Organisations are rolling out AI globally to improve customer communications, deliver better experiences and bolster trust. Recent research reveals that ahead of revenue growth, the main goal driving AI investment decisions is improving customer satisfaction (36%). The benefits are definitive, making a more efficient communications network and freeing up customer service teams to focus on more complex issues.
However, there is a paradox emerging. Enterprises are actively rolling out multi-channel agents to innovate and improve customer satisfaction, but a few years into AI implementation, customer communication projects have left pilot stage and hit a critical blind spot. Now, when an issue crops up, it is the customer who often sees it first, not the business.
As AI rollouts continue across customer communications, businesses will need to achieve the right level of governance to unlock AI innovation and ensure that ambitious customer experience projects live up to their full potential.
Assessing the risk at production level
No longer stuck in pilot purgatory, customers across the globe are benefitting from the successful roll out of AI customer communication agents. As it stands, by the end of the year, nine in ten organisations will have a live AI agent in production across one or more of their channels. However, it is at the production level that many businesses are starting to see AI innovation stall.
Many businesses have been racing to the production finish line. But now that they’ve reached it, innovation is being held back by governance failures as enterprises balance AI ambition with customer trust. So much so that almost three quarters (74%) of agents that have already shipped have been forced to roll back or shut down completely.
As a result, engineering teams are focusing on building guardrails and putting out fires, instead of continuing to innovate the solutions that will drive customer satisfaction.
Evolving customer communication expectations
Across customer communications, the margin for error is small. If a customer-facing AI agent fails, the customer sees it immediately, often before the brand does. To them, there is no distinction between the AI, the platform and the company – the entire customer journey is reflective of the brand.
For this reason, businesses are right to be cautious with their agent roll outs.
Customer expectations around how brands communicate have become increasingly discerning. Consumers expect slick, timely and personalised communication with their favoured brands. In particular, Millennials and Gen Z consumers are largely responsible for everyday messaging apps becoming the go-to channel for brands, making it even more apparent when an AI agent fails.
As a result, failed AI interactions can damage trust and create regulatory or reputational risk. In fact, over a third (34%) of those responsible for their organisation’s AI strategy cited a reputational damage as the single biggest impact when an AI agent fails.
Governance as the innovation solution
Governance and innovation should not be opposing forces, but AI’s probabilistic nature requires a new way of thinking about safety guardrails. Done well, governance gives businesses the confidence to experiment. The current challenge is that many businesses are letting regulation requirements consume engineering talent, time and imagination, preventing IT teams from building better customer experiences.
When incidents happen at peak times, such as a product launch, a service outage or usage spike, these can become a real operational and reputational issue. Given that many of these issues are based on PII or hallucination failures, each issue often requires a different approach or solution, meaning engineers have to assess and react to the situation in real time. This quickly escalates, taking precious time away from new AI projects.
However, getting the right governance in place before rolling out AI communication agents means that operations are set on a level of trust that allows the team to innovate. Instead of potential crisis alert, teams have the trust that what they have built is less likely to go wrong, and that they will be able to catch it if it does, giving them the freedom to focus on innovating these products further.
Addressing the infrastructure gap
Engineering teams need to make governance work with them, not against them. This means implementing platform-native controls that bridge an infrastructure gap, as well as the right observability to detect model-level failures.
For example, more than half of enterprises (55%) are custom-engineering the ability to preserve customer context when people switch channels, such as from chat to voice or from WhatsApp to a phone call, rather than having this function built in natively. This is time consuming and could be easily regulated in a native platform.
However, with platform-native controls, engineers are able to come out of the weeds and look at the AI roadmap, rather than the AI architecture. This means they can see what their team is actually building and prioritise which parts need protection and governance, or whether they are working on guardrails that could be provided by the platform. This way, engineers can ensure they are innovating and developing their roadmap, not just working on current capabilities.
Trust is what lets AI scale
As enterprises race to roll out ambitious AI-led customer communications to bolster customer satisfaction, the focus has been on the immediate implementation line. Along the way, governance has taken more of a reactive stance. However, successful customer experience AI agents have governance embedded from day one.
This layer of security needs to be made as efficient as possible so as not to take away time from the innovative talent driving these rollouts. Otherwise, enterprises will be stuck in a paradox where the very thing their AI is supposed to build – trust – will be whittled away by its mistakes.
Robert Gerstmann
Robert Gerstmann is Co-Founder and Chief Evangelist, Sinch, a global leader in cloud communications for mobile customer engagement and business messaging. Its enterprise service portfolio includes SMS, MMS, RCS, WhatsApp and other OTT channels, email, verification, voice, and video.
