Vantage Point is a new monthly rotating column featuring timely insights from Wharton faculty.
Since 1482, when Leonardo da Vinci sent the first cover letter and resume to the Duke of Milan, candidates have — often begrudgingly — been including cover letters in their applications to potential employers. But like the Duke’s reign in 1500, the reign of the cover letter is coming to an end.
A French invasion brought down the Duke. Today, generative AI is doing the dethroning.
As a professor in the Business Economics and Public Policy department at Wharton, I have spent the last decade and a half studying markets. The labor market is a prime example of what I call a “hidden market,” which operates without relying exclusively on prices. Firms don’t just lower their offered salary until one person applies to the job. Instead they have to sort through many candidates — sometimes 250 or more — and find the best person for the role. (I discuss these labor market dynamics at length in my new book on hidden markets called Lucky by Design: The Hidden Economics You Need to Get More of What You Want.)
There is a major transformation happening in the labor market right now thanks to generative AI. Large language models are changing how candidates can effectively signal their quality, and their interest in a job, to potential employers. This is making traditional signals like cover letters obsolete.
To understand the dynamics, it is useful to remember the two signals cover letters used to send.
The first was about the candidate’s quality. Strong candidates — those with excellent writing skills and impressive backgrounds — could more easily produce impressive cover letters. The second was about the candidate’s interest in a role. A detailed, tailored, thoughtful cover letter (e.g., one that tied a candidate’s background and skills to a particular job posting) signaled that a candidate was very interested in a role. Signaling interest is particularly important if a firm is unsure whether a candidate is serious about the position and willing to invest the time and energy required to succeed in the role in the long run.
Large language models are changing how candidates can effectively signal their quality, and their interest in a job, to potential employers.
When writing an excellent cover letter was difficult and took hours, only strong candidates who were very interested in a job would invest the time to optimally tailor them.
Enter large language models like ChatGPT and anyone can write, in a matter of minutes, a deeply thoughtful cover letter that might have previously taken an impressive candidate hours. This innovation has dramatically lowered the cost of the signal, substantially degrading its ability to showcase that a candidate is a good fit for a job.
Recent data from Freelancer.com confirms this dynamic. A new working paper by economists Jingyi Cui, Gabriel Dias, and Justin Ye analyzes over 5 million cover letters submitted to 100,000 jobs during a period when the platform introduced an AI-powered cover letter generator for some of its users.
The researchers found that those with access to generative AI wrote stronger cover letters that were better tailored to the job postings. The candidates with access to the tool secured more interviews.
But the researchers found that the AI tool substantially decreased the value of the cover letter as a signal on the platform. Before the AI tool, tailored cover letters were strongly predictive of higher interview rates and job offers. After the tool was introduced, having written a good cover letter was much less predictive of being given an interview or being hired.
Because of the AI tool, a tailored cover letter became a prerequisite rather than a differentiator.
Finding More Costly Signals
If AI is making cover letters obsolete, how should firms respond? And how can applicants cut through the noise?
Firms need to rely on different signals — ones that cannot be easily replicated by AI — to assess candidate quality and candidate interest in a role. Candidates need to learn to send these signals.
One way for a worker to signal quality will be leveraging personal connections, including recommendations from former employers.
In research I’ve done with University of Michigan’s Sara Heller, randomly providing young workers with a short letter of recommendation from the supervisor at their summer job — a letter they could share with potential future employers — had big impacts on their labor market outcomes.
Firms need to rely on different signals — ones that cannot be easily replicated by AI — to assess candidate quality and candidate interest in a role. Candidates need to learn to send these signals.
Just having access to a letter of recommendation increased youth employment by 4.5% over the next year and increased earnings over four years by 4.9%. We estimate that the effect of including the letter in a job application is likely much larger: on the order of 10% to 15% for employment and earnings.
The key difference between a cover letter and a letter of recommendation is what they signal. A recommendation signals that a former employer is willing to vouch for a candidate and their skills. It doesn’t matter whether the former employer uses AI to write a cover letter — what matters is that they stand by its content and are willing to be a reference.
This logic is backed up by the data from Freelancer.com. The researchers mentioned above found that in the presence of many AI cover letters, employers started relying more heavily on other signals about worker quality like prior work history and reputation.
Personal connections — again, the kind that are not replicable by AI — can also allow candidates to signal their strong interest in a job. Candidates who invest time and energy networking with current employees — perhaps by attending industry events, information sessions, or coffee chats — can show that they are particularly keen on working at that firm.
A candidate can now generate and send hundreds of tailored cover letters at the push of a button (or two). But even in a world of AI, if they spend an hour taking someone out for coffee, that’s an hour that they cannot use any other way. It remains a costly signal of interest.
These in-personal signals may become even more valuable in our increasingly virtual world. When Zoom interviews are commonplace, volunteering to do an interview in person has becoming a costly signal of interest that a candidate can send to a potential employer.
Like the leaders of great world cities, hidden markets change. New signals take the throne. In these periods, you just need to be dynamic and be willing to adapt to succeed, which I call getting lucky by design.
