From Silos to Services: Cloud Computing for the Enterprise

Mar 31 2018   8:57PM GMT

Adding Artificial Intelligence into Software Platforms

Brian Gracely Brian Gracely Profile: Brian Gracely

Tags:
ai
Artificial intelligence
Machine learning
ml
Zugata

Image Source: FreeImages

It’s difficult to go a day in the tech industry without hearing a prediction (e.g. here, here, here, here, and many, many more) about Artificial Intelligence or Machine Learning. Jobs for these skills are in high demand and companies in nearly every industry are trying to figure out how to embed these capabilities into their products and platforms before their competition.

The question for many CIOs or Product Managers is, “How do we add AI into our platforms?”. Do they hire a few PhDs or Data Scientists? Do they try using one of the AI/ML services from public cloud providers like Google, AWS or Azure? Or is there some other path to success?

Recently I asked this question to Srinivas “SK” Krishnamurti (@skrishna09; Founder/CEO of Zugata) as his company recently announced a new AI service to augment their existing SaaS platform. I wanted to understand the complexity of the technology, the difficultly in finding engineering talent, and how to make the product more attractive with the embedded AI capabilities.

The first thing we discussed was what types of business (or customer) problems that AI/ML could potentially solve. SK highlighted that it was important to understand who the would be using the software and how much experience or expertise could be assumed about their use-cases. Once this was a well understood domain, then it was possible to understand if or how AI/ML should be used.

The next thing we discussed was how AI/ML advanced would be perceived in those use-cases. Would they create a measurable difference from what could be manually accomplished now, and would the difference be perceived as valuable enough to make the investment?

Once we got past the business value and use-cases, we began to focus on how to find the right staff to start the AI/ML process. SK shared with me that their journey lasted well over a year before they began to feel confident that their efforts would be valuable. This included hiring talent, looking at data models, building the models, and the long process of training the models with data. He said that the longest amount of time was in training the models, as they had to frequently ask themselves if they were biasing the system to get the answers they believed were needed vs. the system coming to those answers by itself.

Finally, we talked about the challenges of building AI/ML models that were influencing non-human systems (e.g. electronic trading or IT Observability systems) vs. systems that would directly impact human decisions (e.g. hiring, firing, evaluating emotional state, etc.). He said that this added yet another layer of complexity into their analysis of their models, as again they needed to make sure that a broad set of scenarios were being properly evaluated by the system.

It was clear to me that there is no single way to add AI/ML into a software platform, but the guidelines and guidance from SK may prove to be valuable to the readers as they begin to explore their journey to improve their software platforms. I’d love to hear of any experience that the readers have about their own systems.

 Comment on this Post

 
There was an error processing your information. Please try again later.
Thanks. We'll let you know when a new response is added.
Send me notifications when other members comment.

Forgot Password

No problem! Submit your e-mail address below. We'll send you an e-mail containing your password.

Your password has been sent to:

Share this item with your network: