The Economy Is Growing. Your Jobs Are Disappearing. Both Are True.

The Jobs Report That Confused Everyone

The February 2026 US jobs report, released on Friday February 6, created a communications problem for everyone who covers the economy. The headline number — 187,000 non-farm payroll additions, unemployment at 3.9%, average hourly earnings up 3.4% year-over-year — told a story of a resilient labour market in an economy growing at 2.7% annualised in Q4 2025. Stock markets closed higher. The Federal Reserve signalled no imminent rate changes.

The detailed breakdown of the same report told a different story — one that several economists and technology analysts spent the week following the report attempting to explain to business audiences who found the two stories impossible to hold simultaneously.

In the specific sectors where AI deployment had been most aggressive and most operationally mature — software quality assurance, content moderation, junior financial analysis, customer service management, entry-level legal support — employment was contracting. Not catastrophically, but clearly and consistently. The Bureau of Labor Statistics categories covering these roles showed aggregate job losses of 34,000 in February, against the broader gain of 187,000. The macro numbers obscure the sectoral reality. Both are correct.

Understanding both — and more importantly, understanding what they mean for workforce strategy — is the work that most boardrooms had not yet done by the end of February 2026.

The Sectoral Data Behind the Headlines

The organisations that published the most clarifying analysis of the February labour data were not government statistical agencies — their data is aggregate and lagged by design — but research teams at major financial institutions who track AI deployment alongside employment data for their clients.

JPMorgan’s Equity Research division published its “AI Employment Displacement Tracker” quarterly update the week following the jobs report. Its methodology: monitoring publicly-available job posting data alongside company announcements about AI deployment in specific functions, then tracking whether roles in those functions are being backfilled after natural attrition. The finding: organisations that had announced AI deployments in specific operational functions were backfilling vacancies in those functions at a rate 40 to 60% below pre-deployment levels. They were not eliminating roles en masse. They were simply not replacing roles that became vacant through natural turnover.

The aggregate effect of this pattern across a large number of organisations is visible in the sectoral employment data, even though no single company has made a dramatic announcement. Gradual non-replacement is harder to see than a 15,000-person layoff. It is not less consequential.

“The story the data tells is more subtle than ‘AI is replacing workers’ and more consequential than ‘the labour market is fine.’ What we are seeing is selective non-replacement concentrated in specific categories of knowledge work — content review, junior analysis, templated documentation, standard customer interactions. The macro numbers look stable because those roles are a small percentage of total employment. But within the organisations where they concentrate, the change is significant.”

— Dr. Daron Acemoglu, Institute Professor, MIT Economics, speaking at the Brookings Institution, February 24, 2026

Goldman Sachs published a complementary analysis focused on the relationship between GDP growth and employment growth. Its finding: labour productivity — output per worker — had increased by 1.9% in Q4 2025, the highest quarterly reading since 2002. The source of the productivity increase was not primarily capital investment in the traditional sense. It was AI deployment in knowledge work functions. The economy was producing more output per worker because, in specific sectors, AI was augmenting worker productivity significantly. That is good for GDP. It is complicated for the workers whose productivity was augmented — because in some categories, “augmented” is transitioning to “replaced.”

The Two Economies Running in Parallel

The most useful framework for understanding the February 2026 employment picture came not from economics but from business strategy analysis. Harvard Business Review published a widely-cited feature in its February issue titled “The Bifurcated Economy.”

The argument: the US economy is running two parallel labour markets that are increasingly decoupled from each other. The first is the broad economy: resilient, growing, characterised by tight labour markets in physical-world services (healthcare, construction, skilled trades, personal services) and in the kinds of complex knowledge work that require genuine expertise, relationship management, or creative judgment. The second is the specific economy of routine knowledge work: content processing, standard analysis, templated communication, basic support functions. In this second economy, the supply of AI systems capable of performing the work is growing faster than demand is being created for the affected workers.

The macro statistics average these two economies together. The result is a headline number that looks reassuring and a sectoral reality that looks significantly more complicated.

“I am not predicting mass unemployment. The labour market has absorbed technology transitions before and will absorb this one. What I am saying is that the specific workers in the specific roles being structurally changed are not going to be protected by good aggregate employment statistics. They are in a different economy from the one the headline numbers describe.”

