Siri, Find Me a New Job
Structuring the complexities of simplicity
Technically, a simple enough question.
Wouldn’t it be great if we had agents who just handled the whole “next job” trials?
What if we could just automate all the dehumanizing hoops we have to jump through and find the perfect next job exactly when we need it, and move forward with our lives? Maybe ignorance is bliss, and we don’t need to know which company has declined us with a generic “unfortunately for you, we were so lucky and got loads of applications, and you didn’t match with what we were looking for”, only to see the job post resurface a couple of days later.
This post is about process considerations for building a zero-click job search engine.
Context
For most of us, job search is a high-friction/high-pain task. And if you have never experienced this, you should consider yourself lucky. You need to know the right people at the right time and make them aware that you could be the next perfect hire for the role. But how could you? The paradox is that the more we are connected, the more we are separated. The more people we have in our network on LinkedIn, the more our friends’ cries for help are drowned out by high-engagement success stories.
How would such an agent work, and what friction points are on that agent's journey?
Or in other words, what background workflows need to exist so we can just ask the question and find a fun and lucrative new career opportunity?
Since the dawn of time, I have seen influencers on LinkedIn promote a process like this (not necessarily in that order):
Networking Outreach (3-9 months)
Skill Development (1-24 months)
Resume Optimization (1 month)
LinkedIn Profile Optimization (1-3 months)
Job Description Analysis (1 month)
Personal Branding (9-18 months)
Cover Letter Crafting (1 month)
Interview Preparation (1 month)
Salary Negotiation Strategy (2 months)
The problem with this approach is (a) that it relies heavily on human-in-the-loop, and (2) can’t easily be done as a zero-click execution. Also, you might have noticed that I have added time frames to the list. The reason here is that while it’s easily possible to update your LinkedIn profile, developing a personal brand will take a long time.
If you want to learn a new skill, depending on the complexity, this might also take a long time.
This post is not a technical article or implementation. It’s also not about what can go wrong, but instead what we need to do to get this right.
Yet I also want to highlight what the implications would be if we get it to work.
Zero-click Job Search
In general, I’d think this could be a simplified basic flow:
A digital twin of ourself that knows our preferences (location, salary, industry, etc), skills, executes interviews (at least first-round) on our behalf, and can also negotiate salaries.
A Planner Agent that structures the work that needs to be done for the state of a current application and creates the right context for the worker agents.
The Search Agent scrapes news sites, API’s and websites daily to build, in regular intervals, a database of all currently truly open and likely upcoming roles. Alternatively, the agent scrapes precurated lists of career websites on demand.
A hiring manager identifier who searches out the hiring manager and their respective hiring preferences.
An ATS-optimizer Agent for our CV that allows the CV to pass automated tracking systems’ keyword match.
A matching algorithm between a job opening (incl. the company) and a digital twin
An application manager agent that files and tracks applications and organizes ongoing conversations (maybe including salary negotiations).
That said, I think it would be important to also identify solid touch points where the human (you) returns into the discussion if things go wrong.
So, let’s identify some friction points.
Ghost Jobs
About 1 out of 5 jobs listed on popular aggregation portals and career sites are considered “ghost jobs”. I.e., it is assumed that they don’t exist for a variety of reasons. While legitimate companies put out illegitimate roles where the budget has been closed or a hiring freeze has been announced after posting, observing a reduction of 8 hires per 10 postings in 2019 to only 4 hires per 10 postings is an atrocious statistic. Factoring this in and also that, especially startups post positions to show traction (not only in India), our agents should therefore also be able to predict the likelihood of the post being a “ghost” job.
The Shotgun
Admit it, you have done this. “Easy Apply” on LinkedIn or Indeed is an application approach that seems easy but never really works. Recently, in the nascent age of genAI you see startups (AIApply, LazyApply, JobCoPilot) promote products where you pay them and they apply to as many jobs as possible in a short time frame. This spamming increases the risk of your personal information being made available to non-trustworthy parties. But also, employers might associate this with unmotivated applicants. In my opinion, less is more. Therefore, the agent should only execute on high-probability opportunities where you have someone in your network, might even know the hiring manager, and have the right skills.
So the agent would not do the “shotgun”.
Personalization
Personalization is helpful when used for job/applicant matching, as it is for content recommendations. Based on the search parameters, we specify the title and city and get some job listings. We all know the dangers of filter bubbles. In my opinion, personalization, while heralded as the ultimate discriminator for decades, creates an environment that, by definition, does not create a shared social context. Quite the opposite. In most cases, recommendation algorithms do not understand context. Therefore, when we are building out our matching algorithm, I think that agents will be helpful as they understand context much better than a user-product or user-user-product, i.e., you are similar to user B and user B like product X, matching concept.
The Networker
One human aspect of the one-shot job search agent is to identify and connect with the hiring managers. Well, one of the key aspects of success at work is that you get along well with your boss. So, finding out the most about your future boss, within legal limits, might be a good approach. Further, hiring risk reduces if your future boss already has an impression of how you work, and one possible way this could be achieved is by connecting and then sharing related content.
But this is, in reality, a long-term strategy.
Timing differentials
When we utter those words, “Siri, find me a new job”, we task our virtual personal recruiter to go out and hunt. I’d argue that this hunt would be triggered in a moment of emotional distress. Maybe anger over a missed promotion or an argument with your boss. These are not good moments to start searching. Even if supported by such a system, finding the “right” next thing for you will still require patience and an understanding of your needs and wants 5-10 years into the future. Changing industries is incredibly hard; maybe even impossible.
In closing
We all know job hunting today feels much more like playing a game with unclear rules, inconsistent referees, and an ever-moving goalpost. And DEI, whatever you may think of it, hasn’t helped but even further muddyied the mud. The main idea behind a zero-click job search agent will only remove friction if we fully redesign the entire flow. To reframe the job search from a hustle-based, burnout-inducing process into a discovery journey where agents work with and for us would be a benefit. In my experience, while many people claim that careers can be planned, I don’t believe in it. In more than 20 years, I had never seen an HR department actively support the growth of employees by pushing for promotions. Of course, one could argue that this is also not really their task.
I think the main benefit is that such agents could support those switching industries, re-entering the workforce, or quietly exploring change while employed. And I hope it would push companies to clean up outdated hiring processes and standards, and ideally and guide them to rethink how they approach transparency and intent in hiring.
Of course, we’ll always need human touch-points. A machine can suggest a job, craft the application, and even prep for interviews, but the moment of yes (or no) still holds emotional weight. And so, the human must stay in the loop not just for decision-making, but for meaning-making.
In the end, maybe ignorance isn’t bliss. Maybe knowing true context, i.e., what’s real, what’s relevant, what’s worth our time, is indeed the first step toward something better.
A job search that finally starts to feel less like rejection... and more like alignment.
