Sending AI cover letters to AI resume screeners

Welcome to Runtime! Today: why tech hiring is in a weird place right now thanks to the end of pandemic flexibility and AI tools, Microsoft rolls out its plan for agentic AI, and the latest funding rounds in enterprise tech.

(Was this email forwarded to you? Sign up here to get Runtime each week.)

Artificial all the way down

It's the best of times and the worst of times for people looking for tech jobs or tech candidates. The economy is growing, with plenty of new jobs to go around. However, the competition is brutal in the tech industry.

Since January 2024, at least 933 tech company layoffs have impacted 222,026 people, so more people are looking for new jobs. In a new story on Runtime, Esther Schindler examined how remote work policies and AI tools are changing the employment equation.

Job hunting is never fun, but it’s disheartening to see that “100 people have applied for this job” on LinkedIn for a position that has been open for only an hour.
The situation is just as frustrating for the hiring managers, who are inundated by applicants for each job requisition.
For example, one tech lead interviewed for this story received more than 2,000 applications for a mid-level software engineer position.

There has been a pretty monumental shift in the difficulty of recruiting, said Ross, who leads a small climate tech startup in California. “We’re getting an absolute deluge of candidates due to some combination of AI and desperation,” he said.

In 2020 and 2021, Ross got between 30 and 80 applications for job openings that he was tasked to fill as hiring manager, such as front-end engineer or data scientist.
Over the last year, he received anywhere from 400 to 1,000 applications for each position. That’s despite posting openings for jobs with specific technical skills, such as data scientists who have experience in developing carbon and hydrology models.
One reason HR inboxes are filling up is lifestyle changes due to “return to the office” policies at companies like Amazon, where employees are expected to work in the office five days a week beginning in January, following years of pandemic-related flexibility.

But applying for a job can be a soul-crushing experience, like shouting into a void, and it’s time-consuming.

One study found that applications take an average of five minutes and 51 clicks.
That doesn’t count HR departments that respond to the deluge by shifting more work onto applicants such as programming quizzes, essay questions, and unattended video interviews before you hit Submit.
That experience makes it easier to understand why people are motivated to use automated AI tools to apply for jobs or turn to ChatGPT to write resumes and cover letters.

Yet, good hiring managers — the ones whom applicants long to work for — are overwhelmed. In addition to the 400 resumes he received on his last hiring run for a staff platform engineer, Jonathan Eunice, director of Platform and Security at 3Play Media, said people contacted him on Slack and LinkedIn.

“It was overwhelming,” Eunice said. “There is no sane way to read 300 or 400 resumes and truly give them care.”
The HR industry offers tools to automate the triage and suggest which candidates to reject in so-called applicant tracking systems (ATSs).
According to a 2021 Accenture survey, more than 90% of employers used a Recruiting Management System to filter or rank candidates’ skills.
However, experienced job seekers recognized the buzzword bingo game long ago and responded accordingly, reinforcing a race to the bottom supplemented by AI hiring tools for both applicants and hiring managers.

Gartner’s 2025 HR technology trend predictions include AI-driven recruitment, personalized employee experiences, and advanced analytics. “AI-powered candidate assessments can infer a candidate’s cognitive traits, behaviors, emotions, or personality. AI also aids in sourcing and ranking candidates and job fit scoring,” its report said.

But the logical extension of that practice is humans using AI tools to fill in job applications that an AI evaluates.
Nobody we talked to for this story believes that’s the ideal path, except perhaps the tool vendors.

Read the rest of the full story on Runtime here.

I'm running out of agent puns

Microsoft took its AI show on the road to London this week, rolling out new capabilities in Copilot Studio that allow customers to build AI agents, the hottest fashion accessory of the fall. It also announced new agentic capabilities for Dynamics 365, which will allow its ERP software to track components in customers' supply chains and allow its CRM software to qualify sales leads.

A parade of embargoed stories positioned Microsoft's news as a response to Salesforce's agentic AI bonanza last month, but Microsoft actually introduced agentic capabilities in Copilot Studio at Build last May, and even used the word "agents" in that press release. As of Monday, however, those capabilities are now available as a public preview, rather than a private preview.

It's certainly true that Microsoft has invested a ton of time and money in promoting its Copilot brand as the face of its AI strategy, but it has been talking about the potential of AI agents — which are supposed to be able to handle complicated tasks in sequence and in response to unstructured data — for quite some time. The question of whether enterprises and their customers actually find agents worthy of the all-out blitz that almost all enterprise software vendors have called during the last several months remains open.

Enterprise funding

Lightmatter raised $400 million in Series D funding to continue its work on optical interconnects for AI datacenters, in hopes of improving the speed at which GPUs can exchange data.

Zip landed $190 million in Series D funding for its procurement software, which helps businesses purchase the materials and services they need more quickly and cheaply.

Socket scored $40 million in Series B funding as it builds out a security tool for monitoring vulnerabilities in open-source software that companies use to build their own apps.

Everstage raised $30 million in Series B funding for what we're apparently calling "RevOps," or software that helps companies pay commissions to their sales teams.

CrewAI landed $18 million in seed and Series A funding to help companies build their own AI agents.

Oxla secured $11 million in seed funding for its data processing engine, which will compete with engines from companies like Snowflake and Databricks as businesses shift their data into open formats.

The Runtime roundup

The SEC will collect $7 million in fines from four enterprise tech companies that downplayed their exposure to the SolarWinds hack.

SAP raised its full-year guidance for cloud revenue and overall revenue coming off a strong quarter.

Cisco cut off access to one of its developer sites after an attacker stole customer data and offered it for sale to the underground hacking community.

AI models are very different from software source code, and there may not be a way to release a truly "open source" AI model, according to Redmonk's Stephen O'Grady.

Thanks for reading — see you Thursday!