Lara Santos | Product Design

Designing an AI-powered ATS to improve retention

Welcome to the Jungle is a two-sided marketplace connecting companies and candidates. While the marketplace performed well for job visibility, companies used us as one of many other tools for hiring like ATSs, for this reason, retention was limited by fragmented hiring workflows outside the platform.

The business goal: Evolve our B2B offering from marketplace to a one-stop hiring solution, to increase our value, stickiness and long-term retention. Starting by building an ATS with the most up to date technology.

I designed an AI-powered ATS, leading discovery and design from concept to beta launch. Working with a cross-functional team while owning:

Research and opportunity framing
End-to-end UX/UI design
Design decisions around AI-assisted workflows
Learning from beta users and iterating

Discovery & research

I conducted qualitative research with hiring teams across different company sizes to understand their existing hiring toolkits and workflows.

We didn’t just focus on feature gaps, we learned about:

The problems hiring teams were facing
Where were they losing time
What tools were their using today and what they’ve stop using
What was their workflow like
Who uses the tools and what do they need
How much do they trust and use AI across their hiring process

These conversations revealed that teams:

Were overwhelmed by high volumes of applications
Spent disproportionate time sourcing and triaging, rather than evaluating quality candidates
Struggled to get back to candidates quickly, creating a poor candidate experience
Did not need feature-heavy ATSs but rather a simple tool that’s intuitive and easy for everyone

Sharing learnings with the team after every conversation helped build knowledge and excitement. This was really important when working with a brand new global team.

This helped us get a good understanding of the problem space and define our mission:

How can we help hiring teams move faster?

From job creation to confident decision-making

How can we help hiring teams move faster?

From job creation to confident decision-making

Product strategy & decisions

Based on the insights, and the nature of the beta we prioritised the early stages of the hiring process like creating a job and screening applications. This meant making some trade-offs along the away, like deciding to focus on speed over personalisation, testing bold ideas over a comprehensive cover of features and approaching AI as a tool for designing experiences not a solution in itself.

Making it clear that 6 months, we won’t be competing with bigger, more complex tools used by mid-enterprise companies but we would be able to offer a great solution and test our ideas with smaller companies with less complex hiring needs.

We focused on three main solutions:

Job creation, making it faster and less repetitive.

Application screening, speeding up the process while ensuring it’s safe and fair for candidates.

Analytics, giving users straight forward access to their analytics.

Opportunities we identified but weren't prioritised:

Smooth and seamless collaboration - Because we focused on smaller companies with less people on their team

Scorecards for faster decision making - Because it only be needed further along the hiring process

Automated scheduling - Because it had dependencies with other features like user roles and candidate messaging

Product strategy & decisions

We focused on three main solutions:

Job creation, making it faster and less repetitive.

Application screening, speeding up the process while ensuring it’s safe and fair for candidates.

Analytics, giving users straight forward access to their analytics.

Opportunities we identified but weren't prioritised:

Smooth and seamless collaboration - Because we focused on smaller companies with less people on their team

Scorecards for faster decision making - Because it only be needed further along the hiring process

Automated scheduling - Because it had dependencies with other features like user roles and candidate messaging

Solutions

In 2 months after launching the beta, we were able to onboard 12 daily users and gather bi-weekly feedback with the main 3 features we focused on plus other basic pieces to make adaption possible.

Before being able to offer the product commercially, we wanted to build product market fit, understand our current strengths and weakness to set the product up for success.

Making job creation faster

We set out to enable hiring teams to go from 0 → live job in minutes using AI to create a draft and an easy way to generate descriptions from similar roles.

This solution reduced cognitive load and setup friction. It also enabled us to align job creation with our marketplace insights, nudging recruiters to create applications that are candidate friendly and prioritise intention and transparency.

Impact so far
Users highlighted this feature as the most useful in the beta product.
They reported feeling like the time taken to get a job live decreased by at least 30% compared to their previous tool.

Loading state

Result/Editing state

AI chat window

Empty state

Making job creation faster

We set out to enable hiring teams to go from 0 → live job in minutes using AI to create a draft and an easy way to generate descriptions from similar roles.

Loading state

Result/Editing state

Empty state

Making job creation faster

We set out to enable hiring teams to go from 0 → live job in minutes using AI to create a draft and an easy way to generate descriptions from similar roles.

Loading state

Result/Editing state

AI chat window

Empty state

Automating application screening

This feature allows recruiters to set 5 criteria per job, a criteria could be anything from “5 years of startup experience” to “Product impact in Fintechs”. Once this is set, the AI can scan future applications against the criteria and show the recruiter how many it could find per application.

We wanted this feature to be transparent and keep decision making in the hands of the recruiter, while ensuring applicants have a fair process.

Impact so far
Users praised the transparency of this feature and the use of AI throughout the product.
“ I really like that the AI in the screening feature doesn’t feel like a blackbox”

But we haven't seen enough usage of this feature end-to-end.

AI suggestions

Filled in state

Pipeline view

Application view

Automating application screening

We wanted this feature to be transparent and keep decision making in the hands of the recruiter, while ensuring applicants have a fair process.

AI suggestions

Filled in state

Pipeline view

Application view

Impact so far
Users praised the transparency of this feature and the use of AI throughout the product.
“ I really like that the AI in the screening feature doesn’t feel like a blackbox”

But we haven't seen enough usage of this feature end-to-end.

Automating application screening

We wanted this feature to be transparent and keep decision making in the hands of the recruiter, while ensuring applicants have a fair process.

Impact so far
Users praised the transparency of this feature and the use of AI throughout the product.
“ I really like that the AI in the screening feature doesn’t feel like a blackbox”

But we haven't seen enough usage of this feature end-to-end.

AI suggestions

Filled in state

Pipeline view

Application view

Analytics

The vision for this feature was to take advantage of the AI chat to create an experience that allowed users to personalise their analytics page, prioritising the data they cared about and building their own dashboard.

We were unable to get this vision working for the beta launch. After finding lots of limitations with the LMM structure, we found that we wouldn't have enough time to get this feature to a good standard. We still launched a super basic version that could allows us to learn if users would find this feature interesting and what type of questions they would ask.

Impact so far
Despite not being able to launch the full experience, we saw a lot of interest when testing this feature with users. Which we're using to develop it for future improvements.

Empty state (vision)

Basic analytics chat

Filled in state (vision)

Analytics

Empty state (vision)

Filled in state (vision)

Impact so far
Despite not being able to launch the full experience, we saw a lot of interest when testing this feature with users. Which we're using to develop it for future improvements.

Analytics

Impact so far
Despite not being able to launch the full experience, we saw a lot of interest when testing this feature with users. Which we're using to develop it for future improvements.

Empty state (vision)

Basic analytics chat

Filled in state (vision)

We're still learning and building…

The beta is live, we're gathering insights and working towards improvements as you're reading this! If you want to learn more about the process and what we're currently working on, let's chat.

Back to homepage

Lara

Santos

Thanks for having a look!

What next?

Next project

£5,000 - £15,000

From £3.52 per month

Pet damage cover

Cover your stuff against accidental damage by your pets.

Add to Policy