Hello, I am here with the next article on AI automation. When we create any automation, we create a workflow, so for the employer for whom we are creating this automation, for the client, for the outside perspective, one important thing that the factor counts is its cost, and calculating its cost is very important because there are many factors involved in it.
This development of automation is just one section of the cost, while there are many other things that are required to create any automation or workflow.
It is very important to factor them in so that it can be communicated transparently to the client, the employer, or the outside perspective, because for them, it is very important to do this cost-benefit analysis, this ROI, and this is mostly the one who is developing or who is It has to be done by the solution provider.
As you have already understood, if you have developed automation, you must have understood that there are multiple factors in it. For example, whenever we create any automation or workflow that has language models, API integrations, and various subscription-based tools, we have to look at their usage. For example, if you are using a language model, what is the cost of the part token?
What will be its cost per month, and what will be the usability of this user?
i.e. how much is its client base, how much will it use, and how much will it cost per month. In the same way, when we talk about APIs, the APIs that are available are also available on a monthly basis or they are also on a call basis. Mostly, they are on a call basis. Now, how many calls will you have in it? You have to estimate the cost based on how much it will be daily and monthly. In addition, if the solution you have is to host on the cloud,
What can be the cost?
If it is to be self-hosted on premises, what can be the cost? You have to factor in all these things, and you have to communicate this cost so that the client or the outsider of your product can decide whether this solution can afford this cost or not. And many other things that, if you look at,
When we look at the future, the cost that AI automation is charging the client or the business every month in most cases, is comparatively very less than the cost that they are already paying for the resources that they are utilizing, but the cost that they are paying on a monthly basis is still a burden. It is very important to communicate this cost,
and in today’s article, we will do its cost and keep in mind the factors. I will tell you step by step what things you have to evaluate in it and how you have to take care of them, and let’s go. There are eight steps for today’s video cases that I have listed out. You can also combine some of the steps that are there, but I have done it because it will clear up a little understanding.
Let’s break them down one by one in a little detail. The first step is to break down the workflow. That is, breaking down means that you have to see its different components separately, which are its main components, because the requirements of people are different every time, and its components and nodes are used differently. So you have to see which major processes are inside it or some services are used, for example, what are we using for input, are we using any form API for input or which reasoning model are we using,
what are we using for storage and what are we using for notifications, so you have to list out these things, which applications, tools or services are you using the most. The next step in this automation is that you have to see which are the paid components in it, that is, whatever services or tools you are using, the components are just those components that are there in them that are paid for. For example, if you are using a language model, you are using LMUs, then if you take your own token, for example, then
The tokens that are counted, if you look at the example, can be charged from one cent to three cents. Similarly, the Google API is requested. The number of calls, how many calls are being made, can be requested, for example, at five dollars, or it can be less or more. You can confirm these things. If it is subscription-based, if it is monthly, what is the subscription. Similarly, if you host it, if you run it on Netrun’s own cloud, which is official, then they can have any fixed fee, which is for 5 runs. If you have to do executions for 5 months, then what will be the charges for it? If you have to go beyond that, then what will be the charges for it? These things have to be identified by you and hosting; if you have to do self-hosting, that is, you have to do self-hosting on Google Cloud or AWS. By renting the instance of PHU to the model,
What will be its cost per month?
Similarly, the next step is to estimate what its usage volume will be. This means how much print your language model will have to receive or process. Now you can do a small calculation of it, such as if you have a chat summarizer, if you do a little bit, how many tokens will it run and how many tokens will it use in the input and output, and how many tokens will it read per month? So you have to do a little estimate of how many runs, how many times will that chat summarizer or chat, if you use a chat trigger for example, how many times will your chat brown take input, and how many tokens will be used on it, and per month. What will be the cost, and if you want to do a total cost, you can do it on total tokens and multiply by the token price, which will be the price of the token, and the token, which is the total token divided by its cost, and multiply by the token price,
which is its price. After that, if you look at the included NA 10 execution cost, that is, what will be the execution cost of Any 10, that is, here we are only talking about the hosting of An Aston, if you do hosting, then it also has its own cost, the server is intended, the database is separate and the backup and monitor are separate. In this, if you are using Anytime, the pricing is based on the number of workers executing. If you use the official cloud of NA 10, then it is based on the number of executions. But if you post it yourself, you will have to estimate some things like I mentioned that the server that you have, if you take EC2, then it has a configuration, EC2 which is AWS,
then which machine you use in it, it can cost $10 per month, $20, it can be $30, similarly, in the database you are using, are you using RDSU or are you using it locally, it can vary, the price backup storage you are using,
monitoring and logging if you want, it has a cost and as an example, if you look at hosting, you can read it at $10 to $30 per month, self-hosted and similarly the storage that is there can cost $2 per month, if you want to run a workflow, you can run unlimited, there is no limitation in this, self-hosting Similarly,
if you look at step four, in it you have to see the cost of the integration APIs because the APIs that you integrate are very important in your automation because they are charging calls on a sim time basis. You can see the method as a little formula that the number of API calls per month is how much,
and multiply it by the cost of the call that how much a call costs you, and multiply it by the number of APIs that you have to call in a month. That will be your monthly cost. Well, now, in this, which will be your main revenue, or you understand what you have to charge, will come now. If you look at step six, in this step, there will be your development cost for the workflow that you have to create, what you have to charge, and
What do people do at this time?
That you see how many hours it will take me to make this work kilo and give it a number that I have, so much my rate, and it will take so many hours, and it also has a maintenance cost. Sometimes the employer or client or someone outside tells you that you have to maintain it too, so for that, you have to decide on a separate cost along with that.
