Canada needs to build major projects and build them faster. The current way of doing impact assessments cannot deliver that. Reviews are slow, expensive and structurally vulnerable to legal challenge — not because regulators lack diligence, but because the evidence base they work from is fragmented across departments, formats and decades.
The federal government holds reams of relevant data: decades of environmental assessments submitted to regulators; studies and samples accumulated by the Geological Survey of Canada; fish, wildlife and land use information held across departments. Current systems for using these datasets remain in their infancy.
Assembling and using impact assessment data to its full potential would save governments and project proponents valuable time and money. More importantly, it would mean the decisions that flow from the data are defensible.
Under the current system, each major project application requires the proponent to commission its own baseline studies on water, wildlife, air quality, vegetation, traditional land use and more. These studies are filed with the Impact Assessment Agency of Canada, which leads the federal process. For cross-provincial pipelines, the agency works jointly with the Canada Energy Regulator, and for nuclear projects, it works with the Canadian Nuclear Safety Commission. These other government bodies co-ordinate review with the federal expert departments that hold the related data: Environment and Climate Change Canada, Fisheries and Oceans Canada, Natural Resources Canada, Health Canada and others. Proponents typically lean on outside consultants to compile this complex information, adding time and cost to the process.
When the assessment ends, those proponent-funded data sit in PDF appendices on the Canadian Impact Assessment Registry, the agency’s public online repository for all project documents (Impact Assessment Agency of Canada, 2025a). The registry is a filing system, not a database. You can find a study, but you cannot query the data inside it. Project proponents for subsequent assessments can barely use the information.
For example, a company proposing a mine near a sensitive watershed must commission its own hydrology studies on stream flows, water quality, groundwater and fish populations. Yet much of this information already exists within the federal government: Environment and Climate Change Canada monitors water quality, Fisheries and Oceans Canada holds decades of fish data and Natural Resources Canada maps geology. The information does not arrive in usable form, so proponents often end up repeating work the government has already funded.
The Crown pays for this fragmentation indirectly — and so does industry. Proponents pay consultants to collect data the federal government already holds. Federal staff then spend months reviewing and integrating those studies, often resolving discrepancies with their own department’s existing records. The administrative burden falls on both sides of the table and lengthens every assessment.
The problem runs deeper than storage. The baseline studies that anchor each assessment (e.g., water or geological studies) are inconsistent across projects.
Project-specific Tailored Impact Statement Guidelines specify what each proponent must produce (Impact Assessment Agency of Canada, 2025b). But the underlying methods — how a wildlife survey is conducted, how cumulative effects are scoped, how Indigenous land use is documented — are set largely by each proponent’s consultants, not the federal regulators.
Peer-reviewed research confirms the consequences. A 2025 Dalhousie study of 266 impact assessments on 227 mining projects across 13 jurisdictions found that 20 per cent of projects had public records that were incomplete or missing. For projects assessed by both provincial and federal regulators, 60 per cent showed significant discrepancies in reported project sizes, footprints and durations (Westwood et al., 2025).
Data collected for one project cannot easily inform the next, even when the projects share the same watershed, species at risk or Indigenous territory.
Most federal regulators have digitized their document repositories. Almost none have done the harder work of extracting structured, searchable data from those documents. The Canada Energy Regulator’s (CER) Biophysical, Socio-Economic, and Regional Data and Information (BERDI) tool makes data from 40 pipeline applications usable across assessments (Canada Energy Regulator, 2026). It is the exception, not the norm. Most mining applications remain locked in PDFs, as do forestry permit applications, water-taking permits, and project files for infrastructure such as roads and transmission lines.
BERDI is also a learning resource. The CER team developed an open-source method that uses Natural Language Processing and Machine Learning to extract tables, figures and maps from environmental and socio-economic assessment PDFs (Canada Energy Regulator, n.d.). The same method identifies and protects sensitive Indigenous knowledge so that it is excluded from public search results, establishing an early Canadian precedent for data sovereignty architecture. The proposed Sovereign Impact Assessment Agent can build on this work rather than reinvent it.
This memo does not specify the technological architecture required to build a Sovereign Impact Assessment Agent (SIAA). Artificial intelligence and data platforms are evolving too quickly for a policy proposal to lock the federal government into a specific approach. Those decisions belong to system designers and engineers working with data experts, operational teams and project proponents.
Similarly, this memo does not recommend where the SIAA’s daily operations should sit. The Impact Assessment Agency of Canada would seem a natural home, but it is under-resourced and does not currently have the technical capacity for such an undertaking.
