We have quickly moved past the hype phase and, in just a few years, transitioned to a stage of concrete AI usage. Now, we have a clear understanding of how AI is transforming the lives of individuals and humanity as a whole.
Let’s focus on the practical side—how AI is used today, especially in startups. Initially, and still to a large extent, AI adoption revolves around interacting with commercial models developed by other companies. The basic setup looks like this: your product sends user requests to these models and receives responses.
Some startups integrate their own or open-source models, adapting them to their needs, but the core principle remains the same.
A couple of years ago, startups often kept the models they used a secret, considering it a competitive advantage. They either didn’t disclose specific models or concealed the combinations they used, offering users a complete solution without technical details.
Now, I believe the trend is shifting toward transparency. Successful startups no longer hide which models they use. Take, for example, the latest hyped-up product—Cursor (a developer tool). They openly share details about the models they use and don’t see it as a threat to their business.
Their magic and competitive edge lie in how these models interact, how input data is structured, and how results are interpreted. They have built a business around seamlessly integrating models into a code editor. Users can work with the results, make modifications, and update the code with a single click.
I think this is a fair and effective way. At this stage, there is no real point in hiding information about the models used. The competition should either focus on developing proprietary models (which is extremely complex and costly) or on creatively interpreting and applying the results of existing models. After all, results can always be tailored to best fit your product and users.
For example, what Sendforsign is doing?
We have an API that interacts with our servers. We configure models to not just provide results but to take into account the specifics of our product.
You might ask the model to check a document for risks. Instead of simply listing the risks, it will suggest contract modifications to minimize them—and integrate these changes directly into the relevant parts of the document, considering its specifics. All that remains for you is to review and accept or reject the suggested changes.