Custom Model Creation Process

Creating a custom model for your business is simple with Content Villain and we want you to have the best experience possible whilst we help you out with this so sit back, relax and enjoy our overview of the process involved when creating a custom model.

Getting the Concept

Every business has some repetitive processes that they can automate. We see it time and time again and often the simplest automations have the most impact. Getting clarity on what you want your custom model to do will help us to understand exactly what you need and help us to build it out in a quick time. We often like to do this over a phone call. It will help us to understand you and your business. Our expertise can shoot down any unrealistic ideas and use our business consulting knowledge to help you formulate your idea.

Once we have the concept then it is really about determining the technology to use and what path to go forward. Do you need this with an API? Is this going to be against any of the NLP providers TOS? We'll figure all of that out as we are in the business of solutions rather than computer says no.

Examples are Invaluable

The next step is to get a few examples together. We tend to find that 3-5 is the sweet spot. This is really us saying to you, "Give us your ideal AI output. Give us some examples of what you would be super happy if the AI could produce." This process works in two ways and helps us out a lot.

The first way it helps is by actually showing us what you would be happy with. We can take those examples and build on them to get a dataset together to ensure that the AI can output similar responses each time it is run. It is better for us to get your ideal AI outputs than guessing and putting our own ideas forward as ultimately this is one of the main selling points of a custom model.

The second way it helps us with building your custom model is determining what we would need to provide as the input fields. AI is there to help you. It is meant to speed up your content creation and give you quality but it is not going to be able to produce anything of quality without the right input fields. Our team reverse engineers your ideal AI outputs and determines the minimum but essential input fields we feel are required to get to that output. We'll let you know what we think is needed and you can give us feedback on this.

Building out the Model

Now the fun begins! Once we have determined the inputs, outputs and the general concept of the custom model you want to be built, our team can get to work and start building it out. We'll give you a timeframe of how long this usually takes once we complete the previous conversations on input fields. The more complexity, the longer it will take and we're determined to give you something that works for your business so do not rush these.

Usually we'd have everything we need at this point and get a beta version together. Our team will then test this internally and see what we think of it. There is a chance at this point, we may ask for a few more examples or additional information if we see a quick fix to improve the model substantially otherwise we'll sign off on our initial testing and give you a version of the model that you can do your own testing with.

Beta Testing Concept

At this point, we'll send you an email to let you know that we have put a beta version of your model live onto your account. You then can test this yourself and enjoy getting the outputs from it. We tend to say whilst in beta format, do not share the outputs with more than 4 other people just whilst we determine that the quality and everything is working to a high level of quality.

After 1 week if we did not hear from you with any feedback, we'll reach out to get your thoughts. Any suggestions for tweaks / changes will be considered and actioned. The next step is the full delivery of the model where you are free to use it as you wish.

Custom Model Delivery

With the final delivery, we can provide you with API access if that was agreed during the consultation and all guidance on sharing outputs etc is lifted. You are able to then use the model like you would with any of the other models on Content Villain.

Some NLP providers have a stricter going live process than others. If your model is created using such technology, we will let you know as there can be additional waiting time due to this approval process.

This brings an end to the custom model process and we often get customers asking then for another one! In terms of ongoing maintenance on your model, we will ensure that it is kept running and up to date. Any changes in terms of training data or purpose would need to be discussed on an individual basis based on the complexity.

Last updated