Vendors are racing to coin their own terms or carve out a specific niche when it comes to AI. To be able to select the correct AI service for your business, you need to understand what these services mean or entail.
Chris Sorel, Catalyte’s AI advisory practice lead, discusses the overarching question of, “What AI do you need?” with CTO Jun Wang and Principal Architect Jason Foster. They explore:
- The landscape of AI services/technologies (eg: productivity, concierge, platform enhancements, ML/predictive analytics).
- The use cases for these AI services/technologies.
- How organizations can determine which service/technology is right for them.
You can watch the full conversation or read an excerpted transcript below.
The following transcript has been edited for length and clarity.
Chris Sorel: I’m Chris Sorel, AI advisory practice lead at Catalyte. I’m joined by CTO Jun Wang and Principal Architect Jason Foster. Today, we’re going to answer the question, “What AI do you need?”
There are many different types of AI. Let’s boil through the hype and get to a better understanding of what AIs exist, how they can help my business and what they’ll cost. Can we break this down into major categories of AI?
Jun Wang: Overall, there are modular or systemic solutions. The primary one people focus on is a chatbot or concierge service. This is a modular replacement for an existing service.
Jason Foster: It’s easier to say, “Take that technology and map it onto my business.”
CS: I try to match solutions to business cases. If you talk about concierge, you’re talking about using AI to provide services to end users for engagement with organizations. It’s all about end-point engagement. More specific use cases might require a more modular approach.
JW: Switching it up a little bit, AI, in general, is enhancing productivity; engineering or software development productivity in particular. If you’re writing code but not using AI, you’re missing out.
JF: You still have to have people to understand context. To have those [AI] productivity tools makes you much more productive and more mission critical.
CS: So we have another category we can call “productivity” [AI]. It’s about using AI to tackle tedious or repetitive problems, leaving people to focus on high-value tasks, innovation and the things AI can’t do.
JW: The problem AI is solving is the information overload problem. We free people from the tedious tasks to focus on decision-making. But, when in decision-making mode, you need access to all the data to make those decisions. We have so much data, so we need information retrieval and summary. Large language models (LLMs) will be very helpful in finding this information.
CS: If we wanted to flip the switch on CoPilot 365 or Gemini, is that enough? Does that solve all my AI problems?
JW: Of course not. If you don’t do anything you’re going to lose out to your competitors. But, it’s not as simple as just throwing AI at the problem. People swing between the extremes of throwing AI at the problem or waiting to see what other people do. The right solution is in the middle. It’s hard, but it’ll be worth it. We need to think of the whole process systematically. And you often need someone who can help guide you through that process.
CS: Most organizations will have use cases for multiple categories of AI, all of which can be tied to business cases and ROI. Those use cases can add up in terms of cost. How can we rationalize those costs?
JF: Businesses look at the money out and the money in. If all they perceive is that, “I’m already paying people, why am I paying more?” then [AI] can be seen as relatively expensive. Those businesses are shortsighted. They don’t understand the revenue they’re losing because they can’t understand where they’re having customer churn. There’s no way they can. There’s so much data. You need to understand that AI can help with revenue, help keep customers and help people be more productive.
CS: If you can use AI for customer service triage, you can route customers to the right person more cost effectively.
JW: The value of using an AI customer service chatbot is clearly understood. That concept is generalizable to other categories. AI, in general, is like cloud computing…it’s elastic. When you have a spike, it can scale. In almost any AI category, you will have more productive gains if you implement it modularized. But, you will have a more effective organization if you design a system for the new generation of technology. I don’t consider AI to be expensive. I consider it to be perceived as a “high price tag.” Beyond that myopic point of view, you can generate more productivity and build a more efficient organization with AI. Implement it now and figure out the secondary technical details later. It’s critical to be in the game.
CS: People said it would be too expensive to migrate to the cloud. But we don’t have those conversations today because everyone uses the cloud. Similarly, we have to shift our assumptions that AI is a part of our systems.
JF: It’s tough to fight that inertia. Businesses just get in a rut.
CS: When we’re looking at AI in general, we’re talking about machine learning. Why would I build my own machine learning model versus leveraging existing AI systems?
JF: It’s your data. It’s your business. They’re your products. You can go and buy data sets, but if you want to make it about your customers, you have to have the right guidance about how to implement it. You can’t look for a panacea.
CS: There will always be special cases where there’s value in building your own model. There will be general systems that solve broader problems.
To wrap it up:
- As a business owner, I have to adopt AI or I will be out-competed.
- I should tie selection of AI to business problems in a way that’s measurable.
- When I can adopt AI systemically, I will get more benefit than implementing simple point solutions.
- When figuring out what AI to adopt, I need to figure out the right size to solve the problem and tie that to business value.
