[{"data":1,"prerenderedAt":429},["ShallowReactive",2],{"home-selected-work":3},[4,161,276],{"id":5,"title":6,"anonymized":7,"body":8,"client":125,"cover":126,"description":116,"draft":7,"engagement":127,"extension":128,"liveUrl":129,"meta":130,"metrics":131,"navigation":147,"order":148,"path":149,"role":150,"seo":151,"stack":152,"stem":157,"summary":158,"year":159,"__hash__":160},"caseStudies\u002Fwork\u002Fpayd.md","Banking The Global Gig Economy From Nairobi",false,{"type":9,"value":10,"toc":115},"minimark",[11,16,20,23,26,30,33,49,52,56,64,67,70,73,77,80,88,91,95,98,105,112],[12,13,15],"h2",{"id":14},"context","Context",[17,18,19],"p",{},"The problem that started Payd was personal. Benaiah was a freelancer working with clients in Zurich, Berlin, and New York while banking from Nairobi. Every invoice lost 15 to 30 percent to correspondent banking fees, exchange spreads, and wires that took three to five business days to clear – if they cleared at all. He was not alone. There are an estimated 400 million independent workers globally, and a disproportionate share of them live in markets where the cost of getting paid is a tax on their existence.",[17,21,22],{},"The obvious consumer play – \"a better Wise for Africa\" – was crowded and hard. The less obvious play was that the same infrastructure needed for a great consumer product was infrastructure that growing businesses needed even more. Agencies paying distributed contractors. Companies running global payroll across M-Pesa, MTN, SWIFT, and SEPA in the same transaction batch. Platforms needing to move money for their users without becoming a bank.",[17,24,25],{},"Payd could either be a consumer fintech competing on UX, or an infrastructure company that happens to have a beautiful consumer front-door. We chose the second.",[12,27,29],{"id":28},"moved-fast","Moved Fast",[17,31,32],{},"In the first ninety days we shipped what a bigger team would have scoped as a six-quarter roadmap:",[34,35,36,40,43,46],"ul",{},[37,38,39],"li",{},"USD and EUR virtual accounts with dedicated account numbers, wired into SWIFT, SEPA, and ACH through a licensed partner network.",[37,41,42],{},"Direct mobile money integrations with M-Pesa and MTN – the most-used rails on the continent – so a freelancer in Lagos could receive a wire from Munich and spend the proceeds at a corner kiosk in Ikoyi by end of day.",[37,44,45],{},"Stablecoin settlement rails on ten chains, giving us a low-cost cross-border layer that settled in seconds and let clients choose between fiat and digital dollar delivery depending on market conditions.",[37,47,48],{},"A payment-link primitive that let anyone generate a branded checkout in thirty seconds and receive from any of the above channels.",[17,50,51],{},"The rule we ran was \"ship something in production every Friday, or explain what broke.\" That cadence did two things. It forced decisions. And it compounded – every shipped primitive was a lego brick the next week could stand on.",[12,53,55],{"id":54},"failed-forward","Failed Forward",[17,57,58,59,63],{},"The B2C-first hypothesis almost killed us. We spent a full quarter optimising consumer onboarding – shaving it from 40 minutes to 3 – and celebrated each milestone. The metric we should have been watching was not signups. It was ",[60,61,62],"em",{},"volume per user",". B2C users send small amounts infrequently. The unit economics do not close at under-$50 average transaction sizes when payment partners charge fixed minimums.",[17,65,66],{},"We learned this the hard way. Two cohorts of acquisition spend produced great vanity metrics and a P&L that refused to work. The failure was useful because it was sharp. We did not have to argue about it. The data was the data.",[17,68,69],{},"The pivot was to stop chasing consumers and start chasing the businesses paying them. Agencies, marketplaces, platforms, and payroll operators all had the same pain we were trying to solve for their workers – but at 50x to 500x the average ticket size. Same infrastructure. Same integration surface. Completely different unit economics.",[17,71,72],{},"Within two quarters, B2B and B2B2C volume crossed 65 percent of total. Six months after that, it crossed 70 percent. The consumer side remained – and remains – because it validated the rails and kept us honest about end-user experience. But it stopped