The world of artificial intelligence is frightening. No, not the danger of an army of AI-powered robots taking over the world (though that is a bit concerning). The real fear is that the wrong vendor is chosen or the rollout handled poorly. After all, AI is complex, not fully mature, in some cases poorly understood, and […]
The world of artificial intelligence is frightening. No, not the danger of an army of AI-powered robots taking over the world (though that is a bit concerning). The real fear is that the wrong vendor is chosen or the rollout handled poorly. After all, AI is complex, not fully mature, in some cases poorly understood, and involves great changes to how an organization thinks and operates.
Much of the complexity stems from the fact that AI has no single meaning or definition. It is a combination of several elements (machine learning, natural language processing, computer vision and others). This means that use cases tend to be unique and complex. Companies not big enough to hire expertise rely deeply on consultants and vendors, likely more than in more familiar areas. AI is not for the corporate faint of heart.
So how should organizations approach AI?
The first step in any corporate initiative is to fully understand what is on the table. It seems almost needless to say that organizations must educate themselves about AI before taking the plunge. But, in this case, it’s so important that it is worth stating the obvious. They must assess what data they have to feed into the system and if remedial work is necessary to enable that data to be used.
Tractica Research Director Aditya Kaul suggests that organizations understand the difference between the AI platforms that process raw data to reach conclusions and perception-driven approaches that focus on the intricacies and nuances of language and vision. The next step is to experiment on a wide variety of use cases and settle on those that bring the greatest value to the organization. It is important to understand the metrics that will be used to gauge success, such as increased productivity or reduced costs.
Moving Ahead with AI
At that point, they are set to move ahead aggressively. “Once companies have a good understanding of the AI technologies and use cases, they can go [choose] a third-party enterprise-grade AI platform and build a robust framework around data and model warehousing that allows for efficient production-grade AI that can be swiftly deployed into client-facing products and services,” Kaul wrote to IT Business Edge in response to emailed questions.
This suggests deep changes, which makes choosing vendors an even more vital decision than better understood limited technology deployments. The stakes are high. It is a nascent field where some companies no doubt are selling vaporware and some perhaps haven’t figured out their own value proposition. It’s best to be very careful. “If your AI vendor won’t promise you real ROI, it’s because they can’t deliver,” wrote Ben Lamm, the co-founder and CEO of Hypergiant. “If a vendor is trying to skirt around a clear statement of value, you know they won’t serve you well in the long run.” Organizations should do the same block and tackling that is done for any other significant investment. Credentials should be checked, deep conversations conducted and a high comfort level achieved. “One of the most important things enterprises can look for in an AI vendor is understanding the success of their customer base,” wrote Peresh Kharya, the director of Accelerated Computing for NVIDIA. “Don’t be afraid to ask which of their customers are successful and how has their new AI solution benefited their business. Asking this question will help you gauge the tangible business value the vendor is promoting.”
Organizations can take steps to increase the odds that they will choose the right vendor. Dave Damer, the founder and CEO of Testfire Labs, offers three tips. The first two focus on precisely what the vendor will be providing. Companies should ask if the prospective vendor delivers packaged solutions, custom solutions or both, and if it has the necessary expertise in house or must outsource. Finally, the organization must understand what will happen after the deployment is done. “A lack of employee training or further customization of models can lead to unusable and/or ineffective technology,” Damer wrote.
Best of Breed or Single Vendor?
A longstanding debate in telecom and IT circles is whether platforms are better coming from a single vendor or “best in breed” arrangements in which the top elements are cherry picked and strung together. The single vendor platforms presumably are better integrated and have deeper and easier to use management functions, while the best in breed approach potentially offers better performance.
The pendulum is swinging toward multiple vendors, at least according to Tracy Malingo, the senior vice president of Product Strategy at Verint, which bought AI firm Next IT last December. “This is actually one of the biggest shifts that we’ve seen in AI,” Malingo wrote. “As major players have sought to lock in ecosystems and as companies have evolved in their understanding and needs for AI, we’ve seen the market begin to shift toward best of breed over single-source vendors. That trend will continue in the future.”
The bottom line is that AI is a slippery slope: That slope can arc toward more efficient operations and a healthier bottom line – or toward confusion, failed implementations and all the headaches that those results bring on. “Organizations should have a clear understanding of what business issues they’re trying to solve with AI,” wrote Guy Yehiav, the CEO of Profitect. “How will the technology they’re evaluating make an impact to both top and bottom line and what is the approach to roll it out across the business? If analytics and AI are done well, the impact should be quick and results tangible.”
Read the source article at IT Business Edge.