Artificial intelligence, especially generative AI, holds remarkable potential for companies. In the next 12 months, AI will begin to augment staff across nearly every department, from service to marketing, demand forecasting, and sales. In the next 24 months, AI will replace human staffers for basic tasks, freeing human resources to tackle more creative and compelling deliverables. However, delivering practical AI in the enterprise is not easy.
AI faces substantial barriers in the frequently risk-averse enterprise environment. Some of these challenges include data sourcing, labeling, AI model development and deployment, resource alignment, intellectual property protection, user adoption, risk management, and ROI reporting, to name a few. In parallel, poor deployments can potentially harm companies and lead to PR crises. On top of this, if we factor in the AI talent shortage and the relative immaturity of many AI solutions, it's not surprising that companies are being extremely cautious, with many even limiting internal access to third-party AI solutions.
The imperative, however, is not simply to stay safe in a world increasingly powered by AI but to leverage this momentum to drive performance. While AI will benefit many divisions and departments, the most fertile ground to successfully launch it will be the service organization.
AI, particularly the more advanced generative AI, is remarkably thirsty, requiring massive amounts of properly labeled data. And unfortunately, the open internet alone will not close this gap.
Sourcing data is critical to generative AI development. This data must be ethically and legally sourced, particularly given emerging regulations around AI data sourcing. Furthermore, this data must be human-generated. As digital writers and publishers increasingly turn to AI-generated content, generative AI models trained by internet-sourced data will face quality degradation. Rice and Stamford university researchers call this phenomenon MAD, or model autophagy disorder.
Once human data is sourced, it must be properly labeled. The cost of labeling this data can be astronomical, leading some to adopt questionable labor practices.
The organization best suited to practically train and deploy next-gen AI in the enterprise will be the contact center and service teams. This is because the service team has the following four key characteristics:
- Data: Every service team sits on a massive repository of product documentation, support processes, support transcripts, and support imagery (photos and videos), and data on the issue (case summaries) and resolution data. Furthermore, ongoing customer and human-agent interactions supply a never-ending fresh data source.
- Business Need: Service teams face ongoing staffing shortages and training challenges.
- Use Case Fit: Service teams face a high rate of repeatable issues that can be more easily identified and resolved.
- User Interest: An internal pool of users (support agents) with tight resolution time and resolution rate metrics.
The Technology is Nearly Ready.
A growing crop of technology providers (disclaimer: I work for one) are introducing AI and other automation solutions to extract and generate highly accurate AI models based on the textual and visual data organically created by service teams' everyday operations. In other words, this new crop of technology is self-generating the AI models needed for service automation simply by monitoring teams' normative operations.
Unlike many early examples of generative AI, service automation requires AI that can understand, diagnose, and guide customers. This will yield a new crop of multimodal and multisensory AI solutions capable of blending text, voice, and visual inputs and outputs for a truly natural user interaction across every touchpoint and channel.
Once a particular AI model is developed, it must be thoroughly tested and tuned. Once again, the service organization is uniquely suited to this task. Support agents have the user incentive required to adopt AI assistance technology: their own performance metrics. Agents are measured by time to resolution and resolution rates, both of which require knowledge and skills well-suited to AI assistance. As agents interact with AI assistance technology, they provide valuable signals to the automation provider. This closed learning loop enables continuous, largely automated improvement.
Like many technological advances, AI will undoubtedly shift the traditional role of the contact center agent. This shift will not occur overnight nor spell the end of this $500 billion global industry. Like most innovations, changes will start slowly at first, building experience, expertise, and momentum.
Over the coming 12-24 months, technologies like generative AI across text, voice, and visuals, coupled with the closed learning loop within internal environments like the service organization, will yield increasingly mature AI automation. Some teams will find that AI automation is particularly well-suited to some challenges but not others. There will likely be scenarios where AI cannot technically provide satisfactory automation. Similarly, there are scenarios in which customers require empathy, creativity, or trust that only humans can provide.
Generative AI isn't yet here for most enterprise leaders, but it will break through in the next few months. As you take your first steps toward this transformative future, start with what you know best; your team, products, services, and customers. In an innovation market thirsty for data and resources, customer service has never looked toward a more promising future.
Jon Burg is vice president of strategy at TechSee, a pioneer in the emerging field of multisensory generative artificial intelligence for service organizations. Prior to TechSee, he led strategy and product marketing at multiple companies including Bringg, AppsFlyer, and Conduit.