Briefing · SDWH Board Briefing #1
Governance in an AI-Native Society
Boards are rightly focusing on how their organisations should adopt AI safely and responsibly, monitoring discovery and experimentation and assessing maturity. Looking ahead: how should organisations be governed as society adopts AI at varying rates, or not at all?
This briefing sets out five governing principles for boards navigating an AI-native society.
- Principle 1 Capability has become distributed
- Principle 2 Institutional advantage has changed
- Principle 3 Stakeholder capability is becoming uneven — not everyone benefits equally
- Principle 4 Governance processes must evolve
- Principle 5 Boards need new questions
The AI-native governance landscape
Figure 1: AI-native governance landscape
| Institutional capability | Distributed capability | |
|---|---|---|
| Opportunity | AI-enabled organisations delivering better services, faster decisions, redesigned operating models and improved assurance. | AI-native stakeholders — customers, citizens, suppliers, investors, journalists, campaigners, researchers and SMEs — contribute richer evidence, stronger challenge, collaborative innovation and more informed participation. |
| Governance challenge | Legacy operating models, fragmented information, vendor dependency, organisational inertia and knowledge trapped in silos. | AI-assisted legal challenge, consultation overload, coordinated influence campaigns, sophisticated procurement scrutiny, and rapidly increasing stakeholder capability. |
How should governance evolve in an AI-native society?
Board questions
Figure 2: Board questions by theme
| Theme | Governance questions |
|---|---|
| The wider stakeholder environment | How will AI change the behaviour and expectations of those we serve? How are AI-native stakeholders currently engaging with the organisation? Who may be left behind as AI adoption accelerates? |
| Governance | How should governance itself evolve? |
| Strategy | How is AI changing the environment in which we operate? What are our competitors or challengers doing? |
| Organisational capability | Which capabilities remain genuinely distinctive? |
| Knowledge assets | What organisational knowledge are we creating — and where does it reside? |
| Ecosystem resilience | How portable are our AI capabilities? Where are we becoming dependent upon particular suppliers? |
Governance implications
AI maturity varies. No two organisations currently look alike. Governance frameworks need to stay on top of this rapid pace of change. The immediate challenge for boards is therefore one of orientation rather than optimisation.
Good governance manages risks within the organisational appetite and embraces opportunity. Governance has always depended on understanding how information flows through a system. AI is changing those information flows more quickly than many governance models are changing with them. The governance agenda is rightly focused on strategy, organisational capability and knowledge assets. Boards must now recognise that AI is changing not only the organisation itself, but the wider environment within which it operates.
Figure 3: Risks and opportunities for ARAC and the board
| Risk management | Commercial opportunity | Board agenda |
|---|---|---|
| Data quality and provenance; security and third-party risk; governance of pilots and scaling; guardrails and accountability; workforce implications including specialist skills, training and adoption. An ARAC agenda more holistic than the established risk management framework, evolving through the technology lifecycle (AI as a successor to cloud, cyber, ERP and major digital programmes). | Productivity, efficiency, quality, innovation. | Reputation, legitimacy, target operating model, resilience, sustainability. |
Distributed analytical capability
Why boards need to rethink governance when analytical capability is no longer confined to large institutions
AI is reducing the minimum efficient scale of knowledge work. Analytical capability that was once concentrated in governments, large corporations and professional firms is becoming widely distributed across customers, citizens, journalists, campaign groups, investors, competitors and SMEs.
AI-enabled analysis and interpretation is now available to the experienced professional and the enthusiastic amateur alike, through multiple agentic workflows under full human oversight — automating provenance, version control, reproducibility, workflow, publication, challenge, peer review and stakeholder engagement. The evidence assembly and analytical workflow can be substantially automated, allowing effort to be concentrated where it creates the greatest value.
Figure 4: AI-enabled analysis and interpretation
| Historically, sophisticated analysis required | Today | Institutional advantage increasingly derives from |
|---|---|---|
| Large analytical teams; specialist software; institutional knowledge; significant research budgets; long production cycles; intelligent customers to procure and manage complex technology delivery. | Small expert teams — and in some cases individuals — can assemble evidence, reverse-engineer presentational veneers, analyse complex systems and publish high-quality outputs using publicly available information supported by AI. Data asset management policies and legacy systems continue to process and present corporate information. | Judgement; trust; proprietary knowledge and unique technologies; access to, or ability to generate, primary evidence; effective execution and agility rather than simply organisational scale. |
Inclusion and the AI paradox
AI reduces the cost of sophisticated analysis. But it does not distribute that capability equally. Current themes include risks of misinformation and model hallucinations. At the same time, AI creates a new spectrum for stakeholder capability, condensed below into three broad bands. That has profound implications for governance. Governance has always sought to promote inclusion. Now inclusion has another dimension: not simply access to digital services, but access to analytical capability. This has implications for how directors demonstrate their Section 172 duties under the Companies Act, and for effective corporate governance throughout the public bodies landscape.
