Erica’s keynote at the 2026 OTT Conference explored what becomes more valuable as artificial intelligence makes information increasingly abundant. Reflecting on the relationship between AI, human judgement and policy research, she argued that the distinctive contribution of think tanks lies not only in the production of evidence, but in the encounters, relationships and forms of trust through which meaningful knowledge is created.
Her keynote at the 2026 OTT Conference explored a question increasingly confronting think tanks across the world: what happens to policy research, institutional trust, and human judgment in an era shaped by artificial intelligence?
Rather than framing AI as a purely technological challenge, the keynote approached it as a deeper institutional and epistemological question. At its core was a simple but powerful argument: the most valuable knowledge think tanks produce does not emerge from isolated expertise alone, but from encounters between people working through difficult problems together.
Erica opened with a reflection on a workshop led by Senzi Bengu at the 2025 OTT Conference in Johannesburg. During the session, Bengu remarked that “all the knowledge we need is in this room”.
At first, the statement appeared to affirm the value of practitioner expertise already present among participants. But over time, Erica explained, the meaning became more profound.
The knowledge needed for today’s challenges does not fully exist yet. It is created through interaction itself — in moments of disagreement, critique, negotiation, and collective problem-solving.
Whether through a colleague questioning an analytical framing, a stakeholder reframing a policy problem, or a difficult discussion that produces a new shared understanding, something important emerges through the encounter that did not exist beforehand.
Knowledge produced through encounter
Erica argued that this form of knowledge production appears consistently across multiple intellectual traditions.
Democratic theorists studying deliberation, economists working on dispersed knowledge, and empirical researchers examining collaborative decision-making all arrive at similar conclusions: certain forms of knowledge are produced only when people with different stakes engage directly with one another.
Some encounters are highly formal and institutional. Others are embedded quietly in everyday research processes.
Erica challenged audiences to think beyond the obvious examples of conferences and convenings. Encounter, Erica argued, is woven throughout the entire production of policy research itself. It appears in the framing of a research question, in disagreements over methodology, in editorial feedback, and in conversations with stakeholders whose experiences reshape how problems are understood.
Even analytical outputs that appear to be the work of a single author are often saturated with encounters that shaped the final judgement.
“The room,” the Erica suggested, “is bigger than you think.”
This argument also extended the long-standing discussion about the value of think tanks.
Referencing the work of OTT founder Enrique Mendizabal, Erica acknowledged previous arguments that think tanks’ value lies in their usefulness. Erica added a further dimension: usefulness depends not only on outputs themselves, but on the human encounters that produce judgment, trust, and shared understanding along the way.
AI as a powerful research partner
Erica did not reject artificial intelligence. On the contrary, she described AI as perhaps the most powerful research partner most think tanks have ever had.
AI systems can analyse and synthesise enormous quantities of information at extraordinary speed, dramatically expanding the analytical capacity of even relatively small organisations.
One example came from Stanford University’s RegLab, where researchers used AI to analyse 5.2 million property deeds for racially restrictive covenants. The work, which would previously have required decades of labour and vast financial resources, was completed in only days at minimal cost. That analysis contributed directly to legislative changes in California requiring the redaction of discriminatory covenants.
Yet Erica argued that AI’s success in such cases still depends fundamentally on prior human work.
Before AI could identify racial covenants, generations of activists, researchers, and communities had already constructed the legal, political, and ethical understanding that made the category meaningful. Humans decided that the issue mattered, developed the concepts, and determined how the findings should be interpreted and used.
“AI,” she argued, “did the complexity work.” Humans created the meaning.
What becomes scarce in the age of AI
One of Erica’s most important observations was that AI is making previously invisible forms of value easier to recognise.
If AI can now generate many traditional research outputs quickly and efficiently, then the real value of think tank work cannot lie solely in producing reports, policy briefs, or white papers.
Instead, value also exists in the process surrounding those outputs: the judgment formed through difficult analytical work, the relationships built during research, and the shared understanding created through collective engagement.
For years, much of this labour remained invisible because the final deliverable absorbed most institutional attention.
