It’s a rainy Saturday morning in Lagos, and the noise inside the public hospital walls on the city’s coastal edge drowns out the pelting rain. In one of its echoey wards, voices overlap, and weary nurses try to manage paper trails that pile up faster than they can process them. For decades, this scene has been typical: overworked staff hunched over files, struggling to juggle patient care with burdensome documentation.
But something is changing. In one consulting room, a young doctor in a lab coat finishes diagnosing a patient and instead of scribbling into a paper chart, he speaks. “Fifty-year-old male, presenting with fever, cough, and fatigue. Suspected tuberculosis. Start treatment protocol.” Within seconds, his notes appear, transcribed verbatim on his screen, complete with punctuation.
This shift is powered by Intron, a startup using voice-based artificial intelligence (AI) to help healthcare professionals enter medical records using speech.
The startup, which launched in 2020, was founded by Tobi Olatunji, a medical doctor with a decade of experience. In many African countries, doctors like Olatunji attend to hundreds of patients daily and deal with a lot of paperwork. Intron Health cuts the time doctors spend writing a patient’s diagnosis. Doctors can enter patients’ medical records, and generate patient reports by voice commands. Crucially, Intron’s AI understands Nigerian accents.
Intron is part of a wave of African startups building artificial intelligence tools that speak to the continent’s realities. Beyond local accents, its voice-to-text system understands medical terminology, allowing overworked doctors in Nigeria and Kenya to dictate patient records hands-free, in real-time. Across sectors, homegrown AI solutions like this are bridging long-standing gaps: helping healthcare workers clear documentation backlogs, enabling customer support in indigenous languages, transcribing local court proceedings, and making radio broadcasts more inclusive. As global models struggle with the underrepresentation of African data, homegrown ventures like Intron are building tools trained on local voices to close service gaps in medicine, education, agriculture, and media.
While AI applications on the continent have been affected by inadequate data, Intron’s speech-to-text AI transcription tool accounts for many African accents. Olatunji says the datasets are trained on over 3.5 million audio clips across African languages, making accommodations for 288 accents. The company currently cares for more than 56,000 patients across over 30 public and private hospitals in Nigeria and Kenya, including the University College Hospital, Ibadan, Aminu Kano Teaching Hospital (AKTH), Kano, Babcock Teaching Hospital, Ogun, and Meridian Health Group, Nairobi.
Intron, which has expanded its AI models to power courtrooms and call centres across Africa, claims it has helped reduce the turnaround time for radiology reporting at the University College Hospital, Ibadan, by 99.3% from 48 hours to 20 minutes.
Source: Intron Health
Intron isn’t alone. Enter Spitch AI
Beyond Intron, a growing crop of startups across Africa is building foundational voice AI, tuned to the continent’s linguistic realities. One of them is Spitch AI, a developer-first audio AI platform.
When Temi Babs realised that OpenAI’s Whisper “wasn’t well-tuned to African voices,” the former engineering student abandoned his note-taking app and set out to build the missing layer himself. “African languages were not being catered to by these large models, so we pivoted to voice AI for Africa,” he says, recalling the October 2024 launch of SpitchAI.
The startup had a clear mission: make AI speak the languages of Africans, in both directions: speech-to-text and text-to-speech. Rather than build end-user apps, the Lagos startup sits at the bottom of the stack, offering simple APIs and SDKs that let any team plug local-language voice capabilities into call centres, media tools, or learning platforms, without the need for Machine Learning expertise.
Starting with Yoruba, the team has since added Hausa, Igbo, Nigerian-accented English and, after demand from East Africa, Amharic. Those rails now power everything from multilingual customer-support hotlines to Nollywood studios that generate synthetic dialogue instead of hiring voice actors. “Producers choose the voices they want, now they can produce their movies in a shorter time and spend less money,” said Babs.
Developers buy pay-as-you-go credits to use the speech-to-text and text-to-speech APIs, while large corporations pay bespoke fees for tailored models. “We price on a case-by-case basis: developers go to our portal to buy credits and make API calls, enterprises want a custom solution,” Babs explains. 
The Nigerian government is in the race too
While private companies like Intron and Spitch AI are building artificial intelligence tools that speak to the continent’s realities, the Nigerian government has quietly entered the race to build a foundational AI model.
Last year, the Federal Ministry of Communications, Innovation, and Digital Economy commissioned Awarri, a Lagos-based startup, to develop what will be the country’s first government-backed large language model (LLM). The model, now nearing completion, is a direct attempt to help increase the representation of Nigerian languages in the artificial intelligence systems being built around the world, according to Nigeria’s tech minister, Bosun Tijani.
Co-founded by entrepreneurs Silas Adekunle and Eniola Edun in 2019, Awarri is building the country’s first LLM model trained on five low-resource Nigerian languages and accented English, in partnership with Data.org.
In November 2023, the Lagos-headquartered startup launched a data annotation lab poised to be an AI talent development hub. The lab employs over 100 workers, who are responsible for gathering and annotating data, creating language models, and developing AI apps.
Awarri also launched LangEasy in April 2024, a platform that enables anyone with a smartphone to contribute to training the model through voice and text inputs. LangEasy gives users sentences to read out loud, and asks them to save the audio on the app. The app will help crowdsource data for Awarri’s LLM, according to the startup founder.
Building Nigerian AI is anything but easy
Building Nigerian AI is anything but easy. Training data remains scarce, especially for indigenous Nigerian languages with low digital footprints. Infrastructure barriers continue to slow progress, and the effectiveness of these AI tools depends on how well they capture Nigeria’s full linguistic and cultural diversity.
“Building in AI requires two major things: computing and data. That pipeline does not exist in Nigeria at scale,” said Babs. To get around it, SpitchAI has stitched together an internal data-collection and labelling pipeline and leans on partners for proprietary corpora. “We have our own data infrastructure plus some really good data partners that help us get the high-quality sets we need,” Babs adds.
For Intron, the process involves managing over 3,000 active contributors, collecting hours of annotated voice data, paying workers across borders, and verifying every utterance for training-grade quality.
“People think it’s just about models, but the real moat is operations. You’re managing massive linguistic diversity—Igbo, Hausa, Wolof—and paying thousands of contributors. That’s not just a tech problem. It’s logistics, quality control, finance, and trust at scale,” Olatunji notes.
On the computing side, costs are brutal. Training a single model takes weeks of GPU time and cloud orchestration. Despite credits from AWS, GCP, and Azure, those budgets dry up fast. “We’ve burned through $100K in AWS credits already. And that’s just one training cycle. You need bigger GPUs, faster clusters, every experiment costs real money,” said Olatunji.
But infrastructure is only half the battle.
Unlike many AI startups that wrap open-source models, Intron builds from scratch. That requires real research talent, and in Africa, that’s hard to find, harder to train, and nearly impossible to retain.
“Zuckerberg is hiring ML researchers at $200M. How do you compete with that? We mentor our team internally, but even that isn’t scalable anymore,” said Olatunji.
The team now runs an open-source research group of 30–40 PhDs globally, publishes papers, and uses that network to hire strategically, avoiding inflated U.S. salaries while still accessing top talent.
Even after training, Intron spent over a year figuring out how to serve models to just 100 users reliably. “People think calling an API is scaling. It’s not. You’re dealing with model orchestration, concurrency, GPU latency, cost optimization. At AWS, we had 30 engineers just managing that layer.” Intron had to compress that into a 30-person company.
The opportunity: a voice-first Africa, defined by locals
Despite the headwinds, the path is clear. Intron and Spitch are two of the earliest and most serious contenders in the race to define what voice AI should look like for African languages. While global players continue to prioritise scale over specificity, both startups are digging into the hard, messy, and valuable work of building AI that actually understands the continent.
And they’re already seeing traction.
Intron is landing enterprise deals across healthcare, legal services, and government, with clients like Ogun State Judiciary replacing legacy voice systems and hospitals adopting its transcription stack for clinical recordkeeping. The company is also powering voice authentication for fintechs and working on voice-enabled banking and KYC flows, allowing users to conduct transactions via natural speech.
“We want a future where someone says, ‘Send ₦5,000 to my brother,’ and it just works in Yoruba, Swahili, or African French,” says Olatunji. “We’re not just doing AI. We’re designing usability for where the infrastructure actually is.”
Spitch, meanwhile, is positioning itself as developer infrastructure for voice, offering a no-code conversational agent builder (currently in private beta) and supporting early users in healthcare, legal tech, customer support, and entertainment.
“We want to sit at the bottom of the stack,” says Babs. “Let other industries innovate on top of us, whether it’s a bank in Lagos or a media startup in Nairobi. That’s how we scale.”
This platform-first approach gives Spitch more surface area than vertical-focused startups and aligns with investor expectations in a nascent market. “We’re not chasing quick revenue,” Babs adds. “The real upside is in reach. Savvy investors know this is a long-term game.”
If both data engines scale as promised, Intron and Spitch could define the default voice layer for the region long before Big Tech learns to roll its r’s in Yoruba.
This report was produced with support from the Centre for Journalism Innovation and Development (CJID) and Luminate.
Mark your calendars! Moonshot by is back in Lagos on October 15–16! Join Africa’s top founders, creatives & tech leaders for 2 days of keynotes, mixers & future-forward ideas. Early bird tickets now 20% off—don’t snooze! moonshot..com
