Jamie Dimon on AI: Layoffs are real - redeployment is the move. What this means for Customer Support
JPMorgan Chase CEO Jamie Dimon is blunt: AI will cut some jobs and create demand for new ones. The bank spends about $2 billion a year on AI and reports roughly the same in direct cost savings so far. AI is embedded across fraud, marketing, service, and internal ops, with a proprietary language model used weekly by 150,000 employees. The message to operators is clear: get ahead of it or pay the price.
The signal for Support teams
Tier-1 work will shrink as AI handles repetitive inquiries, summaries, and suggested replies. Human roles move upmarket: complex cases, escalations, relationship work, and quality. Companies that retrain and redeploy will keep talent and compound efficiency. Those that wait will learn this the hard way.
What to expect in the next 12-24 months
- Higher self-service and bot containment on simple issues (passwords, order status, basic troubleshooting).
- Lower average handle time via AI draft replies, auto-summarization, and knowledge suggestions.
- Fewer roles in pure queue handling; more roles in QA, bot ops, and complex case resolution.
- Managers judged on automation coverage, quality controls, and redeployment success - not just headcount.
Manager playbook: move first, not last
- Map tasks by complexity: automate low-risk, low-complexity first; protect high-risk workflows with human review.
- Pilot inside existing tools (Zendesk, Intercom, ServiceNow, Salesforce): AI macros, suggested replies, summarization, routing.
- Set guardrails: no refunds, PII edits, or policy exceptions without human approval; log every AI action.
- Retrain and redeploy: upskill agents into conversation design, bot QA, knowledge management, and escalation handling.
- Measure what matters: first contact resolution, CSAT, AHT, containment rate, error rate, reopens, audit findings.
Skill stack for Support professionals
- LLM prompting for support use cases: intent, tone control, summarization, safe refusals, policy grounding.
- Workflow automation: triggers, routing, knowledge suggestions, and API actions inside your help desk.
- Quality and safety: red-teaming prompts, bias checks, PII handling, approval ladders, and audit trails.
- Knowledge management: structure, versioning, feedback loops from tickets to articles.
- Analytics fluency: build dashboards for AI coverage, savings, and quality, then iterate based on data.
Quality and risk: keep trust intact
- Always-on human-in-the-loop for money movement, account changes, and policy exceptions.
- Clear escalation paths and SLA tiers when AI confidence is low or sentiment drops.
- PII-safe prompts and redaction; log prompts, outputs, approvers, and outcomes.
- Weekly QA sampling: check for hallucinations, tone drift, and policy mismatches; update prompts and articles fast.
ROI you can defend
Dimon's numbers signal scale: billions in yearly spend and billions saved. Your version is simpler: quantify minutes saved per ticket, multiply by ticket volume and labor cost, then subtract tooling cost and QA time. Track redeployment gains (e.g., backlog cleared, higher FCR), not just headcount cuts. Publish the scorecard monthly and tighten where waste shows up.
What skeptics get right
Returns are uneven, and costs climb without tight scoping and QA. Even large firms admit many AI projects miss clear outcomes. If you can't measure quality and savings per workflow, don't scale it. For a sober view on costs and ROI pressures, see research coverage from Goldman Sachs.
Practical steps this quarter
- Pick three workflows: summarization, suggested replies, and intent-based routing. Ship in 30 days.
- Create a two-tier approval system for refunds and policy exceptions. No exceptions.
- Stand up a weekly "AI QA" meeting: review 20 transcripts, fix prompts, update knowledge, repeat.
- Nominate two agents to become bot operators and one to own knowledge hygiene.
Training resources
- Role-based upskilling for support teams: Courses by job
- Hands-on certification for conversational AI and workflows: ChatGPT certification
Bottom line
AI will cut some roles and multiply output for those who adapt. Follow the operators who are already integrating it across core workflows, then retrain and redeploy your team. Start with clear guardrails, small wins, and measurable outcomes. Be early, be specific, and make quality non-negotiable.
Your membership also unlocks:
