Context is King

Context is a bit of a buzzword in the AI space, but it is central to making tools like MAIA useful. Context is the difference between the general answer to a general question any entrepreneur can get out of ChatGPT, and a targeted next step to solve the entrepreneur's next most pressing problem, along with accompanying assets to make it easy to implement. 

One approach is for tools like Maia is to be integrated with, and even come from, an MSMEs large supplier or customer. So for example a small retailer may buy many of their products from a large distributor. If that distributor creates an AI advisor plugged into all of the shop's historical orders, it could immediately give targeted advice on day 1 with all this rich context. There are many tools being developed along these lines today. But beyond the technical challenges in building them, there are two fundamental problems that keep coming up.

First is the organization's desire to put their fingers on the scales and direct responses towards their own ends. This can start innocently enough- let's make sure the tool has our entire price list and features loaded so it can give accurate information, but we won't spend time doing that for other distributors, that's not our job. And it progresses from there. It makes sense: why would a distributor spend money building a tool that suggests' its competitors products and services over their own? Or that encourages small suppliers to only sell to them instead of diversifing their distribution. In the end that just means the tool is not a real coach, it is not operating in the best interests of the entrepreneur.

Second, the entrepreneurs themselves behave strategically. If that retail shop is having some financial problems, would they really seek help from their AI coach about it? Couldn't the supplier use that early information to proactively limit trade credit?

So the rich context comes at the cost of some very messy unaligned incentives which undermine how useful the tool is for the entrepreneur, and if/how they use it.

The other approach is to build that context over time, which is how Maia works. This takes longer and isn't as elegant as the other short cuts, but has the benefits of being completely aligned with our users. We've found that business profiles become quite useful for giving grounded advice for a user after only a few weeks, and we spend a lot of time finding high-value low-effort first uses of the tool, to give the business owners a quick win while building context (eg making logos, a user favorite).