Ground Truth: Practical AI for communicators working where the playbook doesn't exist
Plus what I'm learning this week and AI intelligence for emerging markets newsrooms
Hi everyone.
I’m working out what half marathon races I should aim for this year. That means I’m working out a training plan needs to look like. Which means I need new gear of course. Which means I need to upgrade my running watch. Please share any running gear tips that you think is worth buying if there are any runners here.
My ask: We work hard to bring you tools, news and cutting edge thinking in the communications space. A paid subscription would be a great new years start for me. $15 a month. Thank you ! (We accept payments in Bitcoin and Solana)
Here’s a good video to watch for this year which is a bit more fleshed out with some of the thoughts and trends I shared with you last week.
I’ll be working with Claude Opus 4.5 this week and I will let you know how I do.
We are looking for partners for The Rundown Studio as we develop more Gen AI intelligence tools for you to use in your comms work. This is about how to apply AI tools to your workflows. Get in touch if you would like to learn more or use our products in your own environments.
I will read through some of the work below I have been doing that is a good refresher as I checked out a bit over the holidays. Here are the best ones for you to review too.
Most Watched Embedded Podcast
An AI Co-Pilot for Reporting on Africa
Africa - 2.0. An update of Binyavanga Wainaina’s original ‘How To Write About Africa’, but for the AI age. After 15 years at CNN and building an African storytelling startup, I got tired of watching smart, hardworking journalists default to the same narrative shortcuts about Africa. Not because of malice but because they’re hardwired into a western system, and stereotype defaults are also just easy. So I built something: a master prompt that forces specificity, centers African agency, and makes it harder to write lazy stories about “Africa”
You can download the prompt for FREE here: AI How to Cover Africa: Best Practice for Journalists Pro.You can read the full substack here.
AI Newsroom CoPilot
How do you deliver high quality journalism with lean teams and zero budget? That’s the question newsrooms across emerging markets have been asking. In this Substack, I walk you through our Newsroom CoPilot and take you through a five-step verification process adapted from BBC/CNN editorial protocols all with a human-in-the-loop at every stage. It identifies key elements, and has extensive source verification. It even gives confidence scores for every claim, and flags what’s unverified or conflicting. If it’s medium confidence, you don’t proceed.
After you’re done working with our copilot, you’ll have broadcast-ready scripts under 200 words exactly the way you’d see on a CNN teleprompter, suggested guests based on who can actually make news and move the story forward, and get detailed interview briefings with tough questions and follow-ups. The potential to democratize elite newsroom standards for limited budgets is massive.
Check out the video to learn more about our Newsroom Copilot here.
Harvard AI Frontiers: The Emerging Economies Edition
What are the top 3 barriers to AI in Africa? Rebecca Enonchong’s answer: “Government, government, and government.” I had a wonderful few days in Boston at Harvard Center for International Development’s annual Global Empowerment Meeting. The theme was catalyzing AI for inclusive change and in this Substack I covered the debate, which brilliantly dove into how AI is playing out across emerging economies. I share some interesting examples of how AI is being implemented, for example, Saudi Arabia’s virtual court system reduced case resolution from 217 days to 30 days, saving citizens 160 million trips annually and 23 billion Saudi riyals. Chile established a National Center for Digital Services developing a localized AI model. Brazil announced $4 billion in government investment for generative AI.
The most pragmatic perspective came from Jalal Charaf: “We cannot afford to recreate what all those rich countries did. We’re asking the question, what do we as Africans want? Do we want to solve problems, or do we want to recreate technology that we don’t have the means for?” With AI implemented correctly, students could possibly learn the equivalent of two years of material in just six weeks.
The practical approaches from this conference (mobile-first infrastructure, solving actual problems rather than recreating Western technology, community-centered AI, legal system solutions for cross-border challenges) are the blueprint for how emerging markets can win in AI without Silicon Valley budgets. More here.
How Africa’s Languages Are Reshaping AI Translation One Community at a Time
Only 0.02% of internet content exists in African languages. That means AI models trained on internet data literally cannot understand how most Africans communicate. Lesan AI built a machine translation system that outperforms Google Translate, Facebook’s MMS, and Microsoft Translator for Amharic and Tigrinya through ground-up community engagement.
I had the real pleasure of speaking to Asmelash Hadgu of Lesan AI for the first Embedded episode of the year. You can listen here: Lost in Translation: Building AI for Africa’s Languages and access some of his most interesting insights in this Substack. We talk about the limitations of current AI models in handling African languages or dialects like Amharic (Ethiopia) and Tigrinya (Eritrea) highlight the digital divide in a way that’s impossible to ignore. I’ve been thinking about how you would handle the vast number of languages in Kenya in datasets, capturing oral stories, and handling nuanced references that are lost in translation.
This conversation was about how Lesan is approaching this problem and trying to break translation technology that can serve communities. Lesan’s community-first approach to data collection is the model for how emerging markets can build AI that actually serves their populations.
What 1,000 CEOs Tell the G20 About Africa’s Digital Future
South Africa’s G20 presidency was historic, the first time the G20 came to Africa. But just ahead of that summit is the B20, meaning Business 20, and in this Substack I break down why it matters.
Around 1,000 business representatives work in task forces throughout the year developing policy recommendations. They delivered four concrete recommendations with specific, measurable targets:
Zero people offline by 2030 (down from 2.6 billion).
Trust in AI jumping from 46% to 85%
Every country teaching computer science.
Digital public infrastructure and policies that actually work. A great example given was that of Kenya:
Eliminating the 16% VAT on mobile phones in 2009 caused mobile penetration to jump 7 percentage points in six months, users nearly doubled in four years, and the World Bank estimates every 10% increase in mobile penetration boosts GDP by around 1%. Then Kenya reintroduced the VAT in 2013—handset prices jumped 16% overnight, grey-market imports exploded, progress stalled. More here.
AI Is Not Replacing Journalists, It’s Exposing Who Actually Knows How to Report
The practical ways journalists are actually using AI framework became essential: BBC uses AI to detect deepfakes with 90% accuracy (not perfect). Financial Times discovered AI-generated three-bullet summaries increased reader engagement by 60%. Glacier Media reporters save 30 minutes to an hour per interview with AI transcription. But AI transcription accuracy rates are only between 60% and 86% in real-world settings, not good enough to use without human oversight.
I write about how the practical journalism workflows I share (content verification, data analysis, format transformation, automation, news monitoring, etc) are the ones that actually work. That 81% of employees are using AI but only 13% work at companies that have policies means that newsrooms are deploying AI tools without editorial standards, verification protocols, or transparency requirements.
I give some tips for emerging markets newsrooms with limited budgets on how to adopt AI practically while maintaining human editorial control. Give it a read here.
Thanks for being a part of this growing Substack community. Appreciate it. A paid subscription would be a great new years start for me. $15 a month. Thank you !