Key Takeaways

  • Christopher Donahue rejects using artificial intelligence for staff reduction at Federated Hermes, which manages over $700 billion in money market assets.
  • Rather than cutting back-office headcount, the firm shifted from sampling one out of every 50 or 100 customer calls to automated grading of 100 percent of interactions for accuracy, tone, and warmth.
  • Donahue retained customer service managers despite automated monitoring, shifting their daily work from random spot-checking to targeted coaching based on complete call data.
  • The firm views employee retention as risk management, citing product liability and fiduciary exposure that require human accountability over automated execution.

The Headcount Trap in Fiduciary Operations

Every sponsor playbook for artificial intelligence starts with headcount reduction. Eliminate twenty seats in customer support, automate fifty back-office roles, expand EBITDA margins, and claim tech-enabled margin expansion.

Christopher Donahue takes the opposite view at Federated Hermes. Running a firm with over $700 billion in money market assets, Donahue sees headcount cuts as a dangerous misunderstanding of what operational software actually does. “I tend not to use the word efficiency when we're talking about AI, especially in an employment circle,” Donahue explained. “Cuz efficiency means FTE, I'm dead. Quality means, 'Ooh, I could do it better.'”

In financial services, the downside of an error dwarfs the margin gained by firing an employee. Custodian banks, transfer agents, and money fund desks run under strict regulatory scrutiny and fiduciary standards. Replacing frontline staff with automated agents introduces legal and product liability risks that software vendors do not indemnify. Donahue tells incoming summer interns to pay twenty dollars a month for consumer AI tools to master them, because human workers who know how to command these systems will hold the line on client trust. Donahue targets zero operational errors, not a smaller payroll.

From Two-Percent Sampling to Total Inspection

The real operational shift inside Federated Hermes centers on quality control. Traditional compliance and quality assurance teams in asset management rely on random audits. A supervisor listens to one out of every 50 or 100 recorded calls, notes infractions, and holds a retrospective review with the representative weeks later. Ninety-eight percent of client touchpoints pass through unmonitored.

Federated Hermes replaced that sampling model with full coverage. “Well, when you have AI do that, all of a sudden instead of only listening to one out of every 50 or 100 calls, you listen to them all,” Donahue noted. “And then the AI grades them on things that are really good.”

This shift exposes the flaw in standard cost-out models. When software evaluates every call, the data stream expands by a factor of fifty. If an operator fires the supervisors, that data sits unread in a dashboard. Donahue kept the supervisory tier intact to act on the findings. “Do we have less managers because they don't have to do that job anymore? No. We have higher quality calls, higher quality training, and a better overall result.”

Instead of hunting for mistakes across tiny sample batches, managers receive precise patterns on accuracy, clarity, and client demeanor across the entire book. Training becomes surgical. The workforce remains the same size, but the operational output jumps in reliability.

Why It Matters

Sponsors building value-creation theses around quick back-office headcount cuts run the risk of mistaking compliance exposure for productivity gains. In capital-management and fiduciary businesses, automated systems create the most enterprise value when applied to total risk detection rather than payroll reduction. The operators who maintain staff while increasing audit coverage to 100 percent build defensible client retention, while competitors cutting staff to meet quarterly EBITDA targets inherit hidden compliance liabilities.