Japanese Insurers and Banks: Cutting Back to Rehire Managers

The series that this outlet has been covering, first on consulting ("BCG and McKinsey, AI against itself") and then on banks ("Morgan Stanley cuts 2,500, JPMorgan hires AI"), now advances to two sectors that operate the largest mass of administrative work on the planet: European insurance and Japanese banks. The anchor of this text is a reversal published by Fortune on September 12: Meta has begun to ask individual engineers within Applied AI to return to management roles, three years after the "year of efficiency" that streamlined the same layer. This is the first AI-first company to admit that the flat organization was not delivering.

The Cutting Numbers Today

Allianz confirmed in July that its Allianz Partners would eliminate between 1,500 and 1,800 positions in call centers and claims over the next 12 to 18 months. The unit has 22,600 employees, of which about 14,000 handle customer service and claims over the phone. Weeks prior, the same Allianz ranked as number 1 in the Evident AI Index for insurance, with 900 AI use cases in production. The reading is straightforward: those who lead in adoption metrics cut first and cut deeper.

Ergo, the primary arm of Munich Re, announced 1,000 fewer positions in Germany, attributing part of the move to efficiency gains via AI. Mizuho, in Japan, was tougher on paper but softer in execution: 5,000 clerical positions will cease to exist over ten years, a third of the bank's 15,000 administrative employees in the country. No one will be laid off, the company stated: the 5,000 will move to commercial and front-office roles. The plan comes with a ¥100 billion ($640 million) investment in AI by the end of the fiscal year 2028.

Where Meta Contradicts the Script

The above set-up draws an arrow: less administrative, more AI, more sales. What Meta did this week breaks the arrow diagonally. Mark Zuckerberg bet in 2023 that managerial layers could be cut because coordination tools (Slack, project tools, and then AI agents) would fill the gap. Three years and a new division later (Applied AI, with about 7,000 people), Meta admits that this architecture produces fragile coordination at the base and cognitive overload at the top. It is offering ICs the chance to return to management voluntarily.

This is not a total retreat, and Fortune was keen to note that. But it is the first significant empirical data against the thesis that "AI flattens the pyramid." The pyramid became flattened where AI was already running with weight, and the organization decided that it needs people to manage people. The criteria that Allianz, Ergo, and Mizuho are using to justify cuts now is the same criterion that Meta abandoned quietly.

Three Geographic Perspectives

Germany: German law on worker participation (Mitbestimmung) tends to turn call center cuts into expensive disputes. Allianz and Ergo are doing this with early retirement and severance because they have no legal alternative. The cost per head cut is higher than in the United States, and the return from AI needs to cover this cost before it turns into profit. Analysts at Berenberg estimate a payback of 3 to 4 years.

Japan: Mizuho is proceeding slowly because collective layoffs in Nikkei-listed banks still carry political consequences. The switch to redeployment is honest, but it creates another debt: the 5,000 who left the administrative side need to sell more consumer credit, treasury products, and foreign exchange to justify the new unit cost. Without nominal revenue growth, the operation turns into an accounting margin gain that evaporates the following year.

Brazil: Shared services hubs in São Paulo, Campinas, and Curitiba operate for American and European clients at a unit cost that only makes sense as long as the back office exists. If Allianz and Ergo cut in-house call centers, the next step is to cut outsourcing. Concentrix, Teleperformance, and Atento have direct exposure to the money that is leaving.

The Counterargument That Needs to Be Made

Jamie Dimon of JPMorgan stated to Bloomberg in October that AI will eliminate positions "but the total headcount may grow" because the company rehires elsewhere. Analysts from Morgan Stanley Research suggest that part of the cuts announced by insurers is recycling reduction programs that existed since 2023, now with an AI gloss. This critique does not invalidate the figures; it explains why they appear together in September. And it reminds us that a profitable hyperscaler financing AI capex is a different accounting creature than a VC lab burning cash.

The point that the next piece in this series will test: audit and legal, two sectors where cuts have not yet appeared in the public domain, and where internal memos are already circulating.