The SIAA would not eliminate political contestation, replace the judgment of departmental officials or guarantee that every project gets approved. It would mean fewer projects quashed for inadequate evidence, fewer projects deferred for methodological uncertainty, and a regulatory environment in which good projects move forward and weak projects fail for the right reasons.
To build the SIAA, the first task is to structure data the federal government already holds and make it interoperable across departments and agencies. This is complex work and should be undertaken in iterative phases within narrowly defined pilot programs.
A dedicated multidisciplinary team — bringing together federal data experts, system designers, engineers and Indigenous data governance specialists — would identify federal datasets across government that are relevant to impact assessments and begin with ones where methodological consensus is largely settled, such as those around hydrology, vegetation or baseline species inventories. Building institutional trust on uncontested data is the precondition for tackling contested data later (e.g., data measuring greenhouse gas emissions).
Priority datasets must then be structured, verified and formatted for interoperability — that is, so different systems can read, combine and act on the data without manual translation. PDF documents, such as those containing static images of tables and graphs, are not structured in a way that supports interoperability.
In the second phase, the team would test the relationships between datasets using AI tools such as reasoning models and verify that the information they produce is correct. The team would also work with federal departments, provinces, Indigenous Nations, proponents and the research community to establish and document the methodologies used by the system, beginning with areas of existing consensus.
The third phase would use an embedded AI agent to pre-populate a file for an impact assessment with what is known about the project area in question, and point out what is missing to the project proponent. The SIAA would tell the proponent what information it needs to bring to the table, as well as the methodology and format for collecting that information. Proponents would send data back to the federal government in a structured format, so the data continues to be collected in a usable manner.
Sustainability over time means that, from this point forward, every department must produce new data in a structured, interoperable form, whether for public consumption or internal use.
The mandate, timeline and delivery reporting should sit with the Privy Council Office (PCO), while the SIAA’s day-to-day operations should reside with a federal implementing agency of PCO’s choosing. The mandate lives at PCO not because it is a technology project but because it requires deputy ministers in multiple departments to treat the work as a priority. Only PCO can hold deputies accountable for cross-departmental delivery.
The Major Projects Office, established in August 2025 as a special operating agency supported by PCO, set the pattern: a delivery-driven file with political backing, a single accountability point and a clear timeline (Privy Council Office, 2026). The SIAA borrows that pattern — not the structure. Deputy ministers in participating departments should carry responsibility for their data contributions in their performance management agreements with the Clerk of the Privy Council. Without that, the work will compete for attention with everything else on their desks and lack momentum.
The selected implementing agency must be funded to staff the team, including the authority to second experts from across government, hire at market value when needed and form direct partnerships with research institutions.
Throughout, both PCO and the implementing agency must align with the Treasury Board Directive on Automated Decision-Making (Treasury Board of Canada Secretariat, 2025), which sets the federal rules for systems that inform government decisions (more on this below).
Co-ordination with provincial and territorial regulators is essential. Major projects routinely require both federal and provincial approvals, and the federal evidence base only delivers full value when provinces are at the table. The goal would be to achieve interoperability between provincial and federal datasets, with established typologies and structure. PCO leads on intergovernmental relations, while the implementing agency leads on operational data co-ordination with provincial counterparts. Both would build on the existing federal-provincial co-operation agreements signed under the “one project, one review” commitment (Impact Assessment Agency of Canada, 2026).
In the future, provincial regulators should have access to the federal evidence base for shared-jurisdiction projects via the SIAA. The longer-term destination is an integrated federal-provincial evidence base. But that destination is reached by starting smaller and proving the model works.
The AI Sovereign Compute Infrastructure Program will deliver approximately $890 million over seven fiscal years to build a national, Canadian-governed AI supercomputing system, beginning in 2026-27 (Innovation, Science and Economic Development Canada, 2026). The SIAA should be a priority public-interest user of that program.
Data are an asset. The SIAA would be the first system to bring together federal regulatory data, Canadian-built computing and Indigenous data sovereignty into a single assessment ecosystem. Other jurisdictions are building pieces; none have built the integration.
A sovereign tool requires sovereign foundations. The value of sovereignty — data control, security, freedom from vendor lock-in — collapses if any major component is foreign-hosted or foreign-owned.
The SIAA’s primary users are federal regulators. Indigenous Nations would access the system on terms they set (see below). Proponents would see results pertaining to their projects, but not the underlying federal evidence base. Sensitive information — security-related, commercially confidential or held under Indigenous sovereignty — remains protected within the federal IT ecosystem.
Anchoring the SIAA in this ecosystem produces benefits beyond impact assessment. It brings together a broader network of collaborators, including the SIAA’s implementing team, university researchers, Canadian technology firms, project proponents and Indigenous data governance bodies. Each becomes invested in the system’s success. The project would also help build a public service trained in modern data tools, strengthen a research ecosystem focused on a meaningful applied problem, create a continuing role for an Indigenous data workforce, and provide a proof of concept for sovereign AI that can extend to health, agriculture, transportation and other regulatory domains.
