AI Business Requirements

by Chris PehuraC-SUITE DATA — 2024/05/22

Email us at contactus@csuitedata.com to see how we make your plans work.


AI talks must have everyone represented. We finally got everyone in the same room. All the right faces are at the table, on the Zoom call. People from business and sales, reps from tech and data, customer service, and legal. We even have some AI workers in here. Everyone’s here and jacked up. We’re all excited to talk about AI and what AI can do for our company. But that’s a big conversation. Every time there’s a big conversation, especially something big like AI, people want to boil the ocean. They want to talk about everything. Not only is there no time to cover everything it can also result in a handful of people taking over the conversation. You don’t want to boil the ocean. You want to keep energy levels high for adopting AI. You need to be conscious that the more conversations you have about it, the more roundabout talks you have about it, the less enthusiasm you’ll have, and the more resistance you’ll face when you adopt AI. Conversations need to be direct and conversational, and they can’t go around in circles. You want everyone to stay excited. To do that you need to scope down the conversations and direct them.

AI talks must be an exciting journey. Talking about AI, especially its requirements, is going on a journey with AI. You make sure that the journey doesn’t have a meandering path. To avoid going down the rabbit hole, you need to do the prep, the leg work to fast-track over every rock, pothole, and sinkhole in the AI conversation. This isn’t just to save time. You don’t want people to be frustrated during the journey. You want them jacked up so they’re going to take what they’ve learned and bring it to their areas and implement AI in the right way. To scope down the conversations with AI you got to make sure AI is not nebulous. People must have a good idea of what AI is and what it isn’t. Nebulous concepts result in very nebulous and long-winded conversations where people can find themselves all on the wrong page. We don’t want that. We want everyone to be on the same page for what AI is and what AI will be doing for us.

AI talks must have definitions. AI is a family of tools that allow you to classify something or control something. Classifiers help do identification and assessments. Controllers help do redirects and rerouting. Classifiers are thinkers. They’re the thinking AI. Controllers are the doing AI. When people talk about AI, they’re really simultaneously talking about a mixture of classifiers and controllers. They don’t talk about how the thinkers and doers are working together. To cover all the AI requirements, you have to separate when it’s a classifier versus when it’s a controller. That’s the key in getting AI to do exactly what you want for the organization. Classifiers do identification like the identification of a person’s face or generating a financial statement or logo image based on a series of prompts. Classifiers do assessments like an assessment of a fraudulent claim or a patient’s health condition. Assessing patients like their risk level when they take a foot into the hospital. Controllers do routing of spam email, routing parts on the assembly line, or redirecting documents through work streams to increase productivity or to monitor suspicions of fraud.

AI talks must be from the business perspective. When everyone’s in the room and talking about how AI is going to be used at our company, this needs to be structured as a high-level business requirements conversation. The fastest way to get through a business requirements conversation is to be focusing on certain words when talking about how AI works. There is certain business language and certain terms you need to emphasize. You can’t just give everyone a vocabulary list. You have to use those terms when you’re speaking, when you’re facilitating, and when you’re interpreting what other people are saying about what they want AI to do. This works extremely well when you have a workflow. It works better if you have a surrogate workflow that is representative of all the workflows at the company. Everyone should be able to see their area within that surrogate workflow. You walk through that workflow explaining the process of AI and how AI can hook into each of the activities and each of the transitions between activities. This isn’t a passive conversation where you’re telling how AI fits here and there. This is a conversation where you are asking questions. Where do you think AI will fit? How do you think it’ll fit?

AI talks must be engaging. Never repeat the exact same questions in succession. It’s like a jogger with a sharp pebble in their shoe. It gets painful. You need to give people prompts, keywords, and phrasing questions a certain way so you invite people to contribute to see how AI fits. Not only will you get a lot more understanding in the room you’ll also fast-track the conversation because people will be able to see where AI is, where it’s going in the company, be able to communicate it, and articulate it very clearly. The best words I found to use are from business language. Business language can fall short so you use Bloom’s Taxonomy. Words like how would the AI help “define” something, how would the AI help “identify”, “hypothesize”, “plan”, “organize”, “compare”, “choose”, “combine”, “apply”, “deliver”, “assess”, “defend”, or “recommend” something. By phrasing everything in a question and having everyone draw out and contribute to what is seen in the surrogate workflow, everyone will have a good idea of what they need to bring back to their own areas and what to talk about when they do. Do not underestimate the power of these terms. They’re crucial for the adoption of new paradigms and technologies because when the terms are consistently used consistent language is formed. Consistent mental maps are formed. Consistent navigations through how we all think is formed. Work is delivered faster. Solutions are developed quickly. Using a strong consistency in our language gives us an added benefit. It is not far from reason to realize that AI could actually look at our business requirements, look at the consistency of that language, and then generate the solutions for you.

AI talks must have legal representation. There’s a reason why we brought legal into this AI journey. What happens when AI goes wrong? When it goes bad? When it does some stuff on our behalf that causes damage or major harm? It’s bad to lose revenue from our AI. It’s worse to have damaged property, being investigated, and costs from litigation. That’s much much worse. We have to make sure we have some mechanism in play to be able to investigate what the AI is doing. We need to be able to monitor the AI from a risk management viewpoint. We have to ask some tough questions about accountability. Who is really accountable for the AI for the work the AI produces? The vendors clearly don’t want to be accountable. So who will be accountable? Another key issue is that AI is constantly changing. It’s constantly learning. It’s very hard to investigate and reproduce the results of what actually happened when AI is a moving target. This means you need to keep a major audit trail to track what’s going on and have the tools to be able to dig into that audit trail. This is extremely difficult to do.

AI talks must talk about ethics and privacy. When doing an audit trail for AI you’re not just auditing the AI you’re also auditing the people that are using the AI and potentially the customers outside of the company too. This is a serious breach of ethics and privacy. Worse, if you decide to collect all of this information in an audit it’s not unreasonable that you could be hacked.

AI talks must happen. Despite the costs of the AI to our reputation and quality of work, AI should still considered and adopted. Every disruptive technology has downsides and are always full of surprises. It will take time to find out what all of those are for AI. And I’m pretty sure with the way AI is going we’ll easily be able to get solutions to help address all of our concerns with AI. As long as the politicians, ministries, and other government bodies don’t complicate things with ill-advised regulations.


Email us at contactus@csuitedata.com to see how we make your plans work.