Since the advent of the PC and home broadband, we’ve already seen significant changes in the way people work, with a rise in the number of self-employed and microbusinesses, more short-term and contract workers, and more people working at home.
This has all had profound implications for both enterprises and the service providers supplying them. Enterprises are increasingly buying and behaving more like small businesses, increasing numbers of contract and homeworkers present new challenges, and employees themselves expect a more consumerised experience of connected services.
Will robots take over the enterprise?
As significant as these changes have been, even more changes are in the pipeline. In the most nihilistic of these future visions, intelligent robots will literally take over the world, and most of the jobs in it.
The pro-AI camp foresees AI replacing human workers rapidly and aggressively. The anti-AI camp thinks AI cannot replace human judgement and levels of service for the foreseeable future. The truth, as always, probably lies somewhere in between.
But how will AI impact the demand for telecoms services and the customer experience provided by service providers?
Benefits and challenges of current AI
AI offers the possibility of reducing human error, speeding service provision, reducing call centre waiting times, reducing costs and dealing with mundane tasks. However, there are a number of challenges with current AIs – including the cost of keeping them updated, their lack of emotional intelligence, and their focus on rules rather than commonsense.
A cautionary tale recently derived from Amazon’s experience, which has scrapped its AI recruitment tool. The tool had been developed to pick through CVs and select the most promising candidates. The problem was that because the AI used data from the preceding 10 years to decide which candidates were ‘best’ for the jobs available, it quickly taught itself that male candidates were better than female ones. It began to systematically and efficiently downgrade any CV that mentioned ‘women’. The team responsible recognised, and tried to fix, the bias, but the project was eventually abandoned.
While there is no suggestion that Amazon’s recruitment practices are, in themselves, biased, this illustrates the dangers of self-learning tools that rely on historic data. This factor is set to continue to pose a significant challenge going forward, because the skills that would have made you successful 10 years ago, are not necessarily the same as those you will need in the next 10 years. Relying on a past record of success has never been more fraught with danger.
In order to address this issue and help AI developers, IBM recently launched Fairness 360 Kit, a tool designed to detect bias in AIs. It helps developers see how the algorithm is actually working, as well as the data it is using to make decisions.
Current service provider plans for AI
Omnisperience recently conducted some research for BriteBill, who were interested to understand what new technologies service providers intended to deploy. We asked tier 1 service providers if they intended to deploy AI into their organisations. 50% of the 40 companies we spoke to told us they either had, or imminently intended to do so.
While AI can undoubtedly be a useful tool, it’s important to consider the pitfalls and challenges when deploying it in the telecoms environment. In the most common service provider scenario, it is being deployed into contact centres to supplement human agents. But take billing enquiries, which typically account for one-third of total enquiries in a telecoms call centre (sometimes far more). While AI can handle a menu of common enquiries or already-resolved simple problems, it struggles to deal with complex bill enquiries which may be difficult to explain and resolve.
Although AI may alleviate the pressure on first line care, by acting as a filter and resolving simple enquiries, it usually doesn’t resolve complex enquiries which still have to be dealt with by specialist (expensive) human agents. While certain customers may prefer non-human interactions, upset customers may prefer to talk to a human or reach the human agent even more frustrated if they failed to have the issue resolved by AI.
The danger comes if service providers rely too heavily on AI without taking action to resolve, pre-empt and learn from enquiries. In such a scenario poor customer experiences will proliferate, customers become more frustrated, and waiting times will rise, as even AIs can become overwhelmed. Performing root cause analysis and analysing data to create virtuous circles has never been more important due to the volume and velocity of innovation that service providers are planning. (Wherever there’s change, there’s also the potential for frustrating customers.)
Using automation and tools such as AI in the right way has great potential to improve business performance and customer experience; used in the wrong way, these tools could make things worse rather than better. (In much the same way that IVR and outsourced call centres were hated by customers.) As is often the case, it is not the AI or any other software tool that’s the problem, but the way these are implemented and utilised.
In short, AI has the potential to be a powerful ingredient in a good customer experience, but it is not a panacea. It can be utilised to reduce pressure on human agents, freeing them up to engage in more positive customer interactions that build loyalty and enhance experience, rather than just have them continually fire-fighting. For B2B service providers, AI is going to be an essential tool for those targeting the SME market in particular. And, as automation and AI become more mainstream, it will get particularly interesting when intelligent agents and devices (within enterprises) request support from other intelligent agents (within service providers). Machine talking to machine sounds like the peak of efficiency, but as we’ve seen from the Amazon example, it also has the potential to go badly wrong without adequate supervision and good design.