09/02/2026
EARLY WARNING, ARTIFICIAL INTELLIGENCE (AI), AND STRATEGIC COMMUNICATION: LESSONS FOR KENYA IN THE CONTEXT OF 2027 GENERAL ELECTIONS.
Muraya Justus
Nairobi, 2026.
As we approach the August 2027 general elections, anyone who has lived in Kenya knows that crises here rarely occur without warning.
Experience from past electoral cycles shows that violence, unrest, political radicalisation, and humanitarian stress are often preceded by subtle but observable social and narrative shifts at the community level.
Beyond narrative shifts, early warning is reflected in behavioural change, information distortion, institutional stress, economic adaptation, spatial reconfiguration, and the normalisation of exceptional language.
This may appear as reduced inter-group interaction, the spread of unverified information through closed channels, coded political signalling, inconsistent or hesitant institutional responses, anticipatory changes in livelihoods and markets, altered movement patterns, and growing tolerance for dehumanising or “inevitable” violence. Together, these signals often indicate rising anticipatory fear and informal mobilisation well before overt unrest or violence emerges.
These warning signals, while often subtle, become observable in the spaces where communities interact daily; markets, transport hubs such as boda boda stage, inside matatus, in places of worship, institutions of learning, social media, and ethnic broadcast networks.Frequently, these indicators appear well before formal incident reports are made.
Electoral risk rarely emerges suddenly. Early warning signals; behavioural shifts, information distortions, institutional stress, economic adaptations, spatial changes, and shifts in language, are often the first to appear. Whether knowingly or unknowingly, mainstream, community, digital, and social media pick up these signals, either propagating them or providing analytical commentary.
Even when communities and media pick up these early signals in forums and editorial commentary, the risk escalates when state institutions responsible for early warning systems such as the National Police Service, National Intelligence Service, National Government Administrative Officers - NGAOS, National Steering Committee for Peace Building, Conflict Resolution Uwiano , National Counter Terrorism Centre - Kenya and independent constitutions commissions such as National Cohesion and Integreation Commission repeatedly fail to detect, interpret, or act on the signals in a timely manner or when institutions operate in isolation leaving preventive options unutilised or uncoordinated.
In this regard, the challenge is not the absence of warning, but the abdication of the responsibility to protect (R2P) and the failure of institutions to act effectively while preventive options remain available despite the high cost of inaction, including escalating fear, social fragmentation, and eventual violence.
Kenya has invested significantly in early warning and prevention architecture, including the NCIC, NSC under the Ministry of Interior, devolved peace and security structures such as peace committees and County Engagement Forums (CEF), community policing programs, and security coordination mechanisms from the division all the way up to National Security Council. Yet past elections demonstrate persistent gaps in translating early warning into timely, coordinated and proportionate action.
This “prevention paralysis” reflects how independent commissions and state institutions mandated with the sacred Responsibility to Protect (R2p) are often constrained by political patronage, limiting their capacity to act early and decisively to tackle information distortions, institutional stress, hate speech and incitement.
As a result, harmful narratives may continue to escalate even when risks are already known, with formal intervention often occurring only once statements meet prosecutorial thresholds, by which time social harm may already be entrenched. This dynamic often pushes responses towards enforcement rather than prevention, meaning interventions focus more on legal or punitive measures and less on addressing risks early and de-escalating tensions before they take root.
This reflects institutional limitation rather than institutional absence, particularly in responding to fast-moving, informal, and locally coded narratives.
A key lesson from 2007, 2013, and 2017 general elections is that early warning in electoral contexts is often perceived as a purely technical domain reserved for IEBC and security actors alone, overlooking its narrative and community dimensions. The truth is, before violence occurs, narratives shift, fragmented grievances consolidate, rumours gain credibility, and fear of violence and displacement becomes normalised.
Regional electoral experience reinforces this lesson. In recent election cycles in neighbouring countries, including Uganda and Tanzania, early warning signals were often visible well ahead of polling day. In Kenya, it was evident that Tanzania’s securitised approach to early warning-focused primarily on enforcement and monitoring, could achieve short-term calm but did little to build trust in electoral institutions or prevent narratives from shifting into less visible, informal spaces. In Uganda, early signals were sometimes detected, but heavy-handed enforcement and politically motivated responses often limited the ability of communities and civil society to engage constructively, leaving tensions unaddressed and trust in institutions weakened.
As a result, in both Tanzania and Uganda, early warning systems were often perceived by communities not as tools of protection, but as instruments of surveillance. While short-term calm was in both cases achieved, trust deficits deepened and narrative contestation moved into less visible and less governable spaces, and that is where the long term danger early warning being perceived as surveillance rather than protection lies; the erosion of public trust, which drives grievances and rumours into hidden or informal channels, normalises tension and fear, and weakens institutions’ preventive capacity. Over time, early warning loses credibility, interventions become reactive rather than preventive, and societies become more radicalized against the state and more vulnerable to electoral violence and political manipulation.
These experiences underscore a critical lesson exemplified by Uganda and Tanzania: early warning systems should not be designed merely to produce quiet elections, but to ensure credible ones. Stability achieved through restriction is fragile and requires constant enforcement, whereas stability grounded in trust is far more durable and resilient.
Across the continent, contexts that preserved space for faith leaders, local media, youth networks, and civil society actors to engage constructively, such as Ghana, Senegal, Botswana, and Namibia, retained greater capacity to interrupt escalation early. In Ghana, coordinated efforts between the Electoral Commission, religious leaders, and civil society helped mediate tensions before and after elections. Senegal leveraged inclusive dialogue with faith leaders and local media to maintain peaceful transfers of power. Botswana and Namibia benefited from strong community networks that provided early signals of emerging grievances and facilitated preventive interventions.
Where these actors were constrained or excluded, institutions lost their earliest and most credible warning layer. For Kenya, these examples underscore the importance of integrating community-driven engagement with early warning and strategic communication, ensuring that preventive measures are not only technically effective but also trusted and credible.
Evidence from across electoral contexts also suggests that enforcement-first approaches shorten lead time. Once early warning is interpreted solely through a law-and-order lens, preventive choices disappear, leaving only coercive tools. In contrast, early deployment of strategic communication, timely, factual, and credible engagement, has proven capability of de-escalating tensions before confrontation occurs.
Artificial intelligence can provide limited but valuable support in this context. Tools for pattern detection, narrative tracking, and trend analysis can help distinguish emerging risks from background noise. Nonetheless, communities remain the most sensitive and immediate sensors of electoral risk, often detecting shifts in behaviour, language, and social alignment long before AI or formal systems can register them. Regional experience, including in Uganda and Tanzania, shows that when data-driven tools are applied primarily for monitoring and enforcement, without transparency, safeguards, or community engagement, they can erode trust and reduce preventive options to a narrow, state-centric security approach.
Detection without a defined preventive pathway has limited value. Effective preventive measures may include moderated radio discussions, coordinated factual clarifications through trusted local channels, and engagement by respected community, faith, youth, or civic leaders. Regional best practice shows that these interventions are most effective when pre-agreed, locally grounded, and activated before escalation, rather than reacting after incidents occur.
At present, strategic communication in many electoral settings remains reactive and weakly linked to early warning outputs. Comparative experience withbother jurisdictions points to the need for integration between;
a) Community-based early warning to capture narrative and social signals early;
b) Analytical support, including AI where appropriate, to validate and contextualise risk;
c) Pre-agreed response options, including communication measures, that can be activated before escalation.
Safeguards remain essential. Early warning must respect rights, avoid politicisation, protect community contributors, and ensure ethical information use. Most importantly, communities must observe that reporting risk leads to timely, proportionate, and confidence-building action.
As Kenya approaches the 2027 General Elections, the policy question is not whether early warning systems exist, they do. The more consequential question is whether institutions are structured, empowered, and coordinated to respond while preventive options are still available.
Picture: Evaluation of SYLO project at Industrial Area Police Station.
