AI-driven job cuts may deliver near-term savings, but the harder test for financial firms is whether they can develop the skills and roles their changing businesses will need. The gap between removing work and building what replaces it could surface later in lost expertise, weaker institutional memory and too few experienced employees. That risk is easy to miss when savings are immediate and the cost of rebuilding knowledge arrives more slowly.
American Express plans to reduce its global workforce by about 5% as it expands AI in customer service and commerce. Barclays is cutting roughly 5,000 jobs while introducing AI coding tools for developers. Those moves raise a strategic question: are the companies reshaping how they operate, or making existing businesses leaner? The distinction matters because adopting tools can change how work is done, while a durable shift in the business may require different teams, judgment and accountability.
New work requires different skills
Barclays CEO C.S. Venkatakrishnan has described AI as “the icing on the cake, not the cake,” placing it within a broader program of technology investment, simpler processes and organizational change. His framing suggests that AI's effect depends on the surrounding operating model: software may speed up tasks, but firms still need people to decide where it fits, check its output and handle exceptions. As software takes on more tasks, employers will need to reconsider which skills their teams require.
Some of those roles are already emerging. Box CEO Aaron Levie has cited AI architects, model evaluators and AI business automation engineers as examples at the cloud content management and file-sharing company. Bank hiring signals point in a similar direction: October data from PYMNTS Intelligence showed AI-related bank job postings up 49%. Mentions of agent orchestration rose 1,721%, while references to responsible AI and AI governance increased 657% and 394%, respectively. The figures indicate a sharper appetite for specialized capabilities, though job-posting growth alone doesn't show how many roles are being filled or whether the demand will persist.
Those figures suggest demand for people who can connect AI systems to business processes, set boundaries for automated agents and manage cases that systems cannot resolve. That work sits between technical implementation and business judgment. Firms need systems that fit the work they actually do, and they need staff who can recognize when automated output is incomplete or unsuitable. But demand for new expertise doesn't mean displaced employees can move directly into those positions. A developer whose routine coding work is automated may still need time and training to qualify for AI governance work.
Efficiency can obscure the transition
Companies are getting clearer about which tasks AI can absorb and which jobs may shrink. Their plans for the people affected are less evident. A headcount reduction can be counted quickly; building new expertise takes longer, and the result is harder to measure. This is more significant than it looks: financial firms depend on accumulated knowledge of customers, processes and past decisions, much of which isn't captured in formal procedures.
If roles disappear before workers can gain the skills needed for emerging jobs, firms may save costs now while building a shortage of talent for later. Similar technology transitions often require employees to learn new tools while continuing to keep existing operations running. That can make training a capacity question as much as a budget question: who has time to teach, and who carries the work while colleagues learn? If you're working in this space, the relevant signal isn't just how many jobs are cut; it's whether new responsibilities are being defined clearly enough for people to prepare for them.
The next test is capability
The measure of an AI transition will be whether companies can develop the capabilities their new operating models require, not simply how much work they remove. That means connecting hiring plans, training and workflow changes to the tasks the business expects people and systems to perform. Here's the thing: a company can install AI tools quickly and still be unprepared for the work of supervising them, improving their use and handling what falls outside their reach.
The next workforce challenge is turning new demand for AI expertise into a practical path for people whose current roles are changing. That path may involve new training, redesigned jobs or hiring for skills the firm doesn't yet have; the article's examples don't establish which approach these companies are taking. And yet, the direction of demand is visible in the roles and job postings described here. The unresolved question is whether firms can build that expertise while retaining enough experience to keep their businesses working. A transition on paper. The operational test comes later.

The post AI can eat jobs, but companies still have to build the workforce appeared first on Tearsheet.