being the business.",[12,74,76],{"id":75},"built","Built",[17,78,79],{},"What exists today:",[17,81,82,83,87],{},"A ",[84,85,86],"strong",{},"money OS for borderless work",". USD and EUR virtual accounts open in three minutes with no paperwork. Stablecoin settlement across ten chains. Direct mobile money across the continent. A business suite covering global payroll (batch payments with single approval, automated scheduling, multi-channel delivery per country), spend management with role-based access and real-time tracking, and invoicing with customisable templates and automated reminders. A WhatsApp-native interface that lets users send, receive, and manage their account from the app they already spend four hours a day inside.",[17,89,90],{},"Partnerships with regulated payment providers and virtual asset service providers across every corridor we operate in, so Payd is compliant by construction – not retrofit.",[12,92,94],{"id":93},"compounds","Compounds",[17,96,97],{},"Payd is the flagship of B3n Labs not because it is the biggest product on the studio's books, but because it proves what Labs is for. Every month, two things get stronger on their own without us pushing them.",[17,99,100,101,104],{},"First, ",[84,102,103],{},"distribution compounds",". Every business customer brings 50 to 5,000 end users, who become their own viral loop. The WhatsApp interface did not just improve retention – it turned every user into a distribution channel because referring a friend now happens in the same thread they send voice notes in.",[17,106,107,108,111],{},"Second, ",[84,109,110],{},"rails compound",". Each new corridor, each new mobile money operator, each new stablecoin added to settlement makes every existing customer more valuable. A payroll client signed twelve months ago is 3x more useful to themselves today than they were at signing, without us shipping them a single product update.",[17,113,114],{},"This is what Labs is supposed to produce. Not a product. A compound.",{"title":116,"searchDepth":117,"depth":117,"links":118},"",3,[119,121,122,123,124],{"id":14,"depth":120,"text":15},2,{"id":28,"depth":120,"text":29},{"id":54,"depth":120,"text":55},{"id":75,"depth":120,"text":76},{"id":93,"depth":120,"text":94},"Payd","\u002Fbrands\u002Fpayd.png","labs","md","https:\u002F\u002Fpayd.money",{},[132,135,138,141,144],{"label":133,"value":134},"Monthly transaction volume","$3.5M+",{"label":136,"value":137},"Countries covered","35+",{"label":139,"value":140},"Active mavericks","30,000+",{"label":142,"value":143},"Fee reduction vs banks","Up to 70%",{"label":145,"value":146},"B2B share of volume","70%+",true,1,"\u002Fwork\u002Fpayd","Founder, Chief Maverick",{"title":6,"description":116},[153,154,155,156],"USD and EUR virtual accounts (ACH, Wire, SEPA)","Multi-chain stablecoin infrastructure","Mobile money rails (M-Pesa, MTN)","WhatsApp-native payment flows","work\u002Fpayd","How we built cross-border financial infrastructure serving 30,000+ freelancers and businesses across 35+ countries – and shifted the business from B2C to B2B-led.",2026,"S78BMRwqlaDvfbsKR3i5zIMJSWYWYtxFbpTkfSWhuH4",{"id":162,"title":163,"anonymized":7,"body":164,"client":243,"cover":244,"description":116,"draft":7,"engagement":245,"extension":128,"liveUrl":246,"meta":247,"metrics":248,"navigation":147,"order":120,"path":264,"role":265,"seo":266,"stack":267,"stem":272,"summary":273,"year":274,"__hash__":275},"caseStudies\u002Fwork\u002Fcyphon.md","Building An AI Agent Practice From Zero To $100K+ Monthly Value",{"type":9,"value":165,"toc":236},[166,168,171,174,177,179,182,185,188,190,193,196,217,220,222,225,228,230,233],[12,167,15],{"id":14},[17,169,170],{},"In late 2023, Cyphon was a small team with good engineers, a big idea, and no operating spine. The thesis was right – AI agents were about to replace not just tasks but entire roles inside professional services firms, and someone needed to be the implementation partner for enterprises that could not build it themselves. But thesis without delivery is a pitch deck, and pitch decks do not send invoices.",[17,172,173],{},"The engagement started with a specific problem. Cyphon's parent company ran client delivery with 10 project managers, 5 QA reviewers, and 70+ analysts. The margin was getting thinner every quarter. Every new client pulled linearly on staffing. The