Where the method began: mapping the field of technological singularity for the Singularity Museum at the International Centre for AI (alem.ai), Astana, 2024
Built for depth. Directed by experience
The foundation of my pipeline was developed in 2024 for the International Centre for AI (alem.ai), a project selected through an international competition, where LLM-based agents supported research, narrative development and co-curation. Since then, I have evolved it into a custom framework integrating three leading AI models with a proprietary knowledge base of international standards and best practices.
The pipeline lets me explore more sources and precedents, challenge assumptions and test multiple creative directions before selecting a concept. It brings the capacity of a small multidisciplinary team into one coherent process under experienced creative direction. When needed, additional specialists - architects, art directors, concept artists and others - can join within the same creative framework.
WHAT THE PIPELINE ASSISTS
Source discovery, research, precedent comparison, benchmarking, concept exploration, synthesis, and quality checks.
WHAT I DECIDE
The problem, strategic direction, which evidence matters, curatorial and creative judgment, recommendations and final deliverables.
WHAT YOUR EXPERTS VALIDATE
Scientific and disciplinary claims, collection facts, institutional policies and specialist requirements. You approve the strategic direction.
CONFIDENTIAL BY DEFAULT
Confidential material stays in a closed, local environment. Commercial AI is used only for public, non-confidential work, with training on your data excluded.
DISCUSS A PROJECT
VERIFIED, NOT ASSUMED
Sources remain traceable, facts are distinguished from hypotheses, and weak evidence is flagged. Every deliverable passes my review.
ASK FOR AN EXAMPLE
SYSTEMATIC BY DESIGN
Six stages. Three AI models. Human judgment at every gate. After each project, the pipeline is reviewed and refined, continuously improving its methods and capabilities. The result is a more rigorous and reliable process from one project to the next.
ONE MODEL WRITES, ANOTHER CHECKS
No model validates its own work. Different models research, develop and challenge the work, so errors and weak assumptions are caught before they reach me. Nothing moves to the next stage until I have reviewed and approved it.
GROUNDED IN THE DISCIPLINE
The pipeline draws on a curated library: ICOM ethics and museum standards, professional literature, interpretation and audience research, published strategies and evaluations, international precedents with live research. Every source is cited, so any statement can be traced back.
01
Research
Client materials, international precedents and professional standards are collected, organised and verified. Sources are evaluated for relevance and reliability to establish a traceable evidence base.
02
Analysis
Patterns, opportunities, constraints and knowledge gaps are identified. Assumptions are tested against available evidence, while critical gaps trigger further research and investigation.
03
Concepts
Several creative directions are developed, each with a rationale, key assumptions, potential risks and initial spatial or visual studies. I evaluate the options and select the strongest.
04
Critique
Two independent AI critics challenge the preferred direction, testing its narrative, evidence, audience experience, accessibility and feasibility. Weaknesses are identified and recommendations prepared.
05
Development
The selected direction becomes an integrated concept connecting narrative, content, visitor experience and spatial design. Technical requirements are identified and go to specialists as clear requirements.
06
Validation
Independent models verify claims and sources. Automated checks assess consistency, completeness and presentation. I review the results and approve every deliverable at each stage before it reaches the client.
WHAT THIS CHANGES FOR YOUR PROJECT
TRADITIONAL TEAM MODEL
MY AI-ASSISTED MODEL
One senior lead with supporting team of three to six specialists
One senior author supported by a supervised multi-agent AI pipeline
Typically three to four months
One to two months
Multiple contributors working within a shared creative direction, with coherence harder to hold
One accountable author maintaining a coherent vision throughout
Research and exploration shaped by team capacity and budget, and available review resources
Broader research, more concepts tested, thorough independent AI checks and final human approval
Reflects longer involvement of a multidisciplinary team
One senior fee, fixed for a defined scope
Indicative comparison for early-stage strategy and concept development. Timelines and team composition vary by project.
Want to know more about the method? Read my article on .
how evolutionary thinking, critical rationalism and contemporary AI research shaped my approach to museum concepts
An example from practice
The research dashboard above shows week 2 of the research stage on live work: the institutional strategy and positioning of a major European museum preparing a new building. Client details are redacted under NDA; the process and headline figures come from the project.
Strategic research benchmarked the museum against 18 institutions and projects across Europe, North America, the Middle East and Asia. Evidence came from national census and museum statistics, national audience studies, peer museums’ strategies and annual accounts, public audit reports, site footfall data and the museum’s own collection and visitor records. The pipeline ran 126 working tasks; 3,482 sources were retrieved and screened, 719 were rejected, and the evidence base was narrowed to the sources cited in the final documents.
Want to know more about the method? Read my article on how evolutionary thinking, critical rationalism and contemporary AI research shaped my approach to museum concepts.