Improvements for Customers that don't need AI

“No regrets” initiatives that complement AI driven improvements

Much has been written around what AI can do for business and customers.  One possible risk of this AI trend is that it creates inaction on other fronts, where either AI can’t deliver or where other work is needed to maximise the benefits from AI.  To put it simply, AI will not fix everything. This paper sets out some of these non-AI changes that we think are needed and how they can complement or pay for your AI investments. We have already written about some of the changes needed to exploit AI (See our paper “AI may be harder than it looks”).

We have summarised four areas of change that are still required regardless of any AI investments. These can deliver value in any case, but where relevant we have described how specific AI solutions can complement these changes. The four opportunities span channels, processes, performance, and demand to illustrate the breadth of non-AI opportunities. In short, we are saying that organisations still have work to do in order to:

1. Increase customer take up and use of self-service (whilst AI may improve the breadth and experience of available automated sales and service)

2. Redesign broken and complex customer processes (whilst AI analysis can help analyse these processes and may play a role in their automation)

3. Get staff to follow the right process more often through design, training and coaching (whilst AI can make it easier to be compliant through measurement and process guidance)

4. Reduce unwanted customer contact demand which AI can help report and analyse.

We will explore each of those in turn and describe the role that AI can now play in complementing these changes.

1.      Increase customer use of self-service channels

Most organisations have more self-service channels and functions than customers are aware of. The net result is that many customers still call or message and bypass these self-service channels.  In some companies, we have measured up to 60% of contacts to human assisted channels are for services already enabled through self-service.   For example, customers are still asking when bills are due or to understand baggage allowances or to buy or renew services using front line staff even though all those interactions are possible in apps or online.

In our experience, staff handle these contacts but rarely promote the alternative channels. Often, they lack training or knowledge on what the self-service can do and aren’t trained in the self-service functionality and in some cases, they fear reduction in workload will reduce jobs.  With AI expanding the available self-service functions, companies will have to work even harder to ensure front line staff know what to promote, when and how.  The non-AI opportunity here is to use front line education, promotion and pricing to increase take up of available self-service.  In our book “The Best Service is No Service” we talk about these activities as “customer change management”, namely the process needed to change customer behaviour to try and adopt self-service solutions.  Unfortunately, the failings of the past make this harder. For example, early automated chat solutions had limited capability and often provided wrong or poor answers. Customers who had prior poor experience will need persuading to return to this automation.

One organisation tackled this by re-thinking every human assisted interaction. They decided that every customer contact with a staff member needed to integrate and promote the new self-service capabilities. This meant re-thinking and redesigning interactions to include tailored personalised promotion and then re-training staff on this new way of working. This also meant changing associated metrics and quality scoring. The changes to this manual contact handling created step changes in self-service adoption and customer satisfaction with the options they were being offered.  It’s a great example of the changes needed to the non-AI enabled process in order to make the automation a success.

Once the next generation of AI enabled self-service comes on stream, it should help increase the levels of “containment” in these self-service channels. However, work remains to inform and persuade customers to try this new and expanded self-service.  The AI may improve the self-service functions and add new ones, but customers need to know that and be prepared to try this way of working. It’s not just a case of “build it and they will come”.

2.      Redesign broken and complex customer processes (because the AI can’t)

One financial services business recognised that it had created a very poor process. It was one of many processes that had numerous steps resulting from increased compliance needs, evolving systems and the like. The process design had evolved and grown over time.  The business was so busy managing this process and all the associated work that they hadn’t had time to stand back and look at the reality for customers.  

They recognised that the current process with its paper forms and signatures would have also been complex and expensive to automate using either robotic automation or AI. However, AI was successfully deployed to help them understand the breadth of the problem and the value of a solution.  They used an AI enabled process measurement tool to quantify the number of steps and amount of rework that was occurring in the process. They were shocked when the volumes of issues and the reality of the customer experience was exposed, such as an average 5 touches to resolve a simple process that in theory needed just one.

It took experienced service designers only a few weeks to create a better solution that enabled AI to automate further. These designers worked out ways to harness simple online forms, streamlined process steps and automatic form loading to minimise transcription errors and save customers and business effort. Error rates and rework rates dropped dramatically, and the process took far less time and effort. The AI tools had helped show the problem however human designs skills were needed to create a working solution.

This example illustrates the benefit of investing time in fixing complex and broken processes.  As with any automation tool, AI will work better if processes are simplified and streamlined first.  AI tools can now help understand the problem, but by themselves they don’t fix process complexity.

3.      Get staff to follow the right process more often through design, training and coaching

In any business it's hard to get front line staff to execute processes consistently and compliantly even when processes are well defined and documented.  This can be even harder with outsourcing, offshoring or a combination as processes become remote and issues of culture and language can add confusion.  Nor can AI make staff have conversations in the right way or force staff to use all the tools and resources available to them to service customers even when systems, procedures and tools should make it easy.

In the best operations, front line management or team leaders know “how” their staff are doing the job and then take time to coach them to improve. In our experience, that is the main purpose of this layer of management. We’ve been amazed sometimes to observe the disconnect between the way senior management think the work is occurring to the reality. In one example, half the front-line workforce was logging every interaction in the company CRM as per the process, but they also logged everything in separate spreadsheets to provide parallel but unnecessary performance reporting. Senior Management didn’t know this was happening. This was an example of management not doing sufficient process observation to know how staff were “really” doing the job.

The solution we advocate is observation and coaching by front line management.  They have to understand how their teams are doing their work, know what “good” looks like and coach to that. AI can help but it can't make that happen.  For example, the latest AI enabled real time analytics can monitor what staff say or do and serve up knowledge articles to assist. AI enabled reporting can provide more sophisticated views of performance. Financial institutions are using real time analytics to provide immediate warnings of compliance breaches and process fails. These solutions provide automated quality assessment on every interaction or process not just a sample and can trigger actions quicker. AI can provide far more data on performance and opportunities.  AI can offer all this support, but it can't coach employees and make them change.  Only humans can do that (at the moment!).

4.      Reduce unwanted customer contact demand (because AI can report but not act)

In our experience demand management is a process and discipline enabled by effective analytics and reporting. The latest AI can remove the need for staff to log contact reasons and the tools can integrate data to give a much more complete picture of the costs and volumes of contacts. However, a process is then needed to use that data to drive change. The analytics cannot evaluate root causes and design effective solutions to demand problems. That needs a process and people.   

Companies are using AI tools to help organisations understand and report on the demand for contact, but the AI tools alone won’t reduce the demand.  For example, a recent client organisation implemented AI analytics and reports continuously in real time on the drivers of all contact.  The reporting has been live for over a year but the rate of contact per customer was still rising. The data and reporting exist but the organisation has no process to use this data to hold people to account for the demand and no process to tackle key drivers of demand. The AI reporting shows the problem, but no one acts on it.

One super fund had even given up on the AI contact reporting because it had been poorly designed and reported 600 different contact drivers.  Each was so low volume that none of them were worth fixing.  A one-off contact sample with a deeper dive and people analysing the contacts identified root causes that could be fixed.  For example, many of the customers seemed to be chasing things they had requested. We were able to trace that back to the