Deploying AI solutions used to take time. Now their game-changing benefits are more affordable and easier to adopt.
By Dominic Gallello
Last year, in the article, “The New Business of AI (and How It’s Different From Traditional Software)” Martin Casado and Matt Bornstein of Andreessen Horowitz took an investor/vendor perspective on AI. They are huge believers in the technology and have invested heavily in it. But they are cautious about how AI will transform businesses and how AI-led enterprises will not likely look like traditional software companies. Casado and Bornstein argue that AI companies will be less profitable because AI requires heavy services to build its models. Those services are expensive compared to shrink-wrapped software businesses and SaaS businesses.
Essentially enterprise AI has been the proverbial “long pole in the tent” for things like lean manufacturing, value stream mapping and smart inventory control. It’s the most difficult task, and potentially the most important if it’s done right.
I have run all three kinds of business. For sure, the most profitable and easiest to manage was the shrink-wrapped software business. It is a fixed-cost business. Once you ship enough product to pay for your development costs, every incremental download from the customer is almost all profit. The move to SaaS was a bit tricker. Providing users with 24 hours, seven days a week, flawless operations can be a very thankless business. The new infrastructure and managing the people in the business who oversee it, has certainly caused me more than a few all-nighters to ensure customer delight. This is not easy to achieve and can be very expensive.
Now we enter the world of AI. Using software that takes advantage of AI is often not so straightforward. Generally, vendors have done a good job on tasks like building platforms to aggregate data, process data, run models and report out critical information that gives companies actionable insight. They use these techniques to optimize production processes for electric vehicles, optimize truck routes and make critical plant equipment more reliable amongst many other processes. The elephant under the carpet of AI has been building the model or models that produce real results and create incremental value that is superior to those techniques used heretofore on the frontlines of industry.
Skill and money
Companies often refrain from launching projects that could provide significant benefits because they lack the skills and funds to build AI models fundamental to their goals. If a project does commence, vendors using often scarce data scientist resources build the models, usually as fast as possible using cobbled-together tools and scripts. It is a tedious process that includes data collection, data pre-processing, tag mapping, feature engineering, AI model creation and model training and tuning. Wow, that’s a lot of work, but we’re not done yet!
Setting thresholds for alerts and notifications, workflow integrations and display configurations are also necessary to bring real value. For customers, this process is very expensive. For vendors, it takes up high-skilled resources, delivers low margins or sometimes is even a money-losing business. Nobody wins.
But there is hope. We are entering into version 2.0 of business-to-business AI. Version 1.0 certainly has showed the value of enterprise AI. But with scarce and very specialized skills to make AI-based prediction real on the frontlines of application, this next chapter in AI will be dramatically different in one very important way. AI model-building will move from the specialist data scientist to the business subject matter expert – the SME, a.k.a. the citizen data scientist, as Gartner has described.
To enable this shift, companies aiming for success in the future will need to tool and create user interfaces that transfer the power of generating solutions to SMEs who will drive AI to rapidly build AI models in a no-code manner. A must going forward is the ability to extend, create, set up and deploy predictive analytics in production, build and extend templatized (that is, standardized) models and retrain models quickly, capitalizing both on AI’s processing power and analyses as well as the SME’s deep knowledge and sensibility.
That’s the out-of-the-box advantage. Long tent poles that can go up right away can make for pretty big tents.
Beyond the centralized data team
Today’s version of the “typing pool” are services organizations with deep data sciences expertise to build and deploy AI models for customers. As enterprise AI 2.0 becomes a reality, expect to see these organizations to become redundant as their tools become more widely available in the same way that advances in word processors supplanted and set free the typists of yesteryear. The result will be a win-win for both users and vendors.
Dominic Gallello is CEO of Symphony Industrial AI. Gallello is an established technology executive with a comprehensive background in engineering vertical software solutions at companies, including MSC Software, Autodesk, Macromedia and Intergraph Japan. He has led three successful public and private software companies over the past 16 years, resulting in $1.3 billion in exit value, with an average increase in value of more than 300 percent. He was named a SaaS Top 50 CEO in 2018 and brings a track record of company culture awards and more than fifty product awards, including two R&D 100 awards.