“AI is what makes it legible.”
This reframes the debate around artificial intelligence.
The key question is no longer simply whether think tanks should adopt AI tools. Rather, it is how AI fits into the broader ecology of human judgment, institutional trust, and policy engagement.
The risk of thinner knowledge
A recurring concern throughout Erica’s keynote was the possibility that AI could quietly displace not only administrative tasks, but also the very encounters through which knowledge and professional judgment are formed.
The visible displacement is relatively easy to recognise: skipped consultations, cancelled convenings, or accelerated research processes.
The more subtle displacement may be far more consequential.
What happens when difficult framing decisions are no longer worked through collectively? What happens when junior analysts no longer develop judgment through literature reviews, stakeholder discussions, editorial disagreements, and iterative drafting because those tasks are increasingly delegated to AI systems?
Erica warned that the final output may still appear sophisticated while the underlying intellectual process becomes significantly thinner.
This concern became especially clear in a discussion of professional formation within think tanks.
Junior researchers traditionally build judgment through years of cognitive and relational work: drafting memos, debating evidence with colleagues, observing stakeholder interactions, and learning how policy decisions unfold in practice.
If AI performs large portions of this foundational work automatically, institutions risk losing the developmental pathways that produce future expertise.
“The design question,” Erica argued, “is what replaces it.”
AI, mediation, and the changing policy environment
In her keynote, Erica also explored how AI is reshaping the wider information ecosystem within which think tanks operate.
Drawing on debates within futures studies, the speaker distinguished between augmentation and mediation.
AI augments human capabilities by helping individuals process information more effectively. But it also mediates how knowledge reaches audiences.
Search engines and AI assistants increasingly synthesise information directly for users, meaning policymakers may engage with AI-generated summaries or recommendations without ever encountering the think tank or institution that originally produced the underlying research.
In what Erica described as a “zero-click world”, the traditional policy brief may no longer serve as the main gateway to influence.
This shift raises new questions about visibility, attribution, and institutional credibility.
If the final document no longer guarantees meaningful engagement with the institution behind it, then the deeper relational and deliberative work of think tanks may become even more important.
Trust, judgment, and the future of think tanks
Despite the rapid advance of AI capabilities, the Erica insisted that certain forms of human responsibility remain irreplaceable.
AI systems cannot ultimately take responsibility for what information means or for the consequences that follow from policy decisions. Human beings still stake their names, judgments, and institutional credibility on those choices.
At their best, think tanks function as spaces where people with different interests, perspectives, and stakes encounter one another’s reasoning and develop enough trust to act collectively.
That trust is itself a product of encounter.
Importantly, the Erica stressed that this work is not confined to traditional policy centres such as Washington. Think tanks across cities, including Rabat, Nairobi, and Budapest, perform similar deliberative functions under very different political conditions.
The institutional forms may evolve dramatically in the coming years. But the underlying function — creating spaces where rigorous and accountable encounters can produce socially useful knowledge — remains essential.
Designing the encounters worth keeping
Erica concluded with a design challenge for think tanks.
Institutions must become far more deliberate about identifying which forms of human encounter should be preserved, which can be redesigned, and which new forms of engagement emerging technologies require.
Erika proposed a practical exercise: imagine that AI had produced a think tank’s most recent deliverable overnight. What would be missing?
The answer, she argued, reveals the institution’s “invisible product” — the judgment formed, the relationships built, and the trust generated throughout the research process.
Those invisible products are often where the deepest value of think tanks resides.
Erica suggested that future organisations may need to think carefully not only about how humans use AI, but also about how communities, stakeholders, and lived experiences become visible within AI-mediated systems themselves.
The session closed by returning to Senzi Bengu’s original reflection from Johannesburg.
The knowledge needed for today’s complex policy challenges does not exist fully formed in advance. It is produced repeatedly whenever people with different stakes work through difficult questions together and arrive at understandings they can collectively live with.
The task for think tanks, the keynote concluded, is to design intentionally for those encounters in the age of AI.