Figure 5: Spectrum of stakeholder capability and representation
| Stakeholder capability | Characteristics | Governance implication |
|---|---|---|
| AI-native stakeholders | Use AI to understand policy, compare evidence, draft submissions, analyse data and challenge decisions. | Rich source of insight and innovation, but capable of sophisticated scrutiny. |
| Occasional AI users | Use consumer AI intermittently with mixed confidence and variable quality. | Need guidance, transparency and clear engagement routes. |
| Digitally excluded or disengaged stakeholders | Limited access to digital tools or confidence to use them; may disengage entirely. | Risk of under-representation and decisions that fail to reflect the wider community. |
Example 1: the AI-native parish councillor as a stakeholder
| Observation | Governance principle | Board implication |
|---|---|---|
| An 86-year-old parish councillor, frustrated with voluminous paperwork and a lack of organisational support, now uses AI to analyse planning papers, compare policies, prepare for council meetings and challenge or support local authority decisions. | AI is reducing the expertise and effort required for effective participation in public decision-making. AI compresses huge volumes of information from a variety of sources, but can be unreliable and prone to hallucinations. | Governance should embrace informed external challenge that is more capable, more frequent and more evidence-based. Governance needs to include occasional users and digitally excluded stakeholders. |
Stakeholder engagement and consultation
Example 2: stakeholder engagement and consultation
Board issue: does our stakeholder consultation and engagement model still achieve its intended purpose in an AI-native environment? Boards ask: are our services digitally accessible? They should also ask: who is benefiting from AI, and who is being left behind? This isn’t simply an equality issue — it is also about legitimacy. If consultation increasingly reflects those with AI capability, boards need to ask whether they are still hearing from the wider population.
Assumptions that are changing
Many stakeholder consultation processes evolved in a world where expertise was relatively scarce, participation was naturally limited, analytical capability was concentrated, institutional knowledge was difficult to replicate, and external challenge was costly.
This creates new opportunities: richer public consultation, more informed customers, stronger stakeholder engagement, better-informed boards, faster organisational learning, and wider innovation.
It also creates new pressures: greater scrutiny, more sophisticated legal challenge, increased expectations of transparency, higher volumes of stakeholder interaction, and accelerated competitive change.
In this context, AI-native governance is not simply about increasing the capability of institutions. It is about increasing the capability of institutions and stakeholders to interact with one another effectively, transparently and inclusively.
Stakeholder engagement and consultations are a key performance indicator for boards and inform evaluations of executive performance. Alongside direct discussions, written consultation processes have typically been designed as a linear, gateway process: initiated centrally, broadcast to stakeholders across a wide spectrum of engagement models, posing questions within themes, and measuring the volume and content of responses to establish the direction of future activity. These processes now need to contend with how AI is changing the capability of participants, the quality, provenance and diversity of evidence submitted, and the agentic ecosystem enveloping the entire process — with the capacity to become a continuous feedback loop, both an opportunity and a risk.
Figure 6: How the AI-native stakeholder engages in consultations
| Traditional consultation model | AI-native consultation |
|---|---|
| Implicit assumptions: consultation teams can manually analyse submissions; responses are individually authored; producing a detailed response is relatively expensive; large organisations have greater analytical capability than most respondents; response numbers provide a useful proxy for stakeholder sentiment; stakeholder consultation is periodic, with concerns over engagement fatigue, pauses, deferrals and gaps. | Consultation organisers can use AI to cluster themes, identify genuinely novel arguments, detect duplicate or coordinated submissions, trace evidence sources, and produce more transparent summaries. Respondents can analyse consultation papers, compare previous consultations, identify inconsistencies with legislation, benchmark against other jurisdictions, generate multiple evidence-based alternatives, and collaborate with others at very low cost. |
Wider implications: this is broader than public consultations. It applies to customer engagement, complaint handling, FOIA requests, whistleblowing, procurement and contract/supply chain management, investor relations, statutory/regulatory/planning consultations, legal challenge, ombudsman processes and regulatory procedures.
AI maturity and the board agenda
Governance is maturing from Governing AI towards Governing organisations operating within an AI-native society.
Most current governance activity has prioritised organisational maturity. It asks: which AI are we using? Is it secure? Is the data protected? Are staff following policy? Have we controlled risk? What is the gap we need to fill? Where do we need to be?
Existing governance and risk management frameworks — the three lines model, government’s Risk Management Framework (the Orange Book) — remain essential. The question is not whether they remain valid; it is whether the assumptions surrounding them have changed.
Figure 7: Governance maturity (status in 1H 2026) and the current board agenda
| Governance maturity (status in 1H 2026) | Current board agenda |
|---|---|
| AI inventories and use cases; AI principles and ethics; data governance, master data management, data quality, metadata and Open Data initiatives, GDPR interactions; AI overlay and access to core and legacy systems; automation of traditional workflows; licensing strategies; shadow AI; major enterprise implementations; monitoring and managing token spend, including incentives and limits; procurement and contract management; emerging agentic workflows; datasets for sovereign AI / compute use cases; national security and cyber security implications of model availability and adoption in critical national infrastructure; management oversight and control, AI assurance and internal audit; board reports and governance processes. | Duty of candour for public bodies; sovereign AI; vendor concentration; knowledge ownership; prompt portability and ownership; AI-enabled procurement and shared services; benefits realisation; evaluations and lessons learned; continuous assurance; AI-enabled organisational memory; autonomous agents; autonomous procurement and legal challenge; AI-assisted consultation and democratic participation; agentic systems; machine-to-machine governance; synthetic stakeholders; business processes. |
Draft briefing. This page presents SDWH Board Briefing #1 (17 July 2026, v1.0) as a working draft. Content, structure and figures are subject to revision. This analysis is produced by SDWH Limited for information purposes only and does not constitute legal, governance or regulatory advice.