As it turns out, artificial intelligence (AI) is a major focus at the Microsoft Build conference this week. While the part that's getting all of the attention is the ability for Microsoft's Cortana and Amazon's Alexa to communicate, there's a lot more going on. What will be important to enterprise IT and developers alike is that Microsoft Build is showcasing Azure as a viable path to AI, and it's also talking about new capabilities in its machine learning (ML) service, which is available using Azure.
Not unexpectedly, AI is a common thread across all of the big-name developer conferences this spring since most of these companies have made heavy investments in the technology. The Google I/O conference not only showcased new AI capabilities, but it also announced that Google Research is being renamed "Google AI." The company also talked about AI on Android and Google Home, though IT professionals and developers also had news about AI capabilities being incorporated into Google's Cloud Platform. And, of course, Facebook's F8 conference announced an open AI framework from that company, which hopefully also has a strong ethical compass attached to it.
All of this attention about AI will lead to inevitable questions about how your organization can use it and ML to improve operations and overall competitiveness. But, since those questions will likely come from someone with only a high-level concept of what's involved, it will fall to IT and DevOps pros not only to prepare now, but also to understand the capabilities and limitations of these technologies and how they apply to particular business cases.
Become Well-Versed in AI
To avoid being swallowed up by AI's hype, you will need to accomplish two things. First, you'll need to become well-enough versed in general AI knowledge, which means understanding what the tech can and can't do and where it's headed over the next 12 months. Second, you'll need to map that knowledge to your organization and its workflows so you'll have a good idea of who in your organization might benefit from AI. Only after answering those two questions will you have some idea of the resources you're going to need if an AI project actually happens.
Building general knowledge of AI and ML isn't all that difficult. The article links listed earlier will get you started and you can flesh that learning out with even more PCMag reading, including this piece on how ML is impacting security and this one on the capabilities of AI databases. Once you've dipped your toe into the general AI pool, it's time to get platform-specific.
Start with what your current cloud vendors offer. One advantage you have is that some of the major cloud vendors—notably Google, IBM, and Microsoft—are offering AI and ML as cloud services, either standalone or in conjunction with their Infrastructure-as-a-Service (IaaS) offerings. For example, IBM cloud customers can simply select IBM Watson as a menu choice when configuring their IBM cloud services.
Additionally, check out your available learning resources. There are a lot of AI webinars out there but it might pay to take a more substantial course in AI, especially if you can focus on the topics you need. Vendors might help here, too. For instance, Microsoft has developed on online course in AI called the Microsoft Professional Program for Artificial Intelligence that will help you master the skills to do more than just talk about AI with your friends. The Microsoft course appears to be comprehensive, it's offered online, and, if you don't need a certificate, then it's free.
And once you've got a general grounding, an effective way to gain more pointed, specific knowledge on how AI can help your organization is to simply reach out to your cloud vendor. Sure, you'll need to deal with a professional services salesperson but that's the only downside. The upside is that these folks are one-stop-shops when it comes to quickly mapping out how their advanced services can help your organization. And what they don't know, they can easily find out with a direct line to engineering. You can avoid a long sales pitch by coming armed with the right questions. All you need to do is some due diligence prep, both on what they offer as well as what your organization needs. This can steer a conversation, the upshot of which you can repurpose as an early blueprint for how your company can implement AI and what you'll get out of that investment.
And remember: these services don't necessarily need to be native to the provider. Rackspace, for example, can provide access to most of the AI services offered by the other big-name cloud vendors and even provide them as part of its managed service.
Know Your In-House Functions
As mentioned, a key part of IT's job in the AI equation is understanding how the business works and how AI and ML map to those needs. Obviously, your customer service operation and your call center are two areas in which there's been significant growth in conversational AI, but that's only good if your company currently employs chatbot tech. If it doesn't, then you'll need to add that cost to your overall AI implementation plan. And, as this week's developer conferences are clearly showing, there's plenty of AI growth in other business areas, especially in analytics, development, and security. If your company currently employs development or DevOps consultants, then don't be shy. Sit down with them to discuss where they see AI and ML, and consider inviting them to your talks with your cloud provider's professional services rep, too.
If you haven't mapped out all of your business processes, then AI is as good a reason as any to start. Even if it turns out that AI isn't a fit for a particular process, understanding what's going on is never a bad thing, and you'll undoubtedly need the data later for different deployments. Mapping out your business processes is also a fairly straightforward task if you just follow these three key steps:
- Identify the process. Generally, this is a meeting-intensive step but you can ease that burden by starting off informally. Go top-down from senior managers to mid-level ones to identify what the business' mission is and how its processes support that. Meetings can be informal, over-coffee discussions, the results of which can power more formal process-mapping sessions.
- Put together a team for each key process. Don't try to go this alone as that's almost certain to fail. Instead, winnow down your process list to a manageable number of core processes and then put together a small team of experts in each. That's your brain trust.
- Map the workflow. Once you've got the what and the who, then just map out the how. Step by step using a standard flowcharting tool, map out what happens, who makes it happen, and what they're using to get that particular job done. Use data from your software and hardware audits to verify your findings. You can go as deep or as light as you want at this stage, but a good gauge is when you start having "aha" moments concerning how AI might help a particular process. You'll probably also have several more such moments concerning unnecessary software and hardware expenditures.
Follow these three steps and you'll quickly determine the most effective areas to consider for AI in your shop. Even for those areas in which you determine AI is not a fit, this is invaluable data to have for the future. And for those in which you think there really could be some benefit, it's time to look at resources. Fortunately, many AI service providers make resource consumption and evaluation easy. A good example is IBM Watson, in which you can find a wide selection of prebuilt AI solutions for all kinds of operations, ranging from customer service to visual recognition. IBM even offers an AI-powered call center solution that can help customer engagement while cutting costs. Evaluating these solutions will give you real-world experience of how AI can work in your organization, fuel your discussions with other business managers, and keep your vendor discussions more focused on what you need rather than on what they want to sell.
Signing up for evaluations is easy. In IBM Watson's case, you'll need to go through an IBM Cloud account, and you'll need to be able to engage some of IBM's services. Or you can use Azure to reach Microsoft's AI-powered language processor, which the company calls "Microsoft Cognitive Services Language Understanding" or LUIS. This service is designed to help with speech recognition services, but Azure also carries a growing number of related offerings aimed at different tasks and verticals.
Figure Out the Costs
Function and capabilities certainly comprise key knowledge points, but an important and unavoidable data point for which you'll undoubtedly become responsible eventually is cost. The major cloud providers that provide access to their AI products can help you figure this out but, as I learned while working on my IaaS review roundup update for PCMag, it's not easy. Therefore, you should start now, and you'll almost certainly need to ask for help. The professional services people with whom you're engaging to help define your needs can also help you figure the costs, though you'll need to temper this with your own knowledge of how your organization works, the timeline constraints it's under, and its staffing resources—all of which can have a deep impact on long-term costs.
Sure, kicking off an evaluation project like this is difficult to prioritize if no one in the management chain has yet made any noise about AI. But like I said, the business process knowledge is invaluable whether or not your company eventually deploys AI. Also, gaining that knowledge now will make you a rock star the day someone with high-level knowledge brings it up in a staff meeting. Besides, once you dig into AI and ML, you will see that, clearly, an AI-powered future is on the horizon for most organizations no matter the industry.
In the end, it will be the IT and DevOps staffers who will most likely need to implement and manage the product or service, provide the infrastructure required, and manage the security and integration processes necessary for the development team to make it all work. Prepare now and you'll save yourself many headaches later.