OLX Jobs for Seekers
From a vision for job seekers to live AI tools and an agent that searches for you
Role
Lead Product Designer · 2025–2026
Team
Product Director, Head of Product, Head of Design, Engineering Director, Head of Engineering, a PM for the Jobs Hub, product designers in the feature teams
Outcome
70% of people who tested CV Check applied with the improved CV · 80% say Interview Practice gave them confidence · Vision adopted at the 2025 strategy offsite · Agentic concept shown at Prosus Investor Day 2026
Summary
Looking for a job on OLX meant scrolling through listings and applying without knowing what happened next. I led the vision for the job seeker experience. Two parts of it are now live, and the work ended with an agentic concept: a job search you don’t have to run yourself.
Looking for a job often feels like sending applications into a void. In 2025 I led the vision for what OLX Jobs could be for the people looking for work. Leadership adopted it, two parts of it are live today, and it ends with the concept on the cover: an agent that searches for you and only asks when something is worth a decision.
A vision for job seekers
OLX is known for blue-collar jobs. The business wanted to grow in white-collar and, even more, grey-collar roles, the skilled jobs in between. These people search differently and expect more help along the way.
I wrote the vision as a route to that goal, which is why leadership backed it at the 2025 strategy offsite. It had six parts:
Goal-based onboarding
A smarter dashboard
Tailored search and nudges
Interview practice by text, voice and video
Support and coaching after applying
“Boost my CV”, with extras such as badges and an intro video
Each part answered a problem seekers hit at a specific point in their search:
Not knowing where to start. Many opened the app with no clear idea of what to look for. Onboarding started from their goals, and Wildcard Jobs suggested roles they hadn’t considered.
Needing money soon. For many people the pressure is income now, so the vision added short-term and side jobs. Our research on flexible work showed that hours, pay and location have to be clear before people commit.
Fear of not being good enough. Fear of rejection stops people before they apply. A 2026 OLX Praca study later put a number on it: 71% of Gen Z job seekers feel strong anxiety while looking for work. CV review and interview practice were there to lower that barrier.
A CV that’s hard to get right. Presenting yourself is hard, and uploading the same files for every application adds friction every time.
Two problems we knew well stayed out on purpose: no news after applying, and ads without salary or hours. Both depend on employers, so a better seeker experience couldn’t fix them on its own.
What reached the product
Two parts of the vision are now live. Another product designer led their design and delivery. I supported her and kept each piece tied to the principles of the vision.
CV Check. Next to Apply, seekers see how well their CV fits the offer and why, and can generate an improved version where every change is explained. OLX reports that 70% of the people who tested it applied with the improved CV.
My concept: the fit and why, then an improved CV where every change comes with its reason.
Interview Practice. An AI recruiter asks questions based on the ad and the seeker’s CV, then gives feedback on answers, tone and confidence. One principle from the vision shipped almost word for word: sessions aren’t recorded and are never shared with employers. OLX reports that 80% of candidates say it gave them confidence.
In my concept, practice started in writing or by voice, with video later. For someone already anxious about interviews, the camera is the hardest part.
My concept started without a camera: a private start, a question about the job answered by voice, and coach feedback after each answer.
The figures come from OLX’s product pages. The screens above are my concepts in the current OLX design system, so the live versions differ in places.
How much guidance people want
The part I took into the product myself was the Jobs Hub, a home for the category. Until then, tapping Jobs on the OLX home screen opened a plain list of vacancies.
In January 2026 I tested a hub prototype in moderated interviews with six active job seekers: freelancers, students, people returning to work and experienced professionals. I built it in Lovable so it could double as the research tool, and each person rated how useful its twelve sections were. Six people show a direction rather than proof, and we read the scores that way.
Four sections stood out: Search, Saved Jobs, Wildcard Jobs and Job Seeker Tools. The hub leads with those and keeps the rest further down. Wildcard Jobs surprised me. It suggests less obvious roles based on skills and interests, like dog walking for an animal lover, and people rated it highly because it gave them hope that there was something out there they hadn’t thought of.
