It’s amazing the sway that two letters can have. Investors’ ears prick up. Rivals fear obliteration. Shareholders smell progress. But for all of the outsized value that a simple mention of AI can bring to the boardroom, there are only a few companies really able to maximise its commercial potential.
This is because applying AI effectively requires a large amount of infrastructure, both cultural and material. For advances in machine learning (ML) and other frontier fields to be truly integrated into a business, there needs to be a considerable quantity of data, a clear process in place and skilled practitioners who can put these to good use. Many businesses, even larger ones, do not have those ingredients to hand. For other, smaller enterprises, there simply isn’t the time or money to create an effective internal AI operation.
As a result, specialist companies, often startups, are offering a leaner, outsourced AI operation as a service. In the process, these specialist companies are expanding the applicability and power of AI in all of our lives. The more that businesses are able to tap into these technologies, the more customers they’ll reach.
“By outsourcing AI to highly specialised tech companies, not only individual companies but also the entire industry can benefit as the AI models can reach some level of generalisation,” says Ramakrishna Nanjundaiah, co-founder of Phantasma Labs, which provides AI-based models for automotive companies and smart factories. “Generalisation is hard to achieve if industries do not share insights on the range of use cases and the intended value expected from the AI,” he explains.
By working across industries, such specialist companies are more likely to make real advances in the field, potentially to the benefit of all. By focusing exclusively on AI tasks, these companies are far better equipped to try out new techniques and approaches, and become true experts at applying breakthrough methods to a variety of real-world cases.
Outsourcing AI expertise
This might strike many CTOs as anathema to their modus operandi. Aren’t they, after all, the harbingers of innovation? Outsourcing an AI project requires a level of self-awareness and humility, tapping into a risk-averse mentality and admitting that such quests may be beyond an enterprise’s core capabilities. Many businesses cannot justify expensive outlays that result in failure. And as a frontier technology, AI projects are far from guaranteed to succeed.
A deciding factor will often be the maturity of the company’s existing data capture and management systems. The unthinking accumulation of data for data’s sake is insufficient to foster a competent AI operation. The appropriate data has to be captured and then processed, tagged, stored and updated regularly in order for it to be useful.
“Before considering AI adoption, it’s important to know where your data lies,” says Chris Royles, field CTO for Europe, the Middle East and Africa at Cloudera. “Here, having a robust data management strategy is key. Since algorithms are essential for machine-learning processes, meaningful results can only be achieved by leveraging high-quality data.”
Data is important but it’s unlikely that an individual business starting an AI operation from scratch will have all of the necessary data points. Fortunately for many, owning proprietary data and pursuing a big data strategy is far less of a mission-critical moat than it might have been a few years ago, when data was heralded as “the new oil”.
“Data becomes redundant over time,” explains Nanjundaiah. “The Covid crisis has shown us that industries cannot rely on past data alone for preparing for future emergent scenarios. There has to be a different way in which we can prepare industries for Black Swan events.”
Techniques such as reinforcement learning are the next natural evolution of AI, he says, “where we can build value-delivering AI models without the need for big data”.
Cutting costs
Another factor motivating CTOs to look beyond an internal operation will be cost. AI specialists such as machine-learning engineers are very well paid. Salaries above £100,000 are expected but, unless combined with an effective or well-informed team around said engineer, their talents will often end up frustrated.
Indeed, hiring a couple of experts internally may be insufficient if there is a lack of documentation, agreed-upon frameworks and tacit knowledge already in place. An outsourced team, however, will be better able to tap into a general suite of best practice, experience and specialties as well as being on top of all the important regulatory and AI safety requirements.
Cloudera’s hybrid data cloud allows companies to tap into AI- and ML-powered data services, regardless of their own data capabilities. Their advice for CTOs looking to adopt a leaner, outsourced approach to AI is to use the opportunity to explore these new technologies at their own pace – and really appreciate the value.
“Artificial intelligence can infuse your business with the information and agility needed to adapt to market changes and deliver innovative solutions,” says Royles.
When outsourcing, he says, it’s wise to play it safe by starting small and implementing one use case at a time, “keeping in mind that the sequence you create may inform the use cases that follow. Then you keep building on.”
A key consideration for any CTO facing the question of whether or not to outsource an AI project is how ready they are internally to cultivate their own team and AI expertise. It’s an expensive and resource-heavy task that requires a real openness to failure and experimentation as well as a clear sight of the necessary goals. At least to begin with, outsourcing the first few projects might be the most efficient way to discover how they want to integrate AI into the existing technology stack.
It’s amazing the sway that two letters can have. Investors’ ears prick up. Rivals fear obliteration. Shareholders smell progress. But for all of the outsized value that a simple mention of AI can bring to the boardroom, there are only a few companies really able to maximise its commercial potential.
This is because applying AI effectively requires a large amount of infrastructure, both cultural and material. For advances in machine learning (ML) and other frontier fields to be truly integrated into a business, there needs to be a considerable quantity of data, a clear process in place and skilled practitioners who can put these to good use. Many businesses, even larger ones, do not have those ingredients to hand. For other, smaller enterprises, there simply isn't the time or money to create an effective internal AI operation.
As a result, specialist companies, often startups, are offering a leaner, outsourced AI operation as a service. In the process, these specialist companies are expanding the applicability and power of AI in all of our lives. The more that businesses are able to tap into these technologies, the more customers they’ll reach.
