
Post-pandemic hiring finds footing as AI transforms tasks, not jobs
Speaking at the OpenAI Forum, Dr Karin Kimbrough, Chief Economist at LinkedIn, said, "Right now, I would say the labour market's actually reflecting a lot of macro cyclical effects," rather than widespread AI-driven displacement.
She noted the hiring boom of 2021-2022 has since cooled, with companies largely holding onto talent and new hiring remaining cautious.
"The labour market's looking for normalcy," Kimbrough said. "There's more competition than there was before, but not so much that people can't find a job."
However, this overall stabilisation masks significant disparities. "If you are in what is considered like the knowledge worker space, it might be a lot more competitive for you right now to find the role that you want," she explained. By contrast, "The world is your oyster if you work in [retail, healthcare, or construction] industries."
Kimbrough attributes these trends to structural factors and post-pandemic realignments, not solely to AI. "We are seeing just a huge demand and increase in roles \[in healthcare]. So we're seeing the number of roles open up—it's not just that people are looking for a role and are able to find one."
While job loss due to AI remains a common fear, Kimbrough emphasised that, at this stage, "It's not elimination so much that we're seeing. It's more just people are adjusting, desperately trying to upskill so they can stay relevant."
What is being displaced, she explained, are specific tasks within roles.
"It's how they're spending their time in that role that is changing."
Ronnie Chatterji, Chief Economist at OpenAI, agreed, saying that workers are "rotating from certain tasks to other tasks" rather than seeing their roles eliminated entirely.
LinkedIn's data supports this, with increased emphasis on "AI literacy" - the ability to use AI tools effectively. "One of the fastest growing in-demand skills on the platform by employers is conflict resolution," Kimbrough added. "They're rising in tandem with the demand for AI literacy."
Surprisingly, it's workers in more disrupted sectors - such as communications and marketing - who are most aggressively updating their profiles to signal AI capabilities.
"They need to signal it because the last thing they want to do is be fighting over a job either because an AI could already do half of it or because someone else can do the other half... with AI."
For early-career professionals, the post-pandemic hiring slowdown has been particularly acute. "It's been materially harder for new grads in the last year or so to find a role," Kimbrough said. "Employers are looking at their talent roles being full. Again, no one has quit."
However, historical data offers some reassurance. "What we found... was that people catch up," she said. "In the first couple of years of their career, maybe they don't progress as fast, but eventually they catch up."
She advised graduates to focus on agility, a willingness to learn, and human skills like communication and collaboration. "Even more important [than AI literacy] is communication skills, collaboration skills... leadership."
Kimbrough also spoke about international labour markets, pointing to India as particularly dynamic: "A very young, well-educated, and very dynamic, large domestic economy." Yet she highlighted the challenge of making AI tools locally relevant, citing "accessibility of the digital divide and the relevance" as critical barriers to adoption.
Adoption also varies widely by sector. Education, for example, lags significantly behind industries like finance or healthcare in hiring AI-literate talent. "Even so, we see them growing at really rapid rates. Is itIt's just not at the same rate."
Kimbrough revealed a shift in hiring strategies, noting that employers have reduced backfilling of vacated roles. Instead, they are prioritising new roles, particularly those related to AI.
"If you are working in the AI space... you are just in a position of choice," she said, listing roles such as AI engineer, consultant, and researcher as the fastest-growing on LinkedIn. However, she noted that such an opportunity is not uniform: "If you are in other roles... it's been a little bit more challenging."
The evolving job landscape also means career paths are becoming less linear. "It's far more organic, it is far more skills-based," Kimbrough said. "AI is so powerful, it's going to allow you to pivot into many more options for roles."
Even at the executive level, LinkedIn data shows leaders now emerge from more diverse backgrounds. "So it's okay to have a broader base," she said. "If the thing you have your heart set on isn't working, go do something else and just get started."
The discussion painted a nuanced picture: a stabilising labour market shaped more by macroeconomic tides than AI disruption, at least for now. But as AI adoption deepens, agility, skills signalling, and adaptability may well define success across all sectors.

Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles


Techday NZ
20 hours ago
- Techday NZ
AI bots drive 80% of bot traffic, straining web resources
Fastly has published its Q2 2025 Threat Insights Report, which documents considerable changes in the sources and impact of automated web traffic, highlighting the dominance of AI crawlers and the emergence of notable regional trends. AI crawler surge The report, covering activity from mid-April to mid-July 2025, identifies that AI crawlers now constitute almost 80% of all AI bot traffic. Meta is responsible for more than half of this figure, significantly surpassing Google and OpenAI in total AI crawling activity. According to Fastly, Meta bots generate 52% of observed AI crawler interactions, while Google and OpenAI represent 23% and 20% respectively. Fetcher bots, which access website content in response to user prompts - including those employed by ChatGPT and Perplexity - have led to exceptional real-time request rates. In some instances, fetcher request volumes have reached over 39,000 requests per minute. This phenomenon is noted as placing considerable strain on web infrastructure, increasing bandwidth usage, and overwhelming servers, a scenario that mirrors distributed denial-of-service attacks, though not motivated by malicious intent. Geographic concentration North America receives a disproportionate share of AI crawler traffic, accounting for almost 90% of such interactions, leaving a relatively minor portion for Europe, Asia, and Latin America. This imbalance raises concerns over the geographic bias in datasets used to train large language models, and whether this bias could shape the neutrality and fairness of AI-generated outputs in the future. The findings build on Fastly's Q1 2025 observations, which indicated automated bot activity represented 37% of network traffic. While volume was previously the chief concern, Fastly's latest data suggests that the current challenge lies in understanding the evolving complexity of bot-driven activity, particularly regarding AI-generated content scraping and high-frequency access patterns. Industry-wide implications Fastly's research, compiled from an analysis of 6.5 trillion monthly requests across its security solutions, presents a comprehensive overview of how AI bots are affecting a range of industries, including eCommerce, media and entertainment, financial services, and technology. Commerce, media, and high-tech sectors face the highest incidence of content scraping, which is largely undertaken for training AI models. ChatGPT in particular is cited as driving the most real-time website traffic among fetcher bots, accounting for 98% of related requests. Fastly also notes that a continuing lack of bot verification standards makes it difficult for security teams to distinguish between legitimate automation and attempts at impersonation. According to the report, this gap creates risks for operational resilience and poses challenges for detecting and managing unverified automation traffic. Verification and visibility "AI Bots are reshaping how the internet is accessed and experienced, introducing new complexities for digital platforms," said Arun Kumar, Senior Security Researcher at Fastly. "Whether scraping for training data or delivering real-time responses, these bots create new challenges for visibility, control, and cost. You can't secure what you can't see, and without clear verification standards, AI-driven automation risks are becoming a blind spot for digital teams. Businesses need the tools and insights to manage automated traffic with the same precision and urgency as any other infrastructure or security risk." The report recommends increased transparency in bot verification, more explicit identification by bot operators, and refined management strategies for handling automated traffic. In the absence of such measures, organisations may encounter rising levels of unaccounted-for automation, difficulties in attributing online activity, and escalating infrastructure expenses.


Techday NZ
a day ago
- Techday NZ
LambdaTest debuts AI tool platform for rapid validation
LambdaTest has announced the private beta launch of its Agent-to-Agent Testing platform, developed to validate and assess AI agents. The platform is targeting enterprises that increasingly deploy AI agents to support customer experiences and operations, as organisations seek reliable automated tools designed to handle the complex nature of AI-powered systems. Need for new testing approaches AI agents interact dynamically with both users and systems, resulting in unpredictability that challenges traditional software testing methods. Ensuring reliability and performance in these contexts has proven difficult, particularly as conventional testing tools fall short when the behaviour of AI systems cannot be easily anticipated in advance. LambdaTest's Agent-to-Agent Testing aims to address these challenges by using a multi-agent system that leverages large language models for rigorous evaluation. The platform is designed to facilitate the validation of areas such as conversation flows, intent recognition, tone consistency and complex reasoning in AI agents. Multi-modal analysis and broader coverage Teams using the platform can upload requirement documents in various formats, including text, images, audio, and video. The system performs multi-modal analysis to automatically generate test scenarios, aiming to simulate real-world circumstances that could pose challenges for the AI agent under test. Each generated scenario includes validation criteria and expected responses. These are evaluated within HyperExecute, LambdaTest's test orchestration cloud, which reportedly delivers up to 70% faster test execution when compared to standard automation grids. The platform also tracks metrics such as bias, completeness, and hallucinations, enabling teams to assess the overall quality of AI agent performance. Integration of agentic AI and GenAI Agent-to-Agent Testing incorporates both agentic AI and generative AI technologies to generate real-world scenarios, such as verification of personality tone in agents and data privacy considerations. The system executes these test cases with the goal of providing more diverse and extensive coverage compared to existing tools. Unlike single-agent systems, LambdaTest's approach employs multiple large language models. These support deeper reasoning and the generation of more comprehensive test suites, aiming for detailed validation of various AI application behaviours. "Every AI agent you deploy is unique, and that's both its greatest strength and its biggest risk! As AI applications become more complex, traditional testing approaches simply can't keep up with the dynamic nature of AI agents. Our Agent-to-Agent Testing platform thinks like a real user, generating smart, context-aware test scenarios that mimic real-world situations your AI might struggle with. Each test comes with clear validation checkpoints and the responses we'd expect to see," said Asad Khan, CEO and Co-Founder at LambdaTest. Impacts on testing speed and team resources LambdaTest says that businesses adopting Agent-to-Agent Testing will benefit from more rapid test creation, improved evaluation of AI agents, and decreased testing cycles. The company reports a five to ten-fold increase in test coverage through the platform's multi-agent system, providing a more detailed picture of how AI agents perform in practice. Integration with the HyperExecute system is designed to offer development teams fast feedback from test results, helping to reduce the interval between testing and product iteration. Automated processes also aim to reduce the reliance on manual quality assurance, with implications for cost efficiencies. The platform includes 15 different AI testing agents, covering areas such as security research and compliance validation. LambdaTest states that this is intended to ensure deployed AI agents meet requirements for robustness, security and reliability. The company's Agent-to-Agent Testing technology reflects ongoing efforts within the software testing sector to cope with the dynamic and evolving risks introduced by the increasing use of AI in business-critical systems.

RNZ News
2 days ago
- RNZ News
Book Critic: The implications of AI on writing
Pip Adam has been reading and thinking a lot about how AI could affect writing. She shares books that offer insights to AI and that show how important human writers are. Empire of AI: Dreams and Nightmares in Sam Altman's OpenAI - Karen Hao Blame it on the Rain (no more poetry 2025) - Hana Pera Aoake Show you're working out - Liz Breslin Hana Pera Aoake Photo: Supplied