MVP Case Study: Learnings from AI-BO’s Minimum Viable Product for Startup Validation

Businesses need to make data-driven decisions about their products and customers. This is where market research comes into play.

At The Delta, we know firsthand how tedious and difficult it is to find people to participate in research, so we built a venture that will change that. AI-BO: Market research reinvented for the African market.

But AI-BO didn’t start out as a market research tool. This is the story of AI-BO’s pivot from costly bot to progressive MVP.

MVP — but not the Michael Jordan kind

In the startup industry, MVP stands for minimum viable product. It’s the leanest (or quickest, easiest and cheapest) possible version of a product which can be used by early customers as a later stage validation experiment.

You’ll be able to gauge whether there’s a demand for your product and whether your idea is viable in the real world.

The term was coined by Eric Ries in his 2009 book, The Lean Startup. According to the book, the purpose of an MVP is to get the maximum amount of validated learnings about customers with the least amount of effort.

One of the pitfalls of using an MVP to validate a business model is that some teams put too much emphasis on the ‘minimum’ aspect and end up with a less-than-viable product. This results in inaccurate data. A successful MVP needs to be as lean as possible, without compromising on viability.

In AI-BO’s case, a lot of money was spent early on, for a bot that went unused. We quickly pivoted to create an extremely lean MVP to take AI-BO forward in a new direction.

AI-BO V1

AI-BO started as a data labeling and acquisition business to build vernacular language models. Our mechanism for collecting and labeling data was WhatsApp as it allows for wide and seamless access to people.

The AI-BO team developed the capability internally to build WhatsApp bots for handling data. This was an expensive process that proved unfruitful. It was never actually used and we didn’t get a single client with it.

In hindsight, we should have worked out how to do the same operation in a very unscalable manner (e.g. message 500 people and manually collect data like that OR use a no-code botbuilder). Then, sell the capability. Once we’re selling beyond what we can manually handle then it’s time to BUILD!

In April 2021, after months of trying to land customers from varying industries with no success, we decided to pivot to utilising AI-BO as a market research tool.

AI-BO V2

Founders who have failed previously have a higher chance of succeeding in their current venture. So it makes sense that the learnings from V1’s failure have set AI-BO on the right track for V2.

Pivoting away from data labelling and acquisition, AI-BO V2 does market research for the African market. The African market is characterised by audiences who are difficult to reach, particularly those in low-income areas.

WhatsApp is the most popular mobile communications app in South Africa which enables AI-BO to have a turnaround time of just 3 days, with a variety of respondents available. Survey participants are incentivised to fill out surveys as a stream of income in the form of airtime — a hot commodity in South Africa.

The real value here is that small businesses can hear the voice of the customer directly, which allows them to create products/services that solve lower-LSM customer problems. Traditionally, research into the mass market is done by small-scale in-person interview campaigns which are slow, costly, and aren’t always representative enough to glean true insights.

For this renewed approach, AI-BO skipped expensive custom bot-building and went for a no-code botbuilder. The no-code builder saves time and money because no developers are required and it gives the team increased speed and agility. We can make changes immediately and experiment frequently to test if different methods work better.

A lot of research went into choosing a botbuilder. It was overwhelming — there are 100s of different botbuilding software. This is the decision-making framework we used to make the decision a bit easier:

• Price: needs to be affordable to prevent further loss
• Can handle conditional logic: route people to different flows based on the answers they give
• Database setup: the ability to remember data
• Ease of use: easy to make changes and build surveys regularly

In the end, Landbot’s price was significantly lower than the competitors that match its feature set.

Bearing in mind that each survey has to be set up individually in the bot builder which takes a lot of time and effort, this isn’t sustainable in the long term. AI-BO is currently operating at a service loss — the salary value of the time spent on surveys exceeds the incoming revenue.

Next Up for AI-BO

After a short time in operation, AI-BO recently achieved a personal best. We got 300 vetted responses for a payments survey in the space of 5 hours. That’s a payout of between R3 000 and 4 500 in airtime to respondents and the chance to clean, visualize and make sense of high-quality data.

For context, you typically need at least 100 and preferably 200 responses for a representative sample size of the population you want to know about. This result of 300 vetted responses proves that AI-BO is capable of achieving its value proposition and that our idea is worth investing more effort into.

Right now, AI-BO is still very much in the MVP phase and its stack is completely outsourced to other great product teams to run/manage. In return, AI-BO gives them a SaaS license fee and gets to play and experiment at no extra cost. But, as mentioned, this isn’t scalable.

To scale, AI-BO needs to be able to handle a high volume of clients. We’ll need to build the software infrastructure to support this — mainly to automate processes. It will need to cover:

• Automated survey creation in WhatsApp
• Automated data checks and cleaning
• Automated data dashboard and visualisation

When 90% of the activities that the team spends time doing become automated, revenue will well exceed costs and AI-BO will be left with high margins and a thriving business.

Wisdom & Learnings from AI-BO

The main lesson we can glean from AI-BOs journey from bot to breakthrough is that you need to validate your business ideas thoroughly and cheaply before you put exorbitant amounts of money into building them. In fact, there are several validation experiments you can use to test your ideas before you build a single thing.

With success at our fingertips, we’re excited to see what AI-BO’s MVP is going to morph into next.

Our advice for founders and future founders is to stop thinking “how can I scale what I am doing now?”. You can always find a way to automate processes and get your unit economics to make sense. Rather, focus intensely on everything your customer does and needs by doing everything with them. Observe them closely and ask loads of questions.

If you’d like to check out AI-BO’s website, you can find it here. Or, if you’d like to work with the venture builders who made AI-BO happen, get in touch with us here.