Steven Backer

Steven Backer's public writing over the past year circles back to one conviction: legal AI in Africa fails the moment it substitutes structured data for actual legal understanding. He calls out the pattern directly, describing a compliance manager at a multinational who collects African laws into "a Microsoft folder, yes, of course neatly organised by country" and then asks an AI system to interrogate it. His verdict is blunt: "That is not legal intelligence. That is legal archaeology!" . This is not a one-off complaint but a structural argument he returns to repeatedly, that law is "a system: definitions, obligations, exceptions, thresholds, cross-references, powers, penalties and institutional logic all interact," and that an AI ignorant of that structure will produce answers that "may look impressive but still be legally fragile" . The risks he lists, "the wrong duty-bearer," "a missed exception," "a penalty not linked to the obligation," are not abstract; they are the concrete failure modes he expects buyers to interrogate before trusting any tool .

Out of this critique comes his clearest methodological contribution, the FACT process: "F — Find the law. A — Ascertain whether it is still in force. C — Check whether another law also applies. T — Talk to an in-country expert before assuming you got it right" . He is explicit that this is not a one-time checklist but an ongoing discipline, "FACT needs to run continuously," and frames skipping it as self-deception rather than progress: "Skipping the FACT process is not innovation, it is wishful thinking" . His tagline for the whole argument, "No FACT, no trust?" , doubles as a challenge to the wider legal AI industry, not just in Africa. The same instinct shows up in his advice to enterprise buyers evaluating vendors like Harvey: "make sure you also have a serious legal data partner," because whatever the AI produces "still needs to be grounded in the law" .

Backer's thinking about how companies should actually deploy this technology has become more architectural over time. Rather than answering the question he says he's most often asked, "Which legal AI tool should we buy?", he redirects clients toward "How should we design our legal AI architecture?" . He sketches this in layers: a document management layer (iManage, NetDocuments, SharePoint), an enterprise AI layer (Microsoft 365 Copilot, Google Gemini), and a legal AI workspace built on top . This layered view feeds directly into his concept of the "full legal intelligence stack," which he insists must include not just AI and structured primary sources but "contextual guidance," "continuous regulatory monitoring," "direct access to local legal expertise," and integration "through API and MCP" . He illustrates why data plus AI alone is insufficient with a live example from Afriwise's own system, a South Sudan telecom question where a yes/no legal answer existed but "the law alone does not give the answer" , reinforcing that local expertise remains indispensable even inside an AI-driven product.

A second consistent thread is his frustration with how slowly perceptions have shifted regarding the value of African legal expertise itself. He connects a recent conversation directly to one from "more than 20 years ago," when an international investor asked him, "Do we really need to take legal advice in Tanzania, is that even useful?" His response then and now is the same open question: "have we really not moved on?" . This is not incidental complaining, it's tied to his broader professional mission of correcting Africa's lag in legal tech adoption, a gap he had already documented through a continent-wide survey on legal tech adoption conducted in 2020 focused on corporate commercial legal needs and COVID-19 impact ¹⁰.

On the ecosystem-building side, Backer positions himself as an advocate for collaboration over competition among African legaltech builders. Reacting to a founder list that included both his own venture and companies he considers peers, he writes that "the ecosystem is maturing" and argues that "collaboration, whether through data contribution, API integration, or shared customers, is how we will really scale African legaltech" . He backs this with action, organizing "Demo Dozen" sessions with Legaltech Hub aimed squarely at African lawyers and in-house counsel, explicitly because "there simply aren't enough opportunities for legal professionals across Africa to see modern legaltech tools in action," given that "most vendor showcases are US-focused or Europe-first" . He frames the format, twelve vendors, fifteen minutes each, live software and real Q&A with "no slides, no pitch decks", as deliberately practical: "This is neither a trade show nor a sales funnel. It's efficient market exposure" .

Finally, Backer's account of his own work reflects the same values he preaches to others: grounding, patience, and giving time back to overworked legal teams. Describing months of "heads-down building" at Afriwise, he says the goal of the product is to "give legal and compliance teams their time back" , and he frames Afriwise's new agentic AI tool, Ask Lexiwise, and its MCP server as tools he's genuinely enthusiastic about, while still insisting "we cannot lose sight of reality" about what data and AI alone can and cannot deliver . Taken together, his output reads as a coherent, repeated argument: build AI, but never let it forget that African law is a system, not a folder, and that trust has to be earned through process, not promised through automation.