
EP 38 | Speed, Open Source, Ecosystems: The Faster Path to AI Value
Loading player...
How do you build AI that serves profit and purpose—without causing unintended harm?
In this episode of The Angle, Atenkosi Ngubevana - Group AI & Automation lead at Vodacom unpacks a values-first approach to AI: design for inclusive growth, embed ethical data stewardship, and measure impact beyond demos. We get into why many corporates stall after flashy PoCs, how quality data and real digital-transformation work unlock scale, and why startups can leapfrog incumbents with speed, open source, and focused ecosystems.
We discuss personalised sales, fraud detection, and what changes with the gen-AI hype cycle. The headline: the tech is not the value—data, processes, and change management are.
Policy and power come up too - South Africa’s AI regulation is lagging; the urgent work is ethical guardrails and data sovereignty (who holds which citizen data, to what ends?). Atenkosi helped set the tone for a national AI strategy - from infrastructure ownership and governance to public-private upskilling partnerships - so we dig into what “sovereign AI” should mean here.
What you’ll learn
-How to frame AI around values, inclusion, and measurable impact, not hype.
-Why data quality + transformation (people/process) decide adoption at scale.
-Where startups beat incumbents: speed, open source, specialist ecosystems.
-The state of AI regulation, ethical AI, and data stewardship in SA.
-National AI strategy pillars and public and private roles in skills.
-Purpose pilots: AI tutor & career guidance with health/education partners.
Subscribe for more conversations on Africa’s digital future—where systems, incentives, and execution matter more than slogans.
Chapters:
00:00 – Cold open & intro
00:11 – Values before tools: building AI that doesn’t cause unintended harm
01:34 – Atenkosi’s role + Vodacom AI Lab: profit and purpose pillars (education, health, water)
03:20 – Why PoCs wow but scale stalls: data quality, transformation, change management
04:04 – From ML to gen-AI: what actually changed (and what didn’t)
08:22 – Startups vs incumbents: renovate the mansion or build the glass house?
10:25 – Speed, open source, outsourcing cybersecurity, and thinking global from day one
14:38 – Using scale without killing innovation: external partners, “David & Goliath” logic
17:12 – M&A realities, Competition Commission guardrails, and consumer protection
21:19 – Are we regulating AI well? Ethical AI, data sovereignty, POPIA limits
24:41 – Incentives over excuses: SARS as a working ML example; why will matters
27:24 – Infusing AI in purpose work: AI tutor and career-guidance pilots with partners
31:07 – Inside SA’s national AI strategy: infrastructure, governance, partnerships, skills
33:23 – Implementing the strategy: corporate lessons for a sovereign AI agenda
34:59 – Public–private complementarity: upskilling, incentives, and execution gaps
36:39 – Close & credits
#TheAnglePodcast #AtenkosiNgubevana #AI #DigitalTransformation #EthicalAI #DataSovereignty #POPIA #OpenSource #AfricaInnovation #sovereignai
Produced and Published by Submedia.co.za
In this episode of The Angle, Atenkosi Ngubevana - Group AI & Automation lead at Vodacom unpacks a values-first approach to AI: design for inclusive growth, embed ethical data stewardship, and measure impact beyond demos. We get into why many corporates stall after flashy PoCs, how quality data and real digital-transformation work unlock scale, and why startups can leapfrog incumbents with speed, open source, and focused ecosystems.
We discuss personalised sales, fraud detection, and what changes with the gen-AI hype cycle. The headline: the tech is not the value—data, processes, and change management are.
Policy and power come up too - South Africa’s AI regulation is lagging; the urgent work is ethical guardrails and data sovereignty (who holds which citizen data, to what ends?). Atenkosi helped set the tone for a national AI strategy - from infrastructure ownership and governance to public-private upskilling partnerships - so we dig into what “sovereign AI” should mean here.
What you’ll learn
-How to frame AI around values, inclusion, and measurable impact, not hype.
-Why data quality + transformation (people/process) decide adoption at scale.
-Where startups beat incumbents: speed, open source, specialist ecosystems.
-The state of AI regulation, ethical AI, and data stewardship in SA.
-National AI strategy pillars and public and private roles in skills.
-Purpose pilots: AI tutor & career guidance with health/education partners.
Subscribe for more conversations on Africa’s digital future—where systems, incentives, and execution matter more than slogans.
Chapters:
00:00 – Cold open & intro
00:11 – Values before tools: building AI that doesn’t cause unintended harm
01:34 – Atenkosi’s role + Vodacom AI Lab: profit and purpose pillars (education, health, water)
03:20 – Why PoCs wow but scale stalls: data quality, transformation, change management
04:04 – From ML to gen-AI: what actually changed (and what didn’t)
08:22 – Startups vs incumbents: renovate the mansion or build the glass house?
10:25 – Speed, open source, outsourcing cybersecurity, and thinking global from day one
14:38 – Using scale without killing innovation: external partners, “David & Goliath” logic
17:12 – M&A realities, Competition Commission guardrails, and consumer protection
21:19 – Are we regulating AI well? Ethical AI, data sovereignty, POPIA limits
24:41 – Incentives over excuses: SARS as a working ML example; why will matters
27:24 – Infusing AI in purpose work: AI tutor and career-guidance pilots with partners
31:07 – Inside SA’s national AI strategy: infrastructure, governance, partnerships, skills
33:23 – Implementing the strategy: corporate lessons for a sovereign AI agenda
34:59 – Public–private complementarity: upskilling, incentives, and execution gaps
36:39 – Close & credits
#TheAnglePodcast #AtenkosiNgubevana #AI #DigitalTransformation #EthicalAI #DataSovereignty #POPIA #OpenSource #AfricaInnovation #sovereignai
Produced and Published by Submedia.co.za