Meaningful consultation with Indigenous peoples is a legal requirement for any project built in Canada that impacts treaty rights. Failure to consult properly is also the reason many projects get sent back to the drawing board. This is challenging work. It requires relationships, trust and sustained engagement from the outset. Technology will not change that.
Indigenous peoples have been sharing knowledge with regulators, governments and proponents for decades — much of it in oral testimony and in formats the existing data infrastructure was not designed to use. The CER’s BERDI project shows that a different approach is possible. The CER transcribed Indigenous oral testimony from pipeline assessments into structured data and returned the structured records to the communities that provided them. The data went into community hands, not into a federal repository (Canada Energy Regulator, n.d.).
That is the model the SIAA should follow at scale across all other types of assessments. The federated model proposed here is grounded in the OCAP® principles — Ownership, Control, Access and Possession of First Nations data by First Nations — which have long guided Indigenous data governance in Canada (First Nations Information Governance Centre, n.d.). The data would reside in community-controlled systems or vaults operated by each Nation. When an impact assessment requires Indigenous evidence relevant to a project area, the SIAA would query the relevant vault, retrieve only what has been authorized for release and return it to the federal evidence base on terms set by the community. Multi-generational ecological observations strengthen federal reviews without ever leaving community ownership. This transforms Indigenous Knowledge from a consultation checkbox into a high-integrity evidence base.
This matters more, not less, as governments seek to compress assessment timelines. Shortened reviews can only be substantively defensible — legally and morally — if the evidence base is already in place when the clock starts.
To date, the federal government has supported Indigenous data sovereignty through ad hoc project funding rather than systematic capacity-building. The SIAA would change that. Capacity-building must be substantive and sustained. Dedicated funding should train and employ Indigenous data officers within communities, finance the technical infrastructure (including servers) required to operate community data vaults, and support the multi-year work of building governance, standards and trust.
This is not pilot funding. It will proceed step by step, beginning with Nations who have already operationalized data governance and expanding as capacity builds — on terms set by each Nation, at the pace each Nation chooses. There is no universal timeline. The system must be patient with itself.
Where a proposed Crown Consultation Hub within the Impact Assessment Agency (Government of Canada, 2026) or a similar federal structure would co-ordinate who is consulted and how, the SIAA would complement that process rather than create a parallel structure. It would co-ordinate what Indigenous evidence can be queried, under what conditions and with whose consent. The SIAA would provide the technical infrastructure required to support meaningful, substantive consultation that goes beyond procedural compliance (Government of Canada, 2026).
The SIAA does not make decisions. It synthesizes evidence, including complex cross-sectoral data and Indigenous knowledge held under community control, so that departmental officials can make better-informed decisions. Legal accountability for every decision remains squarely with those officials.
This is not optional. The Treasury Board Directive on Automated Decision-Making requires it (Treasury Board of Canada Secretariat, 2025). The directive scales its requirements by impact level. As a system supporting impact assessment decisions, the SIAA would almost certainly fall within the higher tiers. As a result, final decisions would need to be made by a human, the system would require senior approval to operate, and an independent peer review by qualified experts from bodies such as the National Research Council or Statistics Canada would need to be published before deployment. The SIAA must be designed and approved within this framework from day one.
The defensibility this creates is not merely procedural. It is evidentiary. In Tsleil-Waututh Nation v. Canada (2018 FCA 153), the Federal Court of Appeal quashed the Trans Mountain expansion approval, finding a critical flaw in the regulator’s assessment of project-related marine shipping and inadequate Crown consultation. Federal evidence existed but had not been systematically brought into the assessment record. A standing, federated evidence base addresses that failure structurally. Decisions based on a complete, traceable, expert-reviewed record are far harder to overturn. This strengthens the government’s ability to defend them in court, reduces the regulatory risk premium for investors and protects good projects from challenges based on inadequate evidence or procedural gaps.
The same logic applies under compressed timelines: a faster process is not made defensible by political will or streamlining alone. It becomes defensible when the evidentiary foundation already exists before the review begins — a single foundation that serves federal and provincial regulators alike, while allowing each to make decisions under their own statutes and accountabilities.
Canada needs to build and to build faster. But speed cannot be added to an unreformed system. Compressing timelines on the current fragmented evidence base produces delays and litigation, not faster decisions. The SIAA changes the foundation those decisions rest on. It assembles structured federal data and brings in Indigenous knowledge on terms set by the communities that hold it. The data are ready before each review begins. The capacity to build it exists. The decision to build it does not yet.