team sensed AI could unlock the equation, but did not know where to start, what to build first, or how to turn experiments into a billable product.",[17,175,176],{},"What they needed was not a strategy. They needed someone to install operating structure while the team kept shipping.",[12,178,29],{"id":28},[17,180,181],{},"Week one, we picked a single workflow that would never make a PR splash but would prove the model: outbound lead welcome emails. PMs were spending 20 minutes per email, drafting in a blank screen, re-using their best lines inconsistently. We built an agent trained on ten years of their top-converting welcome emails, wired it into the CRM, and shipped it to five PMs as a pilot.",[17,183,184],{},"Within two weeks, welcome email turnaround dropped by 90 percent and response rates rose 38 percent. We had our first internal case study. More importantly, we had a pattern: pick a workflow that is costly, repetitive, and already has training data in the client's own history. Ship an agent in two to four weeks. Measure against a hard before\u002Fafter. Expand.",[17,186,187],{},"By week eight, five agents were in production across the parent firm – welcome emails, proposal generation, QA review, lead nurture, and inbox triage. The same team of 85 was servicing the same book of business with two PMs, ten analysts, and zero dedicated QA staff. Margin opened up by a factor we are not at liberty to publish.",[12,189,55],{"id":54},[17,191,192],{},"The first external client engagement was a mess. We had proven the pattern internally, and we walked into a Fortune 500 consumer goods company assuming the same playbook would land. It did not. The client's internal data was not clean. The workflows we scoped turned out to be three workflows in a trench coat. And the champion who hired us moved teams eight weeks in.",[17,194,195],{},"We lost two months and a healthy chunk of scope. The failure taught us three things that now shape every Cyphon engagement:",[197,198,199,205,211],"ol",{},[37,200,201,204],{},[84,202,203],{},"Discovery is delivery."," We now spend the first two weeks of every engagement mapping the actual workflow – not the org chart's version of it – before any agent is scoped. Cheaper to change a plan than an agent.",[37,206,207,210],{},[84,208,209],{},"Champions churn."," Every engagement now onboards at least two stakeholders from day one, with explicit escalation paths. Single-champion engagements are refused.",[37,212,213,216],{},[84,214,215],{},"Training data is the work."," If the client does not have historical workflow data we can train against, phase one is setting up that collection. No shortcuts. Without data, the agent is a chatbot.",[17,218,219],{},"The second external engagement – a mid-market SaaS company – closed in six weeks and shipped three agents. The third closed in four. The pattern that emerged from the failure is now the firm's playbook.",[12,221,76],{"id":75},[17,223,224],{},"In fourteen months we went from two-person experimental team to a 14-person AI agents consultancy with 40+ enterprise clients across finance, healthcare, SaaS, logistics, and consumer goods. 150+ agents deployed in production. A repeatable four-phase delivery model: Discovery & Consultation → Design & Development → Secure Deployment → Ongoing Support. Model-agnostic architecture running on OpenAI, Claude, Gemini, Grok, and private deployments as the use case demanded. OpenClaw framework integration for skill-based, memory-carrying, tool-using agents that operate inside live workflows rather than beside them.",[17,226,227],{},"Internally: a full leadership team with a CTO, a COO, a Chief Experience Officer, a Chief Creative Officer, and seven directors spanning ML, delivery, research, security, analytics, prompt engineering, and predictive modelling. Hiring pipeline, delivery pipeline, sales pipeline, and quality pipeline – all documented, all running.",[12,229,94],{"id":93},[17,231,232],{},"The thing that makes this case study compound is that the work did not end when we handed the COO seat off. Every process we installed – the discovery-to-delivery pattern, the dual-stakeholder requirement, the training-data phase, the internal-first validation model – is still running. Cyphon's growth since we stepped back has been higher, not lower. That is the test of an operator engagement: can the system run without you?",[17,234,235],{},"The second compound: the internal use of Cyphon's own agents inside the parent firm now runs a client base that would have needed 70+ analysts with 10 analysts. The savings throw off cash that funds R&D on the next generation of agents. Every dollar saved internally pays for the agents we sell externally. That is a flywheel that a strategy deck would never have built – it had to be installed.",{"title":116,"searchDepth":117,"depth":117,"links":237},[238,239,240,241,242],{"id":14,"depth":120,"text":15},{"id":28,"depth":120,"text":29},{"id":54,"depth":120,"text":55},{"id":75,"depth":120,"text":76},{"id":93,"depth":120,"text":94},"Cyphon AI","\u002Fbrands\u002Fcyphon.png","transformation",null,{},[249,252,255,258,261],{"label":250,"value":251},"Monthly value unlocked for clients","$100K+",{"label":253,"value":254},"AI agents deployed","150+",{"label":256,"value":257},"Enterprise clients served","40+",{"label":259,"value":260},"Workload reduction across firm","70%",{"label":262,"value":263},"Delivery speed improvement","4x","\u002Fwork\u002Fcyphon","Chief Operating Officer",{"title":163,"description":116},[268,269,270,271],"OpenClaw agent framework","Multi-model orchestration (OpenAI, Claude, Gemini, Grok)","Custom prompt engineering pipeline","Client data ingestion and integration layer","work\u002Fcyphon","COO and build-partner engagement that architected an AI agents consultancy from a team of two to a 14-person practice serving 40+ enterprise clients.",2024,"QKDivq_CRMh_0pfAtANtnFEsaW1PMJqQcokyGql15lE",{"id":277,"title":278,"anonymized":7,"body":279,"client":400,"cover":401,"description":116,"draft":7,"engagement":402,"extension":128,"liveUrl":403,"meta":404,"metrics":405,"navigation":147,"order":117,"path":417,"role":418,"seo":419,"stack":420,"stem":425,"summary":426,"year":427,"__hash__":428},"caseStudies\u002Fwork\u002Fnaiban.md","A Brand And Digital Refresh For Kenya's Most Active Angel Network",{"type":9,"value":280,"toc":393},[281,283,286,289,292,294,297,329,331,334,345,348,350,353,373,375,378,384,390],[12,282,15],{"id":14},[17,284,285],{},"NaiBAN is Nairobi's – and arguably East Africa's – most active angel network. Over 100 angels, 40+ portfolio companies, and a track record of backing serious businesses including Payd, BuuPass, Rescue.co, Zeraki, Uncover, and Flowcart. But the public face of the network did not match the substance. The live site at naiban.co was a single-page WordPress template. Three menu items. A stock image of the Nairobi skyline. A quote graphic. No portfolio. No application flow. No resources. No investor onboarding.",[17,287,288],{},"The network was absorbing top-tier deal flow through reputation alone. The website was, in practice, a liability – the kind of surface where a serious founder landing from a warm intro might quietly downgrade their read of the organisation before the first meeting.",[17,290,291],{},"The brief was short: make the digital face of NaiBAN match what NaiBAN actually is.",[12,293,29],{"id":28},[17,295,296],{},"We scoped and shipped the rebuild in four weeks. The core moves:",[34,298,299,305,311,317,323],{},[37,300,301,304],{},[84,302,303],{},"Positioning reset."," \"100+ active business angels\" was the fact, but \"100+ active business angels actively supporting startups with advice, connections, and capital\" was the story. The hero re-framed NaiBAN from a passive network to an active one – which is what founders actually experience when they get in.",[37,306,307,310],{},[84,308,309],{},"Portfolio as proof."," The portfolio was the single most credibility-restoring asset in the whole site. We built a proper portfolio index with company profiles, sector tags, and visible logos. Seeing Payd, BuuPass, and Zeraki on the same page ended the \"who are these guys\" question in under ten seconds.",[37,312,313,316],{},[84,314,315],{},"A real application flow."," Replacing a generic contact form with a structured pitch application that captures stage, sector, traction, and ask up front. Gave the screening committee a clean intake. Gave founders a clearer sense of what the network expects.",[37,318,319,322],{},[84,320,321],{},"Investor onboarding path."," Separate call-to-action for angels interested in joining the network. The old site had none.",[37,324,325,328],{},[84,326,327],{},"Resource