The Post-AI Analyst
The analyst role is not disappearing. It's expanding so fast that the old job description already feels stale.
For years, the analytics stack created handoffs everywhere. A stakeholder asked a question. An analyst translated it. An analytics engineer transformed data. A data engineer maintained pipelines. A BI layer surfaced metrics. Everyone waited on everyone.
That operating model is collapsing.
With AI agents, one person can now do the mechanical work that used to require a small pod. Querying. Modeling. Dashboards. Documentation. Basic pipeline patches. Drafting PRs. Iterating on definitions. The execution layer got radically cheaper.
So if execution is cheap, what becomes scarce?
Context. Judgment. Trust.
This is not a eulogy. The role is alive. But narrow analysts are done.
The New Moat: Business Intimacy + Data Judgment
If your differentiation was "I can write SQL faster than others," AI will eat that edge.
If your differentiation is "I deeply understand how this business works, what decisions matter, how metrics get gamed, where data lies, and how to build trust into self-service," your leverage just went up.
In the post-AI org, the analyst becomes the business-facing AI enablement layer for anything data-related.
Not the human query engine.
The architect of how business teams ask questions, get answers, and trust those answers.
Role Boundaries Are Melting. Good.
A trend I keep seeing across tech companies: titles matter less, impact surface matters more.
Call it "member of technical staff," call it "full-stack data," call it whatever you want. The pattern is the same: orgs are increasingly saying, "I don't care what your old box was. Work with agents. Work across boundaries. Ship outcomes."
That means:
- The data engineer can do analyst-style exploration when needed.
- The analyst can ship semantic models, docs, and agent instructions that look like analytics engineering work.
- The analytics engineer can move up into business decision quality and metric governance.
The handoff economy is dying. Outcome ownership is replacing it.
Self-Service Without Trust Is Just Faster Confusion
Everyone says they want self-service. Most teams still confuse it with "more people can run queries now."
That's not self-service. That's query decentralization.
Real self-service means people across Product, Sales, CS, Marketing, and Engineering can answer meaningful questions quickly without creating analytical chaos.
AI gets you the speed.
You still need humans to design trust.
In practice, that usually means investing in ingredients like:
- Semantic models and shared metric language
- Curated context for agents (docs, definitions, known caveats)
- Specialist agents with scoped access and domain instructions
- Review loops for high-impact decisions
- Provenance and explainability norms
- Guardrails around permissions, compliance, and data quality
No silver-bullet framework. No one-size-fits-all playbook. Just deliberate system design.
What This Looks Like in the Real World
At Preset, we've seen how far this can go with specialized assistants.
In AI Enablement Engineer: The Highest-Leverage Role in Tech, I described the emerging role focused on scaling AI leverage across teams.
In Meet DatAgor, Preset's AI Data Engineer, you can see a concrete data example: a specialist assistant with access to our warehouse, dbt context, dashboard layer, and operational workflows. It helps democratize access to analysis while staying grounded in the actual data system.
That pattern matters for analysts. You can be the person who defines, curates, and operationalizes this layer for your company.
The Career Move: Become the Leader of Trusted Data Self-Service
If you're an analyst wondering how to stay relevant, here's the blunt answer:
Don't defend your old boundary. Own the new surface area.
Your job is increasingly to:
- Bring business context into AI-assisted analytics workflows
- Raise decision quality, not just dashboard velocity
- Encode definitions and judgment into reusable assets
- Partner with data and platform teams on trust scaffolding
- Teach the organization how to work with data agents responsibly
This is not "learn prompt engineering and call it a day."
This is leadership.
Not necessarily people management. Leadership in system design, decision quality, and organizational leverage.
The Shift
If AI can do more of the execution, then analysts must do more of the meaning-making.
If roles are collapsing, then analysts should widen, not retreat.
If self-service is accelerating, then trust must accelerate with it.
If everyone can ask the data questions, someone still has to ensure the answers are coherent, comparable, explainable, and decision-grade.
That someone can be you.
The post-AI analyst is not a report builder.
The post-AI analyst leads trusted self-service.
Maxime Beauchemin is the creator of Apache Airflow and Apache Superset, and CEO of Preset. He's currently building Agor, a multiplayer canvas for orchestrating AI agents.