Canada Energy Regulator. (n.d.). BERDI’s environmental and socio-economic assessment (ESA) data extraction method [Code repository]. GitHub. https://github.com/CER-REC/esa-data-bank_banque-donnees-ees
Canada Energy Regulator. (2026, January 27). BERDI. Government of Canada. https://apps2.cer-rec.gc.ca/berdi/
First Nations Information Governance Centre. (n.d.). The First Nations Principles of OCAP®. https://fnigc.ca/ocap-training/
Government of Canada. (2026, May 8). Simplifying Canada’s process for all major projects. One Canadian Economy. https://www.canada.ca/en/one-canadian-economy/services/simplifying-canada-process.html
Impact Assessment Agency of Canada. (2025a, June 26). Canadian Impact Assessment Registry. Government of Canada. https://iaac-aeic.gc.ca/050/evaluations
Impact Assessment Agency of Canada. (2025b, August 15). Generic requirements for impact statements. Government of Canada. https://www.canada.ca/en/impact-assessment-agency/services/policy-guidance/practitioners-guide-impact-assessment-act/generic-requirements-impact-statements.html
Impact Assessment Agency of Canada. (2026, April 29). Basics of co-operation in federal impact assessments. Government of Canada. https://www.canada.ca/en/impact-assessment-agency/programs/impact-assessments-101/cooperation.html
Innovation, Science and Economic Development Canada. (2026, June 1). AI Sovereign Compute Infrastructure Program. Government of Canada. https://ised-isde.canada.ca/site/ised/en/ai-sovereign-compute-infrastructure-program
Privy Council Office. (2026, May 4). About us — Major Projects Office. Government of Canada. https://www.canada.ca/en/privy-council/major-projects-office/about-us.html
Treasury Board of Canada Secretariat. (2025, June 24). Directive on automated decision-making (updated June 24, 2025). Government of Canada. https://www.tbs-sct.canada.ca/pol/doc-eng.aspx?id=32592
Tsleil-Waututh Nation v. Canada (Attorney General), 2018 FCA 153.
Westwood, A. R., Mines, S. M., Miglani, S., Sharan, R., Legault, A., Fitzpatrick, P., Sergeant, C. J., & Collison, B. R. (2025). Unearthing trends in environmental impact assessments for mines and quarries across Canada. FACETS, 10, 1-22. https://doi.org/10.1139/facets-2025-0114
This memo was commissioned as part of Capacity for Change: Designing a Public Service Built to Deliver, an IRPP research project that aims to provide solutions to barriers to state capacity through research-backed recommendations from leading experts. The publication was developed under the direction of IRPP President and CEO Jennifer Ditchburn and Professor Jennifer Robson, Director of Carleton University’s School of Political Management. It was copy-edited by Maya Lach-Aidelbaum, proofread by Zofia Laubitz, editorial
co-ordination was by Étienne Tremblay and production was by Chantal Létourneau.
Annette Hester is an economist, policy researcher and strategist with decades of experience in public policy, data governance, and energy and environmental affairs. She builds and leads multidisciplinary teams that translate complex data into usable insight for decision-makers. In Canada, she contracted with the Canada Energy Regulator on a data visualization project and helped build the team that later developed BERDI, the most advanced public-sector environmental and socio-economic data extraction system in the country. She has also worked with Natural Resources Canada on digital transformation and contributed to federal data and policy initiatives across several departments. Internationally, she served as senior adviser to the president of Petrobras, focusing on the geopolitics of the energy transition and the strategic use of energy and environmental data. She was previously region head for data governance and policy at Oxford Analytica. Her work spans Canada, Brazil and multilateral institutions across the Americas. She publishes at The Hester View (thehesterview.com).
This project was made possible in part thanks to support from the Max Bell Foundation and the Metcalf Foundation. The IRPP is an independent think tank, and retains control over the scope, methodology, conclusions and recommendations of our work.
A French translation of this text is available under the title Données plus accessibles, décisions plus rapides : Mise en place d’un agent souverain d’évaluation d’impact.
To cite this document:
Hester, A. (2026). From Fragmented Data to Faster Decisions: Building a Sovereign Impact Assessment Agent. Institute for Research on Public Policy. https://doi.org/10.26070/e82z-5076
This memo has undergone rigorous internal review. The opinions expressed in this memo are those of the author and do not necessarily reflect the views of the IRPP or its Board.
If you have questions about our publications, please contact irpp@nullirpp.org. If you would like to subscribe to our newsletter, IRPP News, please go to our website, at irpp.org.
Illustration: Auni Milne, Sumack Loft
