Cognitive

On the road to AI adoption, slow and steady wins the race

Share this post:

AI adoption slow steadyAbout 15 years ago, artificial intelligence (AI) was still in its infancy, with few companies actively using it. But since the explosion of cloud computing and new specialized processors for advanced statistical analysis, this 60-year-old computing discipline has gotten a new lease of life.

Leading-edge companies have made AI adoption a key strategic differentiator, using it to gain competitive advantage and transform their offerings.

For thousands of other companies, though, AI is still a mystery. The tools and techniques necessary to take advantage of it seem just out of reach, and the business strategies surrounding it are opaque. They know that they must embrace it somehow if they are to stay ahead of their rivals, but how can they take their first steps in the journey?

Baby steps

In a recent episode of the IBM Cloud podcast, Ian Lynch and Steve Choquette asked Nico Frantzen, a senior technical architect in the AI practice at Perficient, an IT consulting company and IBM Business Partner, for pointers. His advice? Start small.

AI has the potential to transform businesses, but that doesn’t mean boiling the ocean from the start. Taking careful steps and monitoring the results enables companies to build out their strategies thoughtfully, identifying areas of real value along the way. It gives them breathing space to build the proper organizational and technological structures to ensure success.

One of the first steps involves having a cohesive cloud strategy. The cloud is a crucial component in any AI adoption because the computationally intense training algorithms involved in machine-learning models require short bursts of heavy computing. Buying that computational power in-house would not be economical for many companies.

Another early step involves creating a robust data strategy. While companies can source data on an ad hoc basis for small projects, as their efforts unfold, they’ll quickly find a need for more data from across the organization. Getting that information may involve breaking down operational silos and encouraging data sharing between departments and teams.

Part of this strategy also involves putting proper data governance procedures in place. The quantity of data isn’t the only important factor in machine learning projects. Quality is equally important. Companies must ensure proper data governance so that they understand where the data came from, how it was collected and how it’s been processed. This is important to filter out bias in AI algorithms and ensure that systems produce accurate, useful results.

Quick wins

These are big tasks, but they create solid foundations. For that reason, senior management buy-in is an important part of the AI journey. That requires educating the C-suite on the technology’s value and the level of investment it requires. With high-level executives spread thin and addressing business challenges from all sides, how can advocates of AI do this? It comes down to securing quick, demonstrable wins straight out of the gate.

Begin by creating excitement around the possibilities for AI in your organization. This process requires input from both the technology and the business side. Technologists must communicate AI’s capabilities, debunking any myths that the business side may have. Bots won’t run the company or replace human employees. AI is good at small, easily identifiable manual tasks where a large portion of data already exists to help AI algorithms.

The business side can contribute by identifying processes that need improvement. It should be possible to measure the outcomes of these improvements using a meaningful metric, for example, transactions processed per minute or cost per transaction.

Frantzen gives one example of a likely candidate for a pilot AI project: audio or video transcription. Many companies have mountains of potentially useful data tied up in phone call recordings and call centers. AI’s transcription capabilities are well understood and easily accessible through online APIs. Transcribing this data into text unlocks its value for further projects, such as sentiment analysis.

The long game

If you’re in this for the long haul, there are plenty of things to consider early on in the journey. There’s no need to rush headlong into these tasks at a breakneck pace. If you do, you may find you underestimated the investment necessary for coherent AI adoption. This could leave you with a collection of embryonic projects that fail to deliver any real business value or return on investment. Worse still, it may lead to your AI initiative to stall as competitors excel.

Instead, build a long-term strategy focused on incremental improvement.

Technology evolves at lightning speed, but when it comes to AI, slow and steady wins the race. Learn more about Frantzen’s approach to AI adoption in the latest episode of the IBM Cloud Podcast and sharpen your skills at Think 2019.

More Cognitive stories

LogDNA and IBM find synergy in cloud

You know what they say: you can’t fix what you can’t find. That’s what makes log management such a critical element in the DevOps process. Logging provides key information for software developers on the lookout for code errors. While working on their third startup in 2013, Chris Nguyen and Lee Liu realized that traditional log […]

Continue reading

Quantum-safe cryptography: What it means for your data in the cloud

Quantum computing holds the promise of delivering new insights that could lead to medical breakthroughs and scientific discoveries across a number of disciplines. It could also become a double-edged sword, as quantum computing may also create new exposures, such as the ability to quickly solve the difficult math problems that are the basis of some […]

Continue reading

A strategic approach to adopting cloud-native application development

Approximately three out of four non-cloud applications will move to the cloud within the next three years, according to a recent IBM report titled “The enterprise outlook on cloud-native development”. In today’s modern enterprise, optimizing the application cycle is critical: it can help companies keep up with consumer expectations, keep business operations agile, and speed […]

Continue reading