Wildcard Jobs suggests roles based on skills and interests, not career history. People rated it highly because it gave them hope.
Below search, a one-minute preferences step asks what, where, when and how far. The offer count on the button drops with each answer, so every question shows what it’s for.
Every answer narrows the offers, from 1 240 to 52, so each question shows what it’s for.
We launched the hub as an experiment to check three things at scale: which sections people used, whether Job Seeker Tools were worth building into the hub, and whether offers found through preferences got more applications than offers people found on their own. The primary metric was how many people went on to open a job ad.
It was our first test of an entry point for the hub, and of OLX Jobs as its own vertical. We didn’t have many alternatives then, and the category icon felt like the natural door: tapping Jobs opened the hub instead of the listings. Grey- and white-collar seekers stayed and used the help. Blue-collar seekers pushed back hard. They didn’t trust our filters, wanted their listings back, and many left.
We had picked the wrong door, so we rolled it back. The help wasn’t the problem; forcing it on people was. The next version keeps the Jobs icon going to the listings and invites people in from a section on the home screen.
The experiment sent everyone who tapped Jobs to the hub. The next version keeps the icon going to the listings and invites people in from the home screen.
An agent that searches for you
Even with the hub and the new tools, job search still meant the same manual work: scrolling listings, adjusting the CV for each offer, typing the same details into every form. At the same time, OLX wanted to show what a jobs marketplace could look like in the AI era at Prosus Investor Day in March 2026. I designed the concept, and the Product Director presented it there.
What interested me most was the repetition it could remove. What if a job seeker didn’t have to come back to OLX at all? You say once what you want, and the agent keeps searching, prepares applications and brings you only what’s worth a decision. Nothing is sent without your OK.
One message instead of another search: the match, why it fits, and an application that waits for your OK.
The hub experiment had shown me the limit: people accept help, but they won’t give up control of their search. So the agent is built around control.
Each moment below uses one of the four interaction models of the agentic experience (AX) framework: conversational, suggestive, assistive and automated. With the Head of Design, I turned the framework into a talk for OLX designers at the March 2026 design review, so teams had a shared direction they could apply in their own products. Three ideas were meant to stay with them:
Think beyond interfaces. Design how the AI behaves, how much it does on its own and how it interacts, as well as the screens and flows.
Design for human agency. The person stays at the centre, keeps control and understands what the agent is doing.
Think in orchestrated experiences. One experience where the interaction models work together, rather than separate AI features.
You say it once
Conversational
The agent shows what it understood before it acts. Any line can be corrected with a tap.
Instead of filters, one sentence. The agent plays back what it understood, and a wrong guess takes one tap to fix.
You decide how far it goes
Automated
Autonomy is the seeker’s setting, with limits they choose.
Some people want every application in their own hands, others want the agent to apply for them. So autonomy is the seeker’s setting, with limits they choose. It’s the Jobs Hub lesson again: the same help is welcome or unwanted depending on who gets it.
Your CV, without invented facts
Assistive
Only what’s true gets emphasised. After sending, the agent offers the next step.
An AI that rewrites a CV can easily invent things. This one only emphasises what’s true, flags what’s missing and lets the seeker decide. After sending, it offers the next step, practising for this interview, so the CV help hands over to Interview Practice.
Measuring something else
This changes what success means. The hub was measured like most marketplace features: job ad views, saves, applications and search engagement. If the agent works, views and searches go down, because people no longer need to search. So the concept needs different measures: how many suggestions people accept, how often an application gets a reply, and how quickly someone reaches an interview.
What I learned
The way in is part of the design. The same hub helped some seekers and drove others away. Help has to respect how people already search.
Principles travel further than screens. The screens changed in the feature teams, but “explain every change” and “your practice stays private” shipped almost word for word.
An agent needs its own definition of success. If the product works when people visit less, engagement metrics stop telling you anything useful.
Next project