Automation is the flavour of the season, as businesses grapple with skills shortage and high HR costs. As the benefits of machine learning pan out, the C-suite is investing in automating more and more complex processes.

The rise in automation lifts Artificial Intelligence (AI). 52% of CIOs now automate complex tasks, and 49% of them apply machine learning in such complex automation. Another 40% of CIOs will apply machine learning soon. But AI is a costly endeavour, and talent is the biggest cost-centre. 41% of CIOs lack the skilled staff to manage intelligent machines. But only 27% of them have hired new people with intelligent machine skills, to drive their machine learning projects.

Here are the ways business leaders put machine learning to work, without hiring a costly team of experts.

1. Deploy Ready-Made Solutions

Custom development is no longer the only mode of implanting Artificial Intelligence. Several vendors now bake machine learning into their products. A case in point is the self-learning wireless network administration software.

Off-the-shelf programs allow businesses to install machine-learning solutions without deep expertise. IT staff may run AI models without mastering its mathematical or scientific complexities. For instance, Dell EMC Ready Solutions for AI simplifies the workflow and unlocks insights for many enterprises. The pre-designed and pre-validated solutions make artificial intelligence simpler.

The cloud offers a host of easy-to-deploy and scalable AI-based solutions. IT teams with a basic level of tech skills can use these platforms to get deep insights into their customers and business, in double-quick time.

Consider the case of risk analysis on check cashing service, or finding out whether a check is likely to clear. Tech teams may tap into the Google Cloud Platform and the open-source TensorFlow machine-learning software library, to create a custom application. The app could apply social media sentiment analysis, traditional credit bureau data, and other techniques to determine the creditworthiness of the employer.

2. Channel the efforts of the employees

Continuous learning is the lifeblood of any tech employee. With technology always changing, a technocrat who refuses to learn soon becomes obsolete. The onus is on the CIO to push or motivate the tech team to learn machine-learning concepts. The CIO will have to convince employees about the benefits of career advancement and more on offer.

Since technocrats already have the basic analytics skills imbibed, mastering the new niche will not be too steep a task. A basic competency in tech, combined with some practical experience, suffices to learn machine learning on-the-job. Employees already having statistical and big-data skills can master the details of machine learning technologies quickly. A good understanding of matrix algebra and statistics helps.

Formal training accounts for 20% of the machine learning skills. 80% of the skills are in the culture and the growth mindset.

40% of CIOs have rewritten IT job descriptions to focus on work with intelligent machines. In many enterprises, employees with no experience in machine learning, but having experience as a software engineer and a reputation of being a fast learner are in the frontlines of developing and implementing AI-based solutions. In the not too distant past, developing AI solutions required a PhD and several years of experience.

A firm understanding of the business is as important as good analytical skills. Consider Lucidyne’s example. In 2016, the company explored software that could learn and improve on its own. The company got the job done in six months without using data scientists. Business managers gave inputs to programmers to develop workable practical software.

Microsoft offers another good benchmark. The company encourages employees to take ownership of their own learning experience. Employees expand their knowledge base by immersing in different areas of business and learning from other teams. Interacting with different functional teams offers different perspectives and enhances outlook.

3. Institutionalize Teamwork

Relying on machine-learning scientists alone to push machine-learning projects is a big mistake. A successful AI implementation team requires someone with an ear to the ground, to focus on the practical details. A data scientist may make perfect theories in the lab. But even the best models fall flat on its face in the wake of reality. For example, an AI model assumes the easy availability of data on the customer’s buying habits. But getting accurate data may be impossible, and if possible too costly to be viable.

Another key task is facilitating the data. 51% of CIOs regard data quality as the top barrier to adoption. But only 45% of CIOs have developed methods for monitoring mistakes made by machines.

The team needs to take care of data pipelines, data cleansing, and the performance characteristics of an application.

With such key members in the team, a machine-learning specialist may become surplus to requirements.

If internal talent is lacking, or not trainable, take recourse to freelancers. Cultivate an ecosystem of partners with competence in AI and machine learning. Hiring independent contractors is a reliable and scalable way to experiment without investing in permanent staff.

Successful enterprises cultivate a powerful ecosystem, complete with strategic alliances with key providers. Investing scarce resources for in-house development of complex solutions goes against agility.

4. Consider Alternatives

Artificial Intelligence is a wide corpus of knowledge. The manner of applying it varies even within companies in the same industry vertical. Different companies’ use machine learning based on their specific business goals.

Consider alternatives. Focus on Artificial Intelligence tasks for which competencies exist within the enterprise. Tap into open-source forums and for insights on how a particular AI-based model works.

Cognitive automation is a ripe area. Cognitive systems automate specific tasks to augment human activity. It performs a narrow task within a much broader job or works not done by humans in the first place, such as big-data analytics.

Identify areas of the business ripe for cognitive automation. Next, identify the right technology for automation. Consider alternatives. Investing in the wrong areas or applying the wrong technology wastes time and money. Rule-based systems and robotic process automation offer transparency but do not learn and improve. Deep learning is capable of self-learning, but it is impossible to understand the underlying logic of the models created. Such a blind spot raises several issues in regulated industries such as financial and healthcare. Here, regulators insist on knowing the rationale of the decisions.

At times, there is no workaround to experimentation or a pilot project. Short, rapid sprints of experimentation yield the best results. Through such experiments, enterprises get a grasp on the skills requiring more focus or infusion of talent.