Skip to content
Agentic AI

Your First Job In Finance May Be Managing AI Agents. Goldman Calls It A 'Virtual Army' - Here Is What That Actually Means, And How To Get Good At It

On 8 October Goldman Sachs' Kevin Sneader told the Milken Institute Asia Summit in Singapore that the bank's new recruits no longer start with years of grunt work: 'When our young folks now start work, they're managing agents. They have to make the most of this virtual army that they've now got.' The data says he is describing an industry, not a firm. Bank job postings mentioning AI rose 49% this year to almost 140,000; mentions of 'agent orchestration' rose from 108 to 1,967. JPMorgan runs agents for hours at a time. HSBC is reportedly cutting up to 70% of its UK wealth advisers while hiring AI specialists. This is the plain-English guide: what an AI agent is, what 'managing' one involves day to day, the skills that matter, and what it means for graduates, career-changers and the firms training them.

AlchmAI Editorial12 min read

'Virtual army'

How Goldman's Kevin Sneader described the AI agents new recruits now manage from their first day (Milken Asia Summit, 8 October)

139,819

AI-related job postings at banks including JPMorgan, Citigroup and Capital One in 2026 - up 49%, per hiring analytics firm Draup

+1,721%

Rise in bank job postings mentioning 'agent orchestration', from 108 in 2025 to 1,967 this year

Up to 70%

Of HSBC's UK wealth advisers reportedly being cut by the end of October, as the bank hires AI specialists elsewhere

For a generation of bankers, the first two years were a rite of passage: building spreadsheets, formatting decks, checking numbers late into the night. Kevin Sneader, Goldman Sachs' president for Asia Pacific outside Japan, told the Milken Institute Asia Summit in Singapore on Thursday that this is over at his firm. 'When our young folks now start work, they're managing agents,' he said. 'They have to make the most of this virtual army that they've now got.' The routine administrative work is done by AI agents; the junior's job is to direct and check them. Sneader was candid about what that leaves unresolved - the role of middle managers, which he called a 'generational challenge for many, many segments' - and other panellists warned that automation may also reduce demand for entry-level analysts.

The hiring data shows it is not one firm's experiment. Draup, which tracks enterprise job postings, counted 139,819 AI-related roles at banks including JPMorgan Chase, Citigroup and Capital One this year, up 49%. Postings mentioning 'agent orchestration' - coordinating multiple AI agents - jumped from 108 to 1,967, which Draup's chief executive called arguably the hottest skill on Wall Street. Mentions of responsible AI rose 657% and AI governance 394%. JPMorgan is deploying agents that run autonomously for hours rather than minutes. In Britain, the Financial Times reported, via Reuters, that HSBC is cutting up to 70% of its UK wealth advisers and about half of related managers and specialists, while the bank has been hiring AI specialists in Singapore.

What 'Managing Agents' Means, Day To Day

  1. 01Briefing. Giving an agent a precise task with the right context and constraints - which documents, which definitions, what the output must look like. A vague brief produces a confident, useless answer; a good one is a skill much like delegating to a new colleague.
  2. 02Checking. Reviewing what came back against the source: does the number in the summary match the filing, did it use the right quarter, did it invent anything? The most valuable junior is the one who catches the error the agent stated with complete confidence.
  3. 03Handling exceptions. Agents do the routine 80%. The remaining 20% - the unusual case, the conflicting data, the client with a special arrangement - comes back to a person. That is where judgement is learned.
  4. 04Escalating. Knowing what an agent must never decide alone: anything that moves money, commits the firm, or goes to a client unreviewed. Escalation paths are part of the job, not an afterthought.
  5. 05Improving. Noticing patterns in what goes wrong and feeding them back - better instructions, a new rule, a missing data source - so the agent is better tomorrow.

“The old apprenticeship taught juniors by making them do the work. The new one teaches them by making them responsible for work they did not do - which is a harder skill, learned faster.”

The Skills That Matter Now

  • Domain knowledge, more than before. You cannot check an agent's credit analysis without understanding credit. The juniors who thrive will be the ones who know what 'right' looks like.
  • Precise writing. Briefing an agent is writing a specification. Clear, unambiguous instructions are now a core technical skill.
  • Scepticism with evidence. Tracing a claim back to its source and refusing to accept a figure without one - the habit that would have caught this week's $20bn OpenAI revenue confusion.
  • Basic orchestration literacy. Not necessarily coding, but understanding how agents use tools, why they fail, and what controls stop them. The frameworks named in job ads - LangGraph mentions rose 679% - are worth recognising.
  • Governance awareness. Banks are hiring for responsible AI and AI governance faster than almost anything else. Knowing why an agent's actions must be logged and explainable is now career-relevant at every level.

Jobs Lost, Jobs Changed

It would be dishonest to pretend the change is painless. HSBC's reported UK wealth cuts are real jobs, and Goldman's own executive acknowledged the open question over middle management. But the evidence from the banks furthest ahead is more nuanced than a simple story of replacement. Evident, whose index of 50 banks was published this week, found that leading banks are growing headcount while improving revenue per employee, with a 21% rise in 'enablement' roles - the engineers who build the tools to manage AI. Its co-founder Alexandra Mousavizadeh put it plainly: roles are changing and some are not being backfilled, but the banks at the top of the index are growing. The work is moving rather than disappearing - towards supervising, checking and improving automated systems.

The Bottom Line

Goldman's description of new recruits managing a 'virtual army' of AI agents from their first day is backed by the numbers: bank AI job postings up 49% to almost 140,000 this year, agent orchestration mentions up more than seventeen-fold, and banks such as HSBC reportedly reshaping whole adviser teams. Managing agents means briefing them precisely, checking their work against sources, handling the exceptions they cannot, escalating what they must not decide, and making them better over time - which rewards deeper domain knowledge, not less. For graduates the advice is to learn the domain and the checking habit; for firms it is to build supervision tools and training that make juniors responsible rather than redundant. That is the agentic AI and internal tooling work we build for financial firms in London, and the career ladder is being rebuilt around it right now.

References & Further Reading

Agentic AIAI agents explainedfinance careersAgentic AI LondonAI Agency UKClient-Facing & Internal Toolseducation
Share Email
AI

AlchmAI Editorial

Research and analysis, London

The AlchmAI team writes about the markets, technology and regulation we work with every day. We build trading platforms, real-time charts and AI analysis tools for brokers, prop firms and fintech teams from our office in Mayfair, London. Every article lists its sources. Nothing we publish is investment advice.

This article is general information and commentary. It is not investment advice or a recommendation to buy or sell any investment. Important information

Keep reading

All articles