrustworthy.
The strongest opportunity was not adding more AI behavior, but making the experience feel calmer, more legible, and more accountable at every state change.
Five principles guided the redesign.
Consistent visuals, a clear AI entry point, subtle helpful motion, a friendly neutral tone, and transparency that makes the system feel accountable rather than mysterious.
I redesigned summaries, states, and feedback flows.
The work covered chat summaries, task summaries, feedback popovers, error and caution states, and future-facing concepts for how AI support could scale more gracefully in product.
Design Directions
Two altitude tracks helped separate what should improve now from what could evolve later.
The first direction focused on practical in-product refinement. The second opened a broader conversation about future behavior, patterns, and system trust inside enterprise AI workflows.
Designed for how Now Assist should feel in platform now.
This track focused on practical in-product improvements: clearer hierarchy, less visual clutter, more legible summaries, and cleaner moments of feedback around AI-generated content.
Explored how AI patterns may feel three years out.
I also pushed beyond the current UI to imagine future patterns for summaries and controls, balancing system trust, collaboration, and more adaptive behavior without making the assistant feel louder.
One project, several interaction surfaces.
The redesign touched chat summary, task summary, feedback capture, error handling, caution states, and entry-point behavior so the entire assistant experience could feel more cohesive.
Outcome
The final proposal made trust, feedback, and visual calm feel like one system.
What became strongest in the work was the consistency between principles and interface decisions: every choice aimed to make enterprise AI feel clearer without becoming colder or heavier.
Less clutter, stronger hierarchy.
Summaries and controls were reorganized so users could parse what AI generated, what it suggested, and what needed review much faster.
States became more explainable.
Loading, caution, error, and feedback moments were treated as core design surfaces rather than edge cases, which made the assistant feel more accountable.
Enterprise AI benefits from restraint.
This sprint reinforced that the most useful AI redesigns often come from reducing ambiguity and noise, not from adding more visible AI signals.
