Dara Sobaloju, a Nigerian design engineer, built an AI tool called Pewbeam that listens to sermons and automatically displays Bible verses in real time across screens in the church. It started as a tweet. It reportedly has 25 paying churches across three continents. The real question isn’t whether Pewbeam works. It’s whether the church is ready to pay for it.

Last year, Dara posted on Twitter: “I want to build a Bible presentation AI agent for church use,” he wrote. “Imagine Bible verses coming up on screen as the pastor preaches just based on what he’s talking about or his paraphrases and quotes. I want to build this completely in public, starting today.”

The replies filled up before the afternoon was over. Pastors. Developers. Churchgoers. People who had sat in the congregation watching a projection team scramble to keep up with a preacher mid-flow and thought (maybe) that someone should really fix this.

Eight months later, Pewbeam claims to have 5,000+ downloads, paying church customers across 30 countries, and is in the middle of closing a fundraising round. It is, by any reasonable definition, an emerging company. The question worth asking is whether this is a real business, as Africa’s tech ecosystem increasingly turns its attention to religion, one of the continent’s most lucrative and least digitised industries.

I attend 6 am morning mass somewhere on Lagos Island.

The priest says, “And the Bible tells us, Jeremiah 29:11, that God has plans for us, plans to prosper and not to harm us.” Or he doesn’t say the reference at all. He just says, “For I know the plans I have for you.” Either way, Pewbeam, running on a laptop connected to the church’s projection system, has been listening since the mass started.

It processes the priest’s speech in real time using an optimised version of OpenAI’s Whisper voice recognition model, detects the scripture reference, and within two seconds, the full verse appears on the screen in the church. No one had to type or pre-upload verses. The AV team, who previously spent Sunday mornings in a low-grade panic trying to keep up with a priest who might freestyle verses, is now free to manage the verses on the screens live.

Pewbeam also handles paraphrasing. If the priest says “God tells us he will be with us in every valley,” the AI’s search layer is trained to understand meaning, not just keywords, and surfaces the most likely intended verse. It works offline, without internet — a non-negotiable design choice for a product built for the Nigerian and African market, where reliable internet during a service or mass is not something you can assume.

That last line is the product philosophy in a sentence. Build for the hardest conditions first. If it works in places where the infrastructure is unreliable and the WiFi is inconsistent, it should work everywhere else.

Now, for the question that underlies all of this: why does this problem exist at all?

This is not a Nigerian or African problem specifically. It is a global church problem. Walk into a Catholic mass in Lagos or a megachurch in Houston, in each of them, some version of the same experience: the priest or pastor references a scripture, the congregation either knew it already or they didn’t, and the question of whether they could follow in real time was largely left to chance.

Some churches have invested in projection teams and pre-uploaded slides. Some have printed bulletins. Some have nothing at all beyond a Bible in the pew and the assumption that you brought yours. What almost none of them have — or had, before Pewbeam — is a system that automatically solves the problem in real time, regardless of what the preacher decides to say.

The pre-uploaded slides system, which is the dominant solution in churches that have invested in any kind of scripture display, has a flaw: it only works if the preacher sticks to the plan. And preachers, in my experience, do not always stick to the plan. The Spirit moves. A reference comes to mind. A verse becomes suddenly relevant that wasn’t in the order of service.

The projection team, who uploaded 12 specific slides for today’s message, now has to manually search the Bible app in real time while the preacher is still talking — and the congregation watches a blank screen or a completely wrong verse while the team catches up.

Pewbeam’s solution to this is not more preparation. It is no preparation. Or rather: preparation that happens automatically, in real time, by a system that is always listening. It is the difference between a human and a speech-to-text API. The human is always slightly behind. The API doesn’t need to catch up.

Pewbeam’s business model is straightforward.

Three pricing tiers. The Starter plan is free — AI verse detection, two built-in themes, 40 minutes of transcription per week. Enough to try it. Not enough to run a real service on. The Plus plan is $14 a month: unlimited transcription, all themes, no watermark, and AI sermon notes. The Core plan is $30 a month: three devices per licence, main and alternate output, priority support, and advanced AI features.

For a church spending a lot of money per month on utilities, generator fuel, audio equipment, and staff, $14 to $30 per month is not a significant line item.

The pricing is a bit underpriced. And this is both a strength and a risk. It is a strength because the barrier to adoption is almost negligible — a church administrator looking at the Plus tier is weighing $168 a year against the alternative of continuing to embarrass their projection team every Sunday. It is a risk because $168 a year per church is not a path to a large company without extraordinary scale.

Pewbeam claims it currently has about 25 customers. If it had 2500 paying customers — 100 times its current base — at an average of $20 per month. That is a solid small business. It is not, at current pricing, a venture-scale outcome without significant expansion beyond the core use case.

The question I come back to when I think about Pewbeam’s business model is whether this is a sustainable business or a beloved feature that never quite scales: does the church, as an institution, prioritise congregation experience enough to pay for it consistently?

The answer, historically, has been no. Or rather: it depends entirely on the church.

The global church management software market was worth roughly $970 million in 2024 and is growing at nearly 9% annually. Over 350,000 churches worldwide use some form of digital management software. In the US alone, more than 95,000 churches use digital giving platforms. The market exists.

The willingness to spend exists — in the right churches, at the right price point, for the right problems. The challenge for Pewbeam is that it is addressing a problem the projection team currently solves for free, even if they do it poorly. The migration from having someone do this manually to paying $14 a month to do it automatically requires the church leadership to decide that the manual solution is not good enough.

There are two things that could disrupt Pewbeam, and neither of them is a direct competitor in the traditional sense.

The first is better church organisation. The pre-uploaded slides system sometimes fails because preachers go off-script. But what if churches got better at scripting — at building sermon preparation processes rigorous enough that the projection team always had the right slides ready?

This is not a technological solution. It is a human and institutional one. Some churches are already moving in this direction: tighter service planning, digital templates, shared document systems between clergy and media teams. If this becomes the norm rather than the exception, the problem Pewbeam is solving — the freestyle clergy and wonky system that the projection team can’t keep up with — becomes less prevalent. Pewbeam’s best customers are the churches that preach spontaneously with outdated systems. If spontaneous preaching declines in favour of tighter scripting, the addressable market narrows.

The second disruptor is big tech deciding this is worth building. Faithlife — a technology company that provides digital tools for church management — has a vast installed base in churches across the world. It has the distribution, Bible data, and development resources to build Pewbeam’s core feature within six months and bundle it into its existing subscription.

Microsoft, which already sells cloud services to churches and has an entire vertical for nonprofit and religious organisations, could embed real-time scripture detection into Teams or a standalone product. These are not imminent threats. They are category risks — the kind that every niche product in a large market faces when it starts to get traction. Pewbeam’s best defence against them is speed, depth of product, and the specific expertise in African church infrastructure that a Silicon Valley company building from the outside would take years to develop.

What Pewbeam represents, beyond its own specific product, is worth paying attention to. It is a company built to address a problem observed in Nigerian churches, using a design philosophy of offline-first, low-bandwidth, built for infrastructure scarcity. It has already found customers in Africa, in the United States, and in Europe.

Should Pewbeam exist? Maybe. It solves a real problem, it does so elegantly, and it was built from direct observation of something broken.

Does it have sustainable DNA? That is the more interesting question, and the honest answer is: potentially, if it expands beyond verse detection. The core product is a strong hook. The sermon intelligence layer—transcription, notes, summaries, searchable archives—is where the recurring value and pricing power reside. If Pewbeam becomes the platform for how churches capture, organise, and distribute the content of their services, then yes, it has sustainable DNA. If it stays only in the verse detection lane, it is an excellent feature that will eventually face commoditisation.

The founder knows this. The product already hints at it — the Plus tier’s AI sermon notes export is not an afterthought.

The verdict will be decided by how quickly Pewbeam owns the sermon intelligence category before it becomes obvious to a larger entity.

Or if my thesis is debatable. We’ll see.