Corporate Failing Chatbots

Why?

When ChatGPT first burst onto the Conversational AI scene back in 2022, it felt revolutionary – and it probably was. That event seems to have precipitated all sorts of other actions including corporate entities rushing to create their own chatbots based on the Open AI LLM model. This gets you about 25% of the way there, in my opinion.

I would like to offer our step-by-step process to get you to 95% instead of 25%.

Steps:

Step 1: Narrowly define the problem you are trying to solve

Step 2: Detail how that problem is being solved today (through workarounds, manual work, etc.)

Step 3: Map out the business process of how the workflow happens today including each step, decision point, and who is responsible for what

Step 4: Imagine a perfect world (but possible) workflow. What does that look like? Document it.

Step 5: Look at each step in the process. Can it be automated? Would you want to automate it? And if both are true, what type of automation makes sense?

Step 6: Evaluate where AI can be avoided. Yes, you read that right – where do you HAVE to use AI? And where can you avoid using AI at all? This seems anathema at the current point in history, but there are problems when people slap AI on to existing processes or applications. Why? Because AI is costly and mostly non-deterministic. It’s costly not just is dollars, but also in processing power, bandwidth, GPUs, etc. Additionally, when you use AI, you can’t be 100% sure of what’s going to happen or how it will respond. When precision is important, AI should only be used where necessary.

Step 7: Calculate how much time or money can be saved by implementing your solution. (Calculate the ROI.) Make sure to consider repetition in your calculation: How many times would each task need to be repeated? How much time would each task typically take? Calculating the time saved per unit of time X cost of doing it manually can allow you to prove ROI before you even begin the project and obtain buy in, if needed.