AI is helping organizations to create experiences that are more human and contextual. With breakthroughs, however, come anxieties and hurdles. A new study by analyst firm HfS Research raises several red flags about the state of today’s AI market.
HfS’s findings, based on interviews with 50 companies, also echo the concerns Cognizant’s clients share:
- Security remains a top concern. AI-related security issues are more nuanced than most companies are used to. To address that change, we suggest organizations build in a strategic view of security from the start. Our advice is to think of security in two distinct ways. One is protecting systems against hacking and abuse. The other relates to AI’s complexity, such as the origin of training data and how the data is used. This view should also cover the channels by which AI keeps its interactions current as well as a vigilance on those channels and experiences to avoid contamination, such as fake reviews.
- Confusion persists over AI’s legal risks. AI has many layers, and it evolves over time. It requires careful oversight. Corporate counsels often worry that AI’s legal risks outweigh its rewards. For example, because AI only learns what we teach it, are companies responsible for the scope of the training context as well as the training itself? We suggest establishing strong governance and an ethics observation committee to oversee AI’s interactions and learnings from its experiences.
- The vision for AI is to augment, not replace, employees. No one wants to see a collapse of society as AI makes human labor unnecessary. Through our client engagements, we see the most impactful and interesting uses of AI as a decision-buddy, helping to illustrate and guide us, but not to replace us. For example, complementing a call-center experience by noting emotion, fact collection, and other observations that result in a better experience than a pure chatbot or conventional support call.
- AI will shine in the as-a-service economy – but only when data is actionable and accessible. Rapid AI training and calculations, require as-a-service architectures for quick proof-of-concept testing and scalable deployment. A leading health insurance company uses our as-a-service platform, called BigDecisions, to integrate 300-plus distinct data connectors with 200 analytical models, KPIs, and dashboards – and connects to AI cognitive and machine learning services to improve AI’s time to value.
- The need to ramp up machine learning. The biggest AI bottleneck for many organizations is the ability to add enough data and touchpoints for the systems to learn from. The number of qualified programmers remains a limiting factor. Rapid scaling across business functions can only happen when AI is able to learn over time and make use of relevant, unbiased training data. Getting the most value from your data will depend on machine learning from multiple data points.