This article is the second in a two-part series on how CSPs can implement AI and automation. You can read the first article in this series here.
Implementing change to any process can be quite the feat, regardless of industry. The stakes are only raised when seeking to implement artificial intelligence (AI), as Communication Service Providers (CSPs) look to automate processes they can expect to face common challenges in three specific areas: Strategy, Data, People.
It may sound like common sense, but to be successful in introducing AI the CSP needs to have a clearly defined strategy, accompanied by clear business objectives and the support of senior management.
AI is so broad that it can theoretically be applied everywhere, which can a problem as many AI projects fail due to a lack of thoughtful strategy. Thorough planning and preparation are essential, there are a variety of questions telcos need to examine when initiating an AI implementation:
- What are the key areas and processes that require automation before any others?
- What areas will produce the quickest Return on Investment?
- How will success be measured?
- How will we communicate the initiative to staff to ensure strong employee engagement and support?
Aggregating data effectively can take a great deal of effort to implement AI successfully. Telcos in particular utilize a large variety of data sources – OSS counters, device measurements, drive testing, probes, CDRs from billing systems, customer care records, trouble tickets plus external data like weather (impacts equipment reliability) and social media sentiment (an indicator of problems). At the same time, 56 percent of CSPs have issues with data quality.
Interpreting and managing large amounts of data is critical, as typically only a small fraction of all this data is used. The goal is to identify spikes, drops or instances where something crosses a particular threshold, to train AI on what to look for so the appropriate remediation response can be taken.
Telcos with mature data hubs that can ingest multiple sources of information, subset and share will be in a better position to implement AI into processes, compared to those without data hubs.
Having the appropriate skills and expertise of individuals responsible for implementing AI is extremely important, yet 55 percent of CSPs say their organization lacks the skills needed to implement AI. A combination of data science and telco expertise is what is needed to implement AI successfully.
However, data scientists are difficult to find and have very high salaries – CSPs are competing with companies like Google, Apple and Amazon for those high demand resources. One alternative is to retrain your best telco engineers to obtain additional attributes to help ensure AI is implemented successfully.
At Nokia, we trained 7,000 engineers on the basics of big data, analytics. They are not data scientists but are comfortable handling data and using or customizing existing solutions as needed. We call them ‘Citizen Data Scientists’.Andrew Burrell, Head of Marketing, Ultra Broadband and Analytics at Nokia
Also, people tend to be naturally reluctant to change and resistant to anything that threatens the status quo. Implementing AI does require a great deal of trial and error, so having team members comfortable with an acceptance of risk is important.
At Nokia, we implemented an AI solution to help troubleshoot network issues, the AI was trained and more than 90 percent accurate. However, initially, most senior engineers were reluctant to follow the AI recommendations – even though the AI was almost always right and often could help solve the issue much faster if the advice was followed.Andrew Burrell, Head of Marketing, Ultra Broadband and Analytics at Nokia
When implementing AI, it is important to develop a culture of communication and transparency amongst staff, people should understand that their ability to action AI recommendations in a timely manner can play a key role in the project being a success or failure.