— US Secretary of Labor, speaking at the National Economic Council, February 19, 2026

The Political Economy Problem

The political dimension of this data is significant and was immediately apparent in the week following the jobs report. In an election cycle, a GDP number of 2.7% and unemployment of 3.9% are good news. The political incentive is to lead with the good news and address the sectoral reality only when the political cost of ignoring it exceeds the cost of acknowledging it.

That calculation is shifting. The workers affected by AI displacement in knowledge work are disproportionately in specific demographic categories — urban, college-educated, professionally experienced — that have historically been less likely to be mobilised around automation-related concerns than manufacturing workers. February 2026 showed early signs that this pattern was changing. A series of viral posts from laid-off content policy managers, displaced junior financial analysts, and former customer service professionals with graduate degrees were gaining significant traction in media coverage and policy circles.

The legislative response was tentative. A bipartisan Senate bill requiring large enterprises to report AI deployment and its employment effects on a quarterly basis to the Department of Labor was introduced in the week following the jobs report. Its introduction signals that the political conversation has begun in earnest, regardless of whether this particular bill advances.

What Organisations Are Doing — and Not Doing

The range of organisational responses to the employment dynamics of February 2026 falls roughly into three categories, each representing a distinct strategic position.

The first is active restructuring — the Meta model. Organisations in this category are using AI deployment as an explicit opportunity to reduce headcount in specific functions, making the restructuring visible and defending it on economic grounds. The market has rewarded this approach. The regulatory and reputational risks are not yet material.

The second is quiet non-replacement — the most common approach by a significant margin. Organisations in this category are not announcing restructuring programs. They are simply not backfilling roles vacated through natural attrition in functions where AI handles the work adequately. The aggregate effect is employment reduction without a specific decision point that attracts public attention.

The third — and smallest — category is organisations explicitly investing in workforce transition: reskilling programs for workers in affected categories, active redeployment of people whose roles are changing rather than elimination, and public commitments to specific workforce transition metrics. The organisations in this category include a cluster of European multinationals responding to regulatory pressure and a smaller number of US organisations whose leadership has made workforce transition a stated strategic commitment.

The Competitive Arithmetic Behind the Non-Replacement Strategy

The economic logic of quiet non-replacement is clear and, in the short term, compelling. A role that generates $150,000 in annual compensation and benefits, that can be replaced by an AI system costing $2,000 per year in API costs and $30,000 in implementation, represents a $118,000 annual cost reduction once implemented. Multiply by the number of roles in the affected category and the business case is straightforward.

The risks are less immediate but not less real. Organisations that reduce their knowledge work capacity without simultaneously investing in the human capabilities that AI cannot replace — complex judgment, relationship management, novel problem-solving, organisational memory — are not optimising. They are hollowing out. The distinction becomes visible when the AI systems encounter situations outside their training distribution, when regulatory or reputational issues require nuanced human response, or when competitive advantage requires capabilities that cannot be automated.

The organisations doing this well are the ones making both moves simultaneously: reducing headcount in genuinely automatable categories while investing in the capabilities that make the remaining human workforce more valuable, not just cheaper to maintain.

ZHC Implication: The Honest Workforce Assessment Is the Strategy

For Zero Human Company strategy, the February 2026 employment picture clarifies a requirement that many organisations have been deferring: the honest assessment of which functions in your organisation are in the first economy — requiring human judgment, expertise, and relationship — and which are in the second, where AI systems are now capable of performing the work at or above current human baseline.

The organisations that have done this assessment and acted on it are not the ones that made headline announcements. They are the ones that have been quietly restructuring their workforce composition, redeploying people from affected functions to higher-value roles, and building the AI operational infrastructure that makes autonomous operations sustainable at scale.

The organisations that have not done this assessment are carrying structural cost premiums — paying for human labour in categories where AI performs the work — while their competitors are redeploying that capital toward the capabilities that actually differentiate them.

The economy is growing. That is true. The specific roles that constitute a meaningful portion of your current workforce are undergoing structural change. That is also true. Holding both of these realities simultaneously — and making strategic decisions that reflect both — is the requirement. The organisations that can do it are positioned for the next decade of competitive advantage. The ones that cannot are exposed to disruption they are not currently pricing into their planning.