failed-ai-deployments-blog

What Service Leaders Can Learn from Failed AI Deployments

Common Pitfalls & How to Avoid Them

The list of benefits for artificial intelligence in the enterprise is long and includes improved agility, more accurate decision making and reduced customer churn. Unfortunately, those benefits aren’t automatic. Transformation requires sweeping organizational change, better data policies, user buy-in and an end product that’s easy enough for the entire workforce to use. As more analysts detail the state of AI deployments across the enterprise, we are taking a closer look at how these lessons apply to the service industry.

Why are AI Deployments Failing?

Before we dive into best practices, let’s look at problem areas. Last year, IDC released a sobering report about enterprise-wide AI failures. “Most organizations reported some failures among their AI projects with a quarter (1 in 4) reporting up to 50% failure rate; lack of skilled staff and unrealistic expectations were identified as the top reasons for failure.” In addition, the report noted the cost of AI solutions and bias in the data were also major factors holding back successful implementations.
“Most organizations reported some failures among their AI projects with a quarter (1 in 4) reporting up to 50% failure rate; lack of skilled staff and unrealistic expectations were identified as the top reasons for failure.”
The MIT Technology Review talked to transformation leaders about AI implementation struggles. There were a few common threads among those currently building and implementing in-house AI projects.
  • Adopting AI and Machine learning is more difficult than the transition to mobile.
  • One of the biggest hurdles is getting disparate record-keeping systems to talk to each other.
  • Data extraction, cleansing and analyzing is time consuming and requires expertise.
  • To develop a smart system requires the time and attention of your best domain experts.
And while none of these concerns were unexpected, experts note that they were more time consuming and resource heavy than originally anticipated. What is apparent is that organizations are biting off more than they can chew trying to build AI solutions from scratch.

How Can You Boost AI Success?

There is good news, too. The 2019 McKinsey Global Survey on AI found that AI has universal value, but superstar organizations are pulling away from the pack in demonstrable ROI. “While most survey respondents said their companies have gained value from AI, some are attaining greater scale, revenue increases, and cost savings than the rest.” The McKinsey team outlined how high-performing companies develop and scale AI in a thoughtful article in the Harvard Business Review. Using their advice as a guide, we are diving deeper into best practices that center around ways that people and technology can work in tandem, and how a little bit of planning delivers outstanding results, especially for those using AI to improve service delivery.

Best Practices for Successful AI Implementation

    1. Identify Clear Goals and Priorities It’s fundamental, but often overlooked. One of the oft-cited roadblocks to successful AI implementations is that organizations try to do too much too quickly — and with limited financial resources — resulting in plans that are scrapped midway through or scaled back. Experts advise starting with a limited scope and adopting a solution with a fast time to market. Choose a project where AI tools can have immediate impact, for example, implement smart tools that spread tribal knowledge as a way to mitigate the wave of retiring experts Skill up the workforce to empower more techs to tackle increasingly complex products. Put knowledge at the tips of your employee’s fingertips, to bolster KPIs, from improved first time fix to NPS scores.
      “Prioritizing use cases based on feasibility, time investment, and value can help leaders balance short term needs and long-term value,”  the McKinsey team notes.
    2. Choose AI Tools that Dig Deep to Analyze Hidden Data Big data becomes a big mess if AI solutions are fed information that’s dirty, incomplete or can’t be read by the system. The service industry is especially prone to ignoring a large chunk of useful data because historically it has been difficult to read, interpret and analyze. To really move goals forward, AI tech needs do more than assist with answering simple questions and play a role in solving complex service problems. To achieve this, consider an AI tool that:
      • Understands and interprets free text, like service orders and notes, in addition to structured data from CRM or parts databases.
      • Combines Natural Language processing and machine learning, which will learn your company’s service language, and better understand the intent of user queries.
      • Draws from siloed information systems and can make sense of disparate data.
 
  1. Adopt User-Friendly AI Solutions For all employees, not just the handful of data scientists on the payroll. Everyone from customer support and those in the field, to executive decision makers should have access to clear, digestible information that guides them to take the most appropriate action.  A smart tool that makes your workforce’s job more complicated is going to fail. Extra bells and whistles are useless if no one understands how to read the data or configure the dashboard.
    According to the McKinsey report: “Companies will need to redesign workflows so it’s easy for employees to incorporate AI insights into their day-to-day activities. They’ll also need to empower front-line workers to make data-driven decisions, rather than having to seek their manager’s approval.”
  2. Provide Training Employee training is critical even when AI solutions are intuitive and easy to use. In addition to technical training, an enterprise-wide approach to change management, focusing on how solutions will enrich a technician’s job, not replace it, is needed, too. 
    “AI transformations are as much a cultural change as a technological one. They require new skills…and new mindsets,” McKinsey notes.
    Like any new technology, it’s not just about the tools, it’s about getting buy-in from your workforce, and creating excitement to use the technology to drive better results. Work with super users and mentors to create an environment when everyone is briefed on benefits, and fear is addressed honestly.

The Solution for AI Success?

Use AI Tools that Seamlessly Mesh Domain Experts and Technology The dirty secret in many AI deployments in the service industry is that in the short-term it creates horrible inefficiencies and drains your workforce. That’s because your best employees are often locked in a room with data scientists for months creating static fault trees that become obsolete as soon as they are finished. A better solution is to look for technology where AI dynamically generates questions and answers based on your service data and tribal knowledge, is then validated by domain experts.  Be sure to ask how long the process will take. If the answer is more than a few weeks, keep looking or contact Aquant because you need your experts in the field, not a conference room. Learn more about our Service Intelligence Platform 

Share this article on:

Share on linkedin
Share on twitter
Share on facebook