Terzin — AI Workforce Impact Map · European Photon Research Facility

Illustrative assessment. This report is a composite based on publicly known European photon-source facility structures and our peer-reviewed CAS methodology (GenAI and the Future of Government Work, IBM Center for the Business of Government). It is not a real engagement with any specific institution. Role structures, headcounts, and CAS scores are modeled to demonstrate the framework's application to international research facilities. In a real engagement, all data is derived from client-provided job descriptions and organizational structure.

Executive Summary

C Complementarity A Augmentation S Substitutivity · Scale: 1 (low) → 5 (high)

Key Findings & Priorities

Transform Priority 1

Scientific computing transforms experimental physics — from data acquisition through publication

Research Software Engineers (C=4.8), Data Pipeline Engineers, ML Research Engineers, and Beamline Data Analysts all score highest on complementarity. AI dramatically accelerates the data-to-paper loop: fast feedback during beamtime, automated reduction pipelines, AI-assisted scientific writing. Substitutivity stays moderate (S ≈ 2.6–3.2) because interpreting whether a peak is real, whether a fit converged, and whether a result is publishable still requires human physics expertise.

Transform Priority 2

The experimental physicist's week spans the entire CAS spectrum — task-level analysis is essential

A Postdoctoral Fellow's tasks range from analysis code (S=3) and literature review (S≈4) to experimental design (S=1), beamtime execution (S=1), and student mentoring (S≈0.5) — all in a single week. Role-level averages obscure where AI actually adds value. Tasks involving photon-source control, sample handling, and physics interpretation remain protected; those involving Python, plotting, and prose are heavily augmented.

Protect Mission-Critical

The accelerator complex and detector hardware are physically protected from AI substitution

Beam Operators (S=1.8), Cryogenics Technicians (S=1.4), Vacuum Technicians (S=1.4), Detector Assembly Technicians (S=1.4), and Alignment & Survey Specialists (S=1.8) work with physical systems where AI assists through monitoring and predictive diagnostics — but cannot tighten a vacuum flange, cool a magnet, or align an undulator. The physics of the facility itself protects ~430 hands-on roles that only exist at research infrastructures of this kind.

Reskill Structural Insight

~430 transient researchers need zero-training AI tools — they rotate every 2–5 years

Postdocs (152), PhD students (149), visiting scientists (55), and visiting engineers (42) cycle continuously. AI deployment for this population must be self-service with near-zero onboarding. Tool UX — not capability — is the binding constraint.

Interactive AI Impact Map

View by:

Color by:

Organizational Deep Dive

How to read this section

CAS Scores (1.0 – 5.0)

Each role is scored on three dimensions: Complementarity (AI enhances capability), Augmentation (workflow change needed), and Substitutivity (displacement potential). Department scores are headcount-weighted averages of constituent roles.

Composite Score (1.0 – 5.0) and Impact %

The simple average of C, A, and S — a single number summarizing AI impact on a department. The percentage shown below each score normalizes to 0–100% (score ÷ 5) for at-a-glance comparison. Higher = broader AI transformation potential.

Productivity Lift (%)

Estimated equivalent productivity gain at 70% AI adoption — the share of working hours that could be redirected toward higher-value work if the department fully adopts AI tools. Literature-derived; calibratable to your operations.

Deployment Readiness

Qualitative assessment combining technical maturity of available AI tools, organizational change capacity, and regulatory/governance constraints. High = deploy now. Medium = pilot first. Low = wait for tooling or governance to mature.

10–14%

Conservative · 50% adoption

15–20%

Moderate · 70% adoption

20–26%

Aggressive · 90% adoption

Based on: Noy & Zhang (2023), Dell'Acqua et al. (2023), Brynjolfsson et al. (2023). Ranges are calibratable starting points.

4.1Division Comparison

Side-by-side view of all departments, sorted by composite score.

4.2Division Detail

Click any department row to expand and see all its roles, the primary AI driver, and the key risk to manage.

Composite

Lift

Readiness

4.3Full Role Assessment Matrix

All roles across the organization ranked by AI impact score — the percentage of maximum possible AI transformation potential for each role.

▸ Show all roles

4.4Cross-Role Patterns

The hardware–software divide maps cleanly onto CAS scores

Every hands-on role (beam operators, vacuum/cryogenics technicians, sample prep, alignment specialists) scores S ≤ 1.8. Every code/data role scores S ≥ 2.6. AI augments analysis and simulation; the physical operation of accelerators, magnets, and detectors stays human.

Scientific computing is the cross-department catalyst — and it touches the user programme

180+ research software engineers, ML engineers, and data engineers serve Experimental Physics, Accelerator Operations, and Detector Engineering simultaneously. One AI platform deployment can compress the data-to-publication cycle for ~2,000 user-programme experiments per year — directly improving the facility's scientific output.

Administrative AI impact matches corporate benchmarks — but governance does not

Finance, HR, and procurement roles score similarly to private-sector equivalents (S ≈ 2.6–3.4). Standard enterprise AI tools apply technically. The constraint is institutional: member-state procurement rules, staff association consultation, and audit transparency requirements shape what can actually be deployed and how fast.

Mission-critical roles need a "do-not-substitute" policy layer

Unlike a corporation optimizing margin, EPRF exists to produce science. Beamline Scientists (S=1.4), Instrument Scientists (S=1.6), and User Support Scientists (S=2.0) are the connection between the facility and ~3,500 external user-programme researchers per year. These roles must be evaluated through a mission lens, not a cost lens — even where scores suggest automation potential.

4.5Organization-Wide Readiness

Data Governance
Scientific data often pre-publication and embargoed. AI tools must respect data access policies, collaboration agreements, and intellectual property conventions across member states.

Multi-Stakeholder Governance
Member state councils, scientific advisory committees, staff associations, and funding bodies each have input on workforce decisions. AI deployment plans need defensible justification across all audiences.

Transient Workforce
~430 fellows, PhDs, and visiting researchers rotate every 2–5 years. AI tools must be self-service with minimal onboarding. Training investment per person is structurally limited by tenure duration.

EU AI Act Compliance Mapping

The EU AI Act (Regulation 2024/1689) creates specific obligations for organizations deploying AI systems. This section maps our CAS findings to the Act's requirements — converting this analysis from a strategic document into a compliance artifact your regulatory file requires.

Annex III High-Risk Identification

Three CAS findings correspond to potential Annex III high-risk categories:

Category 4 (Employment): AI tools that influence researcher assignment, fellowship selection, or performance evaluation

Category 6 (Education): AI-assisted assessment of PhD students and fellows where outcomes affect career progression

Category 1 (Safety): AI integration in accelerator control and radiation safety systems

Article 26 Deployer Obligations

The CAS assessment partially satisfies four Article 26 requirements:

Art. 26(1): Organizational measures to ensure AI use in accordance with instructions — CAS maps which roles interact with AI and how

Art. 26(2): Human oversight assignment — CAS substitutivity scores identify where human oversight is mandatory

Art. 26(5): Fundamental rights impact assessment — CAS workforce impact analysis provides the required evidentiary basis

Art. 26(7): Worker information obligations — CAS role-level reports serve as the structured disclosure the regulation requires

In the full engagement: A complete compliance mapping table specifying which CAS findings correspond to which Annex III categories, a gap analysis of current practices against Article 26 obligations, and recommended documentation templates for your regulatory compliance file.

Deployment Roadmap

Findings become decisions when they're sequenced. This section translates CAS priorities into a phased deployment plan — which pilots to launch first, what governance must be in place, and what success looks like at each stage.

First 90 Days

Pilot 1: Scientific Computing

Why first: Highest C-scores (4.6–4.8), most AI-literate staff, lowest mission risk. Research Software Engineers and Data Pipeline Engineers already use AI tools informally.
Target: Formalize AI coding assistant deployment for 180 scientific computing staff.
Measure: Code generation adoption rate, time-to-analysis reduction, user satisfaction.

Months 4–9

Pilot 2: Research Writing & Literature

Why second: High complementarity (C=4–5) for writing-intensive roles. Postdocs and PhD students spend ~25% of time on scientific writing and literature review.
Target: AI writing assistance for 430+ transient researchers via self-service tools integrated with Overleaf/LaTeX.
Measure: Time-to-first-draft reduction, self-service adoption without training.

Months 7–12

Pilot 3: Administrative Automation

Why third: Highest substitutivity in admin (S=2.8–3.6) means direct efficiency gains. Procurement, finance, and HR workflows are standardized and AI-amenable.
Target: AI-assisted procurement processing, travel expense automation, HR onboarding workflows.
Measure: Processing time reduction, error rate, staff satisfaction.

In the full engagement: Detailed pilot design documents for each phase, success criteria and KPIs, risk mitigation plans, governance prerequisites that must be satisfied before each pilot launches, and a decision framework for scaling from pilot to full deployment.

Governance Recommendation

A facility with member state oversight, scientific councils, and staff associations cannot deploy AI the way a corporation does. This section proposes a governance structure that makes AI workforce decisions defensible across all stakeholder audiences.

Proposed: AI Workforce Transformation Committee

Composition

• Deputy Director General (chair)

• Department heads or delegates (5)

• Staff Association representative

• Scientific Council observer

• Data Protection Officer

• External AI ethics advisor

Mandate

• Approve AI deployment pilots above a risk threshold

• Review CAS assessments before workforce action

• Set "do-not-substitute" designations for mission-critical roles

• Report to Council on AI workforce impact annually

• Establish escalation paths for contested decisions

Scientific Council Pathway

Any AI deployment affecting research methodology or experimental capability is reviewed by the Scientific Advisory Committee before approval. Ensures mission alignment.

Member State Consultation

Quarterly briefing to Finance Committee on AI workforce impact. Annual report to Council including CAS trend data, pilot outcomes, and headcount implications.

Staff Association Protocol

30-day consultation period before any AI deployment that changes job descriptions or performance criteria. Staff Association observer seat on the Committee with speaking rights.

In the full engagement: Complete governance framework document including committee terms of reference, decision matrices, escalation procedures, template reporting formats for Council and Finance Committee, and a model staff consultation protocol adapted to your institutional statutes.

What You Receive

A Terzin engagement delivers a structured set of artifacts, not just analysis. Here's what the deliverable looks like for an organization of this scale.

📊

Interactive Organization Report

The report you're reading now — but with real data, all sections unlocked, all roles assessed. Delivered as an interactive HTML dashboard your leadership team can explore. Department heads can drill into their own teams.

📋

Role-Level Reports (per assessed role)

Full 7-section analysis for each role: executive summary, key findings, impact index, competency assessments with pipeline-generated reasoning, task-level drill-down, financial impact quantification, and readiness & risk assessment.

🗺️

Deployment Roadmap & Governance Package

Phased pilot plan with success criteria, governance framework with committee terms of reference, EU AI Act compliance mapping, and decision templates for your institutional approval processes.

🤝

Presentation & Q&A Session

90-minute presentation of findings to your leadership team, followed by structured Q&A. Designed to be the decision-making meeting — not a pre-read for one. Optional: department-level breakout sessions for detailed discussion.

4 weeks

Assessment timeline

All roles

Full organizational coverage

Quarterly

Optional update cadence

Organizational AI Impact Map