3 New Roles Born in the AI Era
Since 2023, with AI's full spread, "roles that didn't exist before" are demanded within companies. Representatives are prompt engineer / AI PM / AI Ops. This article organizes these 3 roles' duties, skills, salary, future.
1. Prompt Engineer
Role
A specialist who designs/improves instructions (prompts) to the LLM. "A craftsman drawing the best answer from AI." Turning internal knowledge into a Custom GPT, testing/stabilizing prompts when embedding into work flows, etc.
Needed Skills
- Prompt techniques (Few-shot, CoT, structuring, evaluation)
- Domain knowledge (specific fields like law, medical, sales)
- Language sense (writing that eliminates ambiguity)
- Evaluation design (A/B, Eval, quantitative criteria)
Salary (2026, Japan)
- Junior (under 1 year): 5-7M yen
- Mid (2-3 years): 7-12M yen
- Senior (lead): 12-20M yen
- In the US, USD 200K-400K (multiple reports)
Future: "Disappearing" or "Evolving" Role?
Unlike the 2023 hype period, now "side" over "dedicated" is mainstream. The state where Software Engineers and PMs "can also prompt" is standardizing. Meanwhile, prompt engineers specialized in specific domains (legal AI, medical AI) still see growing demand.
2. AI PM (AI Product Manager)
Role
Planning/driving products with AI features. Differences from a normal PM:
- Specifying "non-deterministic" output
- Model selection and cost-estimation judgment
- Deciding hallucination/bias countermeasure policy
- UX design where users can "trust the AI"
Needed Skills
- Conventional PM skills (requirements, prioritization, release)
- LLM basics (tokens, temperature, cost structure)
- Evaluation frameworks (offline Eval, A/B, human evaluation)
- Ethics/regulation (EU AI Act, internal governance)
Salary (2026, Japan)
- Senior level: 12-20M yen (big-IT/startup CTO-class)
- US: USD 250K-500K + equity
Hiring Market
Big IT, SaaS startups, consulting firms hire actively. 5+ years PM + 1+ year AI implementation is the common requirement.
3. AI Ops (AI Operations Engineer)
Role
Handles AI-system production operation. Designs/operates "the mechanism for an ML model to keep running in production." An evolution of MLOps, also called "LLMOps" in the LLM era.
Main Duties
- Model deploy/A-B switch/rollback
- Continuous monitoring of latency/cost/quality
- Prompt version management
- Building/operating the Eval suite
- Security/compliance response
Needed Skills
- SRE / DevOps basics
- Cloud (AWS Bedrock, GCP Vertex, Azure AI)
- Monitoring/tracing (LangSmith, Helicone, Datadog)
- Cost optimization (caching, quantization, routing)
Salary (2026, Japan)
- Mid: 8-14M yen
- Senior: 14-25M yen
- US: USD 200K-400K
The Entrance to These 3 Roles
Career Paths from Current Roles
| Current role | Path |
|---|---|
| Software Engineer | AI Ops, AI PM (implementation-origin) |
| PM | AI PM |
| Data Scientist | AI Ops |
| Customer support/knowledge worker | Prompt engineer |
| Consultant | Prompt engineer + AI PM |
What to Prepare
- Build a track record with personal projects: Custom GPT, internal hacks, OSS contributions
- Certifications: Anthropic Builder, OpenAI API certification, AWS ML line
- Output via OSS: GitHub, tech blogs, social
The 2026 Reality
"AI XX" titles are increasing, but what's essentially required is:
- Domain knowledge (understanding industry/work)
- The ability to actually use AI hands-on
- Judgment that handles uncertainty (living with hallucination)
"3 new roles + existing roles × AI enhancement" is the major change of the next 5 years. It's also said that, rather than a "dedicated prompt engineer," combinations of main job × AI like "marketer × AI," "sales × AI" determine high market value.
