Job hunt tools optimise for volume, more applications, more tracking, more streaks, and the evidence says volume is both emotionally corrosive and strategically worse. I designed and built a companion engineered around the user's emotional state instead of engagement metrics, while leading two UX designers and keeping every decision survivable in code.

Role
Designer and developer, leading the team
Team
3 designers
Timeline
Ongoing
Platform
Chrome Extension

Context

The hunt is breaking the hunters

The modern job hunt is an emotional endurance test, and most tooling makes it worse. The research is consistent:

72%

say the search hurts their mental health

79%

experience anxiety during the search

66%

feel burned out by the process

Hope, effort, silence, disappointment, repeated dozens of times. I was job-hunting myself while building Jobi, designing for a problem I was living daily.

Enter Jobi: the companion, not the board

Leading tools help you do more of the grind, faster. Jobi takes the opposite position: it is not a job board and does not automate applying. You find the jobs; Jobi helps with the judgement, the tailoring, the practice, and the memory of the journey.

Research

Validating the pain

We interviewed five active job seekers and audited the two leading volume tools, Teal and Simplify Copilot. We clustered every quote into an affinity map; four themes kept surfacing, all sitting on top of the same emotional fatigue.

Affinity map: dozens of interview quotes on sticky notes clustered into four themes: research tax, silence and ghosting, a fragile job tracker, and a repetitive application process.
Fragile job tracker
Everyone hand-builds their own system out of Excel, Notion, email folders and screenshots, and it leaks. It relies on remembering to log, which breaks on mobile, breaks under volume, and gets ambiguous the moment ghosting muddies a “status”.
Repetitive application process
Re-typing the same education and experience into every form, even after uploading a CV. And per-role CV tailoring that costs the most effort with no feedback loop to prove it works.
Research tax
Company research paid upstream before you can even start writing, and heavier still for international and junior seekers.
Silence & ghosting
No reply after applying, no feedback after rejection, and managing your own morale through all of it. This is the emotional layer, and Jobi’s defensible ground.
Research summary board: persona Mark, who applies to 10 to 30-plus roles a week, tracking across Excel, Notion and Google Docs. Key findings: 92% never finish a job application, 15 minutes to 3 days per application, mostly silence after applying. Plus goals, frustrations, and a how-might-we to cut the repetitive work and give one clear view of the whole search.
How might weHow might we reduce the emotional cost of the job hunt, without taking the hunt away from the hunter?

That question shaped every surface and decision that follows.

Design Process

Design, build, and lead in one loop

The three of us ran the research, ideation and design together in Figma. From there I took each screen into Claude Code and built it into real, working software. The part that made it move was how I worked with Claude: I gave it everything, the interviews, the affinity map, every design decision and the reasoning behind it, so it carried the same context I did. That made it more than a co-developer. It became a design strategist I could think out loud with, weighing a design against build complexity, API cost and the business case before any of it shipped.

A wide board of Figma exploration: mascot character directions, mid-fidelity chat flows, lo-fi wireframes and reference UI, and the moodboards behind Jobi's visual language.

The Figma side of the loop: mascot directions, mid-fi flows, lo-fi wireframes and the moodboards behind the visual language.

  1. One shared contextClaude Code held every artefact and research piece, so the build never drifted from the why. The design and the code came from the same source of truth.
  2. Crossroads went back to the evidenceWhen two paths were both reasonable, we did not argue from taste. We fell back on a research finding, or on an earlier decision we had already reasoned through, and stayed consistent.
  3. Every call logged as a trade-offSignificant decisions were written down as crossroads, decision and cost, weighing the design against code, API spend and strategy. That record is the discipline this study is built from.

Solution

One companion, end to end

Jobi rides along the whole hunt as a browser side-panel, judgement, tailoring, practice, and planning, each as its own surface.

The chat: judgement on tap

The home surface is a conversation. Paste a role and Jobi tells you whether it's worth your time, tailors your documents, and researches the company, without leaving the page you're on.

Interview practice

A voice-driven mock interview that role-plays the real thing, questions, follow-ups, and feedback in the moment, so the first time you say the answer out loud isn't in the actual room.

Every voice lesson works fully typed

The crossroads

Voice roleplay is the showcase of the interview, but speech recognition can be unsupported, the mic denied, or text-to-speech can fail, any of which would dead-end the lesson.

The decision

Voice and typed paths run in parallel throughout. A 'type instead' option is always offered, a blocked mic falls back to a text field, and even a speech failure degrades to simply showing the line and continuing.

The trade-off

Two near-identical speech subsystems to build and maintain, accepted so the experience never traps a user.

The tracker

Every application in one pipeline, designed so a growing pile of rejections never becomes a wall of anxiety.

A pipeline that protects the user from their own rejections

The crossroads

A job hunt accumulates rejections. The standard tracker gives every status equal weight in saturated traffic-light colours, turning the board into a wall of anxiety as closed applications pile up.

The decision

Statuses split into Active (front and centre) and Closed (de-emphasised, collapsible), so rejection is a real but tucked-away stage. Colour survives only in a small status dot and the summary funnel.

The trade-off

Closed applications are slightly harder to review, and there's less at-a-glance colour coding; the design leans on labels instead.

Emotional Design

The same toolkit, pointed the other way

Emotional design usually chases engagement. Jobi works in a domain people dread, where two-thirds of Gen Z job seekers report burnout. So the mascot, the motion and the voice all point the other way. They lower the cost of a brutal task instead of pulling someone back for a streak. It is aimed at the nervous system, not the dopamine loop.

The mascot is a companion, not a face

The mascot has no mouth, in any pose. It works through its eyes and the set of its body instead: glasses on when it is analysing a role, a gentle bob while you wait, confetti when you land something. That is deliberate. Jobi already watches your hunt to help, so a full talking face would tip “companion” into “something is watching me,” the one feeling it has to avoid. The mismatch is the charm: a soft, near-cuddly creature delivering a dry, level line reads as deadpan, not harsh. It is also why the cursor-tracking eyes were built, tested and cut. Alive became watchful.

A voice that holds you to a standard

The personality is not decoration. It is a written system the team and the AI both build from, so every surface sounds like one being. Jobi is an accountability partner, not a cheerleader. It expects you to show up, and it teaches you to stop needing it. Five qualities stay fixed in every line, steady, honest, loyal, invested, particular. What moves is the tone, which dials down as the moment gets heavier.

The easy version

“Still nothing?! Don't give up, you've got this!”

What Jobi says

“No reply yet. That's normal three days in. Want to line up the next two while we wait?”

Rejection is the moment the companion is won or lost. There the rules tighten. No exclamation points, no clichés, the dry wit goes quiet. Move the blame off the user, then point forward. Get that one moment wrong and the whole idea falls apart.

Measuring success

How we'll know it works

The whole bet is that Jobi lowers the emotional cost of the hunt, so the usual product metrics, time in app and applications sent, are the wrong ones to chase. Those are what the volume tools optimise for. The first real test is our closed beta, and three things will tell us whether the bet is paying off.

Do applications actually get finished?
Jobi runs the tracker, so it already knows what share of saved roles move to an Applied status instead of dying at Saved. We read that applied-to-saved ratio against the 92% who never finish an application elsewhere. If the friction is really gone, more of what people save turns into an actual apply.
Does the search feel less heavy?
A one-tap check-in built into Jobi, asked when someone joins the beta and again at the end, rating how in control and how burnt out they feel on a simple scale. We track the shift per person, not the headline number. If Jobi is doing its job, that line moves the right way.
Do people come back after a rejection?
When a user marks a role as Rejected, Jobi logs the moment and watches whether they open the extension again over the next few days. With no streaks or nudges pulling anyone back, a return after a hard no is a real signal the companion is holding.

What's next

Still building

Jobi is an ongoing project. The next milestone is a closed beta test session in July, putting the companion in front of real job-seekers to pressure-test the calls in this study, with an aim to release in August.

One feature still in design is the career plan: a coaching plan sized to the time you actually have, not a volume quota that burns you out. The hard part is holding the line, capping eager users at a few quality applications a day instead of the bigger number the product knows is harmful. So it is being built slowly, on purpose.