Search inside a company is slow, noisy and scattered.
Genie is an AI-powered plugin that sits beside your browser’s search. It searches across the apps a team already uses, learns what each person cares about, and lets them save, organise and share the links they keep coming back to.
Aravind led the research; I designed the user interface and owned the end-to-end UX. Together, we went from problem to tested prototype in four weeks.
- 4weeks, problem to prototype
- 2designers, research to UI
- 12wireframes before the visuals
- 10people in usability testing
Internal search can’t keep up with how teams work.
- Irrelevant results. People spend extra time filtering through things they didn’t ask for.
- No personalisation. Search doesn’t learn from behaviour, so it can’t suggest what’s new and useful.
- Rich media is a blind spot. Images, videos, documents and audio are hard to find.
Our goal an AI-powered internal search that understands intent, returns accurate results, personalises recommendations and handles every kind of content.
Our solution
An AI plugin that learns from behaviour to personalise results, indexes rich media alongside documents, and lets people save, organise and share what they find — so the right information is always a search away.
Five steps in four weeks: concept, MVP, final product.
We researched the problems people face, defined the issues, explored wireframe concepts and finished with polished interface designs.
| Step | Includes | When |
|---|---|---|
| Research | Problems, existing apps | Week 1 to week 2 |
| Strategy | Challenges, goals | Week 1 to week 2 |
| Content | Copy, master copy | Week 1 to week 3 |
| UX | IA, user flows, paper sketch, wireframes, prototype | Week 2 to week 4 |
| UI | Colours, typography, layout | Week 2 to week 3 |
| Presentation | — | Week 4 |
Who it’s for, and what success looks like.
- Business goal
- A seamless internal search engine powered by AI.
- Target users
- Designers, video editors, product managers, sales and marketers.
- Platform
- Web, as a browser plugin.
- Success means people can find the right information, reach it whenever they need it, and share and organise their links.
- Everyday tasks: searching for information, saving links into folders and sharing them with others.
Starting with the people who search every day.
We interviewed people in person and over video calls (Zoom, WhatsApp, Google Meet), and used an online survey to reach more potential users, aiming for 8–10 participants.
- Target audience
- Designers, video editors, sales and marketers who want to be more productive and keep their links in order.
Meet Robert, who can’t afford to search twice.
Robert, graphic designer · Argentina
He gives his best at work and spends his free time learning new design skills and doing a little photography.
“I need a search engine that understands what I’m looking for and saves me time.”
Goals
Find relevant information fast; spend less time searching manually; keep up with trends and competitors.
Frustrations
Irrelevant results; nothing personalised; poor support for documents, videos and presentations.
Needs
Accurate results based on his preferences, in an interface that makes searching effortless.
What Robert says, thinks, does and feels.
Says
- “I need relevant information quickly.”
- “Search often shows unrelated results.”
- “I want an interface that saves me time.”
- “It should understand my preferences.”
Thinks
- “I spend too long sifting through results.”
- “Search should learn from my behaviour.”
- “It can’t handle different formats.”
- “I need the latest trends and analysis.”
Does
- Turns to colleagues and outside sources.
- Struggles to search across formats.
- Filters results by hand.
- Shares his frustration with the team.
Feels
- Annoyed by wasted time.
- Overwhelmed by the volume of information.
- Impatient when search is slow.
- Unproductive when results miss the mark.
Understanding experiences, not just opinions.
We ran individual interviews and focus groups with employees, managers and administrators who rely on internal search, to gather feedback that would shape the plugin.
- Interview type
- Contextual interview
- Duration
- 10 minutes per person
- Channel
- Google Meet
Questions we asked
- How often do you use your organisation’s internal search?
- What challenges or frustrations have you had with it?
- What features do you wish it had?
- How important is personalisation in search?
- How easy is the current search to use?
- Are you looking for any special features?
- What made an experience memorable or effective?
- What would you expect from an AI-powered internal search?
Putting numbers to what we heard.
A survey measured how widespread each need was, to guide what the plugin should do first.
- 76%need to save important work into different folders
- 52%want to add people to folders and manage their access
- 89%need better integrations with the tools they use
Low-fidelity first: structure before style.
Twelve wireframes mapped the plugin panel beside the search page, from the first click to integrations.
Select any image to see it larger.
The high-fidelity panel, screen by screen.
Home. Genie opens beside search, with recent searches and a bottom bar for Search, Folders, Bookmarks, Notifications and Profile.Search. Results from every connected app, with quick tabs for All, Emails, Files, Messages and Tasks.Filter by app. Narrow results to the apps you choose: Gmail, Notion, Figma, Google Docs and more.Quick actions. Every result can be opened, copied as a link, copied to a folder or added to bookmarks.Folders. Save links into folders, or create a new folder in one tap.
Tap the screen or use the arrows to flip through.
More screens
Ten people, one prototype.
We tested the high-fidelity prototype with ten potential users, asking them to complete everyday tasks.
What worked
- Most people completed the tasks.
- The AI understood their queries and returned relevant results.
- The interface was well received.
What to improve
- Add advanced search options for specific queries.
- Give clear feedback when there are no or few results.
Research changed the product more than any screen did.
Testing showed that people could complete their tasks and found the results relevant, and that personalisation and support for different content types mattered a lot to them.
We also learned how colour, type and components work together to shape the experience. Next, we’d do more research to flesh out the full web experience.
