Since roughly the second quarter of 2023 it has felt like the tech jobs market has been holding it's breath. Still paused, red faced and spluttering, lips turning slightly blue. The shift to remote work during the pandemic, the rabid hiring with remote as an option, then subsequent moves to thin the herd.

Return To Office (RTO) often as part of a 'gentleman's layoff' (pushing folk to jump before they're pushed) while underestimating how hard it is to stuff proven productivity and quality of life sauce back into bottles. The explosion of generative AI with advertised transformative productivity potential. All against the backdrop of global instability and a record number of 2024 global elections.

Job data over the past two years has been a mixed bag of good times and bad; 260,000 tech workers were laid off in 2023, with another 142,000 getting pink slips so far this year, according to Layoffs.fyi. At the same time, US unemployment data released last month showed unexpected growth overall for tech job listings and hiring, along with a marked shift in the kind of workers organizations need — AI talent is no longer at the top of the list.

Computer World, Lucas Mearian, 6th December

Some were just emerging blinking from tech degrees. Others, caught by the shake outs at various firms, are job hunting. A bit like emerging into an Elysium-esque landscape. Up top the world of work is imminently transforming. Everything is shiny, everything is world-beating, everything is boundless AI promise. We will barely need to work at all. The hardest decision will be our choice of leisure activities.

Meanwhile, down below (from the same article)...

Jason Hayman, a tech and data analyst at IT consultancy TEKsystems, which tracks tech employment across industries, said there has been momentum over the past three to four months, with demand for technologists rising 12%. “However, that hasn’t necessarily translated into hires,” he said.

Companies, Hayman said, are likely waiting for more market clarity before adding jobs. They’re still adjusting to post-pandemic over-hiring, high inflation, and external factors such as global conflicts and the recent US election. Additionally, no specific sector or region is showing clear trends, making predictions uncertain...

Talking to peers in various sectors that uncertainty is felt in guts. Hiring practices, themselves increasingly AI powered, are far harder to navigate. Ghost jobs continue to proliferate - jobs without a current role attached, often to test markets and collect candidate data. Third parties automating applications by reviewing adverts and using generative AI to match responses, overwhelming recruiters and HR departments. AI screening mechanisms, including AI generated first interviews and candidates sending avatars to interviews, or using LLMs to answer questions in real-time just off camera.

Piece work is flourishing, some of it to tune AI systems. Hundreds of adverts for generalists or specialists from firms like Stability AI and it's affiliates. Variable hourly rates to review chatbot outputs and improve with better expert responses.

That last part feels a lot like a child's cooking toy, where plastic cookies go in a slot at the top, then the door opens to reveal the 'baked' goods. Take them out and send them round again, ad nauseum.

Successfully converting expert insights and local knowledge into the machine equivalent of reformed steak, or chicken nuggets. Highly processed, lacking a lot of the original nutrients, peppered with surprise bits of fat and gristle.

What related roles exist in this environment? The answer is ones that are fairly precarious. Generative AI in various multi-modal forms has only been about at scale for 2 years. This of course excludes all the industries and professions that require actual physical inputs, local expertise, and other kinds of systems.

The trouble is folk are very blinded by the job replacement potential of generative AI with layered access to selectively automate work with 'agentic' systems.

The Challenges Ahead

Despite significant potential to transform human-machine collaboration and drive greater efficiency and business growth, agentic AI systems are still at a relatively early stage of development. Moreover, despite their greater powers of reasoning and execution, they do not remove traditional workforce management challenges; instead, they change them. Just as in traditional, human workforce settings, managers must still pay heed to issues of team composition and role selection, and they must set the right overall goals to ensure that agentic AI or hybrid teams can be successful. They must also carefully calibrate the conditions under which agentic AI systems can be trusted to make decisions and the circumstances in which human decision makers need to intervene.

Harvard Business Review, Mark Purdy, 12th December

Management consultancy bingo being played in god mode.

I feel deeply for brand new tech graduates for whom this all brewed while they were studying pre-existing ideas of what the market would need. It has a feel of prior big shifts where the relevance of studies bore little relationship to what the market really wanted, or thought it wanted.

How do I personally feel about this world and my place in it? Often teeth-grittingly frustrated, but doing the basics well for all of my clients. The swirly maelstrom of AI positioning and re-positioning bears no relation to the reality inside many organisations. Helping others to let out great big breaths, relax shoulders, and remove the smokey rose tinted glasses. Not a bad description of what I do for a living.

Helping folk to take that pause to survey the organisation. See what works and what doesn't. Be clear-eyed about the part this hyper-novel tech can play in required change. Advice so basic it might make your teeth ache.

We do need to gain substantial familiarity with AI limitations and dependencies. A lot of that is simply sound governance basics. My wonder at what these tools can do has been dampened by the over-selling in a rapid drive to get market share to overcome lack of moats (fundamental differentiating product characteristics) for internet-trained LLMs. The moats are mainly around supply chain dominance in power, materials, chips, compute, 'AI enabled' devices, skills, data, and PR power.

The 'wrappers' for GPTs (Transformer based chatbots) are user interfaces. access schemas, bespoke prompts, specific tuning with purpose focused data, configuration choices, reporting and monitoring elements, linked systems with other capabilities, and selective layers of additional elements. Perhaps using Retrieval Augmented Generation (RAG) to supplement training data with specialist knowledge, or adding real-time search. Perhaps inserting a generative AI element into pre-existing offerings.

All that leans heavily on the availability, stability, explainability, and monitoring capability for the foundation systems and vendors. Perhaps not using one of the biggest vendors, perhaps going with a fully open source offering, with a lot of the same challenges and arguably far scarcer skillsets.

Then there is the deathly skills shortage in most firms. Not enough people who can really pick the bits out of functional, non-functional, and feasibility questions when doing procurement.

Two-thirds (66%) of banking and financial markets CEOs surveyed said that the potential productivity gains from AI and automation are so great that they must accept the risks to stay competitive.

65% of financial institution leaders say that succeeding with AI will depend more on people's adoption than the technology itself and 60% recognize they are pushing for AI adoption more quickly than some might find comfortable.

Half (50%) of financial services CEOs surveyed say they are hiring for generative AI-related roles that did not exist last year and 53% indicate they are struggling to fill key technology positions.

PR Newswire on an IBM Study, 5th June

Pregnant pauses? Absolutely. Execs reportedly tipping into. 'Yes, that's still exciting, but show me the productivity money.' Unsure how most will respond if told they have to have newest agentic iterations of the thing they've been told for two years is ready to rock worlds.

So what about hiring? If folk are dealing with massive macro challenges globally, a workforce shifting slowly, a dearth of readily available people who have any realistic medium or longer-term bead on where this is going...

...not always the most fertile ground for enthusiastic hiring decisions.

That, in combination, is why my head currently feels like it is full of molasses. I'm not fretting too hard as this is a usual pre-epiphany feeling, when threads align and distil into some useful insights. Maybe Christmas will help with that. Goodness knows too much time in the rarified air of LinkedIn often doesn't. Time to ground things in operational reality.

With the US being the engine for this movement and the administration changes due to start in earnest in January, there is not much anyone can do to prompt that exhalation for markets. Will likely events in the next few years foster the kind of certainty that supports productive investment, or feed more wild speculation based on rolling disruption? Lots to muse on. Lots to tune out to get the day job done.

Pregnant pauses in politics, tech, and employment