AI Statement

Overall IDEA Consult AI Statement (e.g. Website)

IDEA Consult has been advising on public policy since 1998, across the full policy cycle: analysis & monitoring, policy design, implementation, impact assessment and evaluation. Tools and methodologies have continuously evolved. AI is the largest such change, and we want to be transparent with our clients and partners about how we deal with it.

Even in the AI era, we maintain our credo: we offer evidence-based and independent advice to improve impact of public policies.

The fundamentals of good public policy consultancy include collecting and analysing data, reading widely, processing large bodies of evidence, and producing and refining drafts and reports. AI amplifies and speeds up part of that work. This could be considered a threat to research-based consultancy. We see it the other way around.

Our clients are not interested in data collection alone. They are most interested in expert insights, in actionable, impactful recommendations and in expert-led implementation. Our work then consists of deciding which evidence really matters, what it means for a particular institution in a particular context, and which recommendation is sound and substantiated enough to be implemented with the highest likelihood of success and impact. That work cannot be automated. Analytical output may become a commodity, but expert judgement that distinguishes a solid conclusion from a plausible one and turns recommendations into implementation within a complex institutional setting is not: it is even becoming scarcer and worth more than it has ever been.

AI lowers the cost of the first draft, but it raises the value of the final product.

IDEA Consult invests in using AI tools well, rather than avoiding them. This is why responsibility for every conclusion we deliver still sits with a named expert.

What we use AI for

We use enterprise-grade AI tools with embedded and tailored Data Protection Agreements (DPAs) that protect your data and do not use it to train models. This includes synthesising literature, scanning regulatory and statistical sources, structuring early drafts, supporting interviews and case studies, processing large amounts of data, and surfacing perspectives and patterns that we then test. AI does not replace our research: it automates some repetitive processes, generates initial inspiration and amplifies our evidence bases, so that our experts spend more of their time on analysis, interpretation, contextualisation, recommendation and the design of the best options for implementation. This is where IDEA Consult adds the most value.

AI amplifies our evidence bases,

but we make the assessment and remain solely responsible for it.

Every recommendation we deliver has a team of experts behind it who has weighed the evidence and is accountable for the conclusion.

How we use AI

Our use of AI follows the responsible research standards of the European Research Area and our own internal guidelines:

We work only with secure, data-protected tools.
We do not enter personal or confidential information into systems that are not contracted to protect it.
We are transparent about where and why AI has been used, naming the tools and their role on request.

We treat AI tools as instruments to be assessed, not trends to be chased. We test what they can and cannot do, and we deploy them only where they raise the quality of our work. Our experts check what these tools produce against the original evidence, and every claim that informs our work traces back to a source a person has verified.

We stay up to date with how standards are evolving: understanding where the risks are heading, not only where they sit today, is part of how IDEA Consult stays up to date on your behalf.

We analyse large bodies of evidence in a neutral and independent way; we connect minds, co-creating with you and your partner networks and empowering people to drive change within their institutions and territories.

AI-assisted but human-owned.

Every factual claim that informs a deliverable traces back to a source a researcher has verified. We are aware of, and actively check for, the known limitations of these tools, including bias, hallucinations and factual inaccuracy.

Please find below a non-exhaustive list of processes supported by AI, where AI support stops, and where human input comes into play. Each category of use states its purpose, the boundary on its extent, and the human input applied before any AI-assisted material enters a deliverable.

Table 1: AI-tool categorisation and use in deliverables

Category of use	Purpose	AI input	Human input
Text summarisation / rewriting	Summarise, restructure or improve the readability of source texts	First-pass synthesis or drafting only	A researcher checks against the original sources; all claims are traced to a verified source
Interview processing	Summarise interview notes and transcripts	Draft meeting minutes	Review and correctly position meeting minutes and interview summaries for the intended audience
Translation	Produce a translated version of a text	Draft translation	Reviewed and corrected by a researcher with the relevant language competence
Text and data mining	Identify patterns, themes or references across large bodies of text and data	Discovery and triage, not a basis for conclusions on its own	Findings re-checked against primary data before use
Data analysis support	Assist with processing, coding or exploratory analysis of data	Supporting computation and exploration	Methods and results reproduced or validated by a researcher
Drafting assistance	Produce an initial draft of text a researcher then rewrites	Initial draft only	Substantively rewritten and verified; final wording is the researcher’s
Image creation	Generate an illustration or visual element from a prompt	Visual content	Reviewed for accuracy, suitability and rights; labelled where required
Generate PowerPoint presentations	Generate first draft presentations	Presentation structures and layouts	Researchers improve presentations towards accuracy and better readability for the intended audience

Source: IDEA Consult