Ethical AI supply chain
The annotation labor behind large language models has been on the front page. Enterprise AI buyers now face real pressure on where their training data comes from — from CSRD reporting obligations, from procurement due diligence, from the press. This is the answer.
Why this matters to you
The EU's Corporate Sustainability Reporting Directive is now in force. AI companies selling into Europe must document the labor conditions, welfare practices, and human rights posture of their data supply chain — and stand behind those documents in an audit. US enterprise procurement is catching up fast, with ESG due diligence questionnaires now standard for AI vendors in regulated industries.
At the same time, journalistic scrutiny of annotation labor is intensifying. The cheapest annotation vendor is no longer the safest vendor. Buyers are being asked, in writing, where their training data comes from — and the answer needs to hold up.
Impact Outsourcing was built from day one to be the kind of vendor that survives this scrutiny — not the kind that scrambles to respond to it.
Supply-chain worker welfare is now part of mandatory sustainability reporting for companies selling into Europe. Your annotation vendor is in scope.
ESG due-diligence questionnaires are now standard in AI vendor onboarding. Buyers are asking, in writing, about employment practices and duty of care.
Coverage of annotation and moderation labor conditions is intensifying. A public-facing data supply chain story is now a reputational asset — or a liability.
AI vendors are reviewed by the same committees that review apparel and electronics supply chains. The bar is rising, and it isn't coming back down.
Operating principles
Six defensible commitments that sit underneath every engagement. Written to be audited. Not a manifesto — a spec.
Every operator on the floor is a full-time employee on a formal contract — not a gig worker, a freelancer, or a subcontracted labeler. The employment relationship is direct, traceable, and documentable to your auditors. The difference is fundamental.
Our pay scales are set above the local market rate for skilled knowledge work in Kenya — not against the data-annotation floor. Wages, benefits, and advancement are reviewed annually against benchmarks, not trimmed to the margin.
Over 60% of our workforce is women. This is a strategic hiring decision, not a quota. Kenya has one of the most educated female talent pools in Sub-Saharan Africa — we built the operation to access it, and the org is shaped by the people we hired.
Operators advance into reviewer, team lead, and delivery roles through defined tracks. The senior layer of the company was hired into entry-level positions. This is how the org sustains itself — and how your account keeps its institutional memory.
For content moderation and red-teaming engagements, operators are briefed in advance, rotated off sensitive queues, and supported through structured wellness programs. We've seen what happens when this is treated as a checkbox. We've built the opposite.
Our employment practices, security posture, and operational discipline are designed to be inspected. ISO 27001 is the floor, not the ceiling. We welcome supplier audits, site visits, and enterprise due-diligence questionnaires — and we answer them in full.
The talent region
Sub-Saharan Africa is a serious technology region, not an exception. A corridor of tier-1 hubs with real infrastructure, real universities, and a time-zone advantage for European and Middle Eastern AI teams.
Nairobi, Lagos, Cairo, Kigali, Cape Town. The continent's tier-1 technology hubs share a pattern — strong English-medium universities, fiber connectivity, modern commercial buildings, and stable regulatory environments for foreign enterprise clients. This is the infrastructure the African AI ecosystem is being built on.
Kenya's talent pipeline is deep and highly qualified — a generation of graduates from English-medium universities across the region, trained in engineering, data, linguistics, and the humanities. The local tech ecosystem has spent two decades building the engineering culture that the region's AI operations now run on.
For European and Middle Eastern clients, the time-zone overlap is a practical advantage — a full business day in the same working window, without the handoff friction of a 10-hour offset.
We are based in Nairobi because the talent density, the infrastructure, and the regulatory clarity are there — not because the labor is cheap. Kenya's educated female workforce is a strategic hiring asset we built the org around; Nairobi's fiber and power redundancy is what lets us hold an ISO 27001 ISMS on real infrastructure.
For clients, this means something specific: you are not outsourcing to “Africa” as a cost play. You are accessing a particular talent region, at a particular tier-1 hub, through a particular operator pool. That distinction holds up in a supplier audit. The generic one does not.
Inside the operation
The operating model is editorial, not gig. Named operators, calibration rooms, written rubrics, senior escalation paths. A place people come to work — and progress.
Shifts begin with a calibration huddle on the day's schema edge cases. Operators sit at their own workstations. Reviewers sit one seniority tier above.
Delivery leads are named and reachable — the same person on week 1 and week 52. Continuity is an operating decision, not an accident.
Reviews are discussed, not dictated. Rubrics are written, iterated, and versioned. Disagreement is a signal — an edge case the schema hasn't yet handled, not a problem to be smoothed over.
Escalation paths to a senior reviewer and then to the delivery lead are documented before production starts, not invented under pressure.
Operators have colleagues, not batch IDs. Teams have continuity, not turnover. People progress internally — the senior layer of the company was hired into entry-level positions.
That is the opposite of how a platform-based operation runs. It is also what lets an account keep its institutional memory across years, not weeks.
Alignment
The frameworks your procurement team already references. Mapped, not reinvented — so the answers are ready before the questionnaire arrives.
For procurement
Procurement-ready documentation, available on request under NDA. Written for the supplier pack, not the marketing site.
Request a supplier pack →Track record
A track record is not a slogan. These are the numbers we put in front of auditors, clients, and journalists — and stand behind.
Continuous delivery for AI teams across North America, Europe, and Israel since 2021.
Full-time employees on formal contracts, based in our Nairobi office.
Every surveillance audit since certification. Cert No. 452AGI102121.
Enterprise client complaints escalated on ethical or labor grounds. Ever.
Let's talk
Schedule a conversation with our team. We'll walk through our operating model, our certifications, and how we'd fit into your ESG and procurement processes. No cold quotes. No supplier-pack vapour.