layer."," A blog \u002F resources section for the network to publish perspective and insight, both as SEO fuel and as a way to let the network's voice extend beyond the portfolio.",[12,330,55],{"id":54},[17,332,333],{},"We shipped a brand palette in week one that leaned into a dark-themed, African-futurist aesthetic – moody gradients, deep teal and gold, long-form editorial typography. It was beautiful. It was wrong. The first round of feedback from two angels and one portfolio founder made it clear: NaiBAN's core audience is split between very senior Kenyan operators and international co-investors, and the moody aesthetic read as \"design studio website\" rather than \"serious investor network.\"",[17,335,336,337,340,341,344],{},"We scrapped the direction in week two. The revised system kept the editorial typographic moves but moved the palette to a lighter, more confident neutral base with restrained accents – more ",[60,338,339],{},"The Information"," than ",[60,342,343],{},"Dezeen",". The tradeoff was ego; the gain was trust.",[17,346,347],{},"The lesson carried into every subsequent decision: when a brand has more credibility to protect than to build, restraint compounds better than expression.",[12,349,76],{"id":75},[17,351,352],{},"What exists now at naiban.b3n.in:",[34,354,355,358,361,364,367,370],{},[37,356,357],{},"A redesigned home that states the thesis clearly (\"Supporting founders with advice, connections, and capital\") with named proof – portfolio companies visible above the fold.",[37,359,360],{},"A portfolio index covering 40+ companies across 16 sectors with filterable views and individual company pages.",[37,362,363],{},"A structured pitch application flow that captures the signal the screening committee actually needs.",[37,365,366],{},"A separate investor onboarding path with the differentiation the network needs to recruit peers, not get mistaken for a generic crowdfunding site.",[37,368,369],{},"A resources section positioned for the network's leaders to publish their perspective, with full blog infrastructure.",[37,371,372],{},"A tightly-scoped brand system – logomark, typography, tone – documented so future additions do not drift.",[12,374,94],{"id":93},[17,376,377],{},"Two things are compounding since launch.",[17,379,100,380,383],{},[84,381,382],{},"inbound application quality",". The founders who hit the structured application now arrive with materials ready, stage clear, ask specified. The screening committee's time per application has dropped meaningfully, and the quality ceiling of applications has risen – sharper founders see a sharper surface and show up accordingly.",[17,385,107,386,389],{},[84,387,388],{},"angel recruitment",". The investor onboarding CTA has pulled in peer referrals from outside Kenya who now have a credible surface to point their own networks at. \"Come join NaiBAN\" used to require a 45-minute call to explain what the network did. The new site closes that pitch in two minutes before the call even happens.",[17,391,392],{},"The work is not glamorous. It is a brand and a site. But brand and site are the lowest-effort compounds in the whole investor playbook – every dollar spent on either keeps paying rent for years.",{"title":116,"searchDepth":117,"depth":117,"links":394},[395,396,397,398,399],{"id":14,"depth":120,"text":15},{"id":28,"depth":120,"text":29},{"id":54,"depth":120,"text":55},{"id":75,"depth":120,"text":76},{"id":93,"depth":120,"text":94},"NaiBAN – Nairobi Business Angels Network","\u002Fbrands\u002Fnaiban.png","sprint","https:\u002F\u002Fnaiban.b3n.in",{},[406,409,411,414],{"label":407,"value":408},"Active angels in network","100+",{"label":410,"value":257},"Portfolio companies",{"label":412,"value":413},"Sectors covered","16",{"label":415,"value":416},"Timeline from brief to preview","4 weeks","\u002Fwork\u002Fnaiban","Brand and product lead",{"title":278,"description":116},[421,422,423,424],"Next.js + Tailwind","Portfolio CMS","Founder application workflow","Investor onboarding","work\u002Fnaiban","Full brand identity and digital platform overhaul for the 100+ angel investor network backing 40+ African startups – from WordPress default to a conversion-optimised platform with applications, portfolio management, and investor onboarding.",2025,"Yra64UtPJqp_djtvATEDPEbS52_wxYzclIlX23FA8nU",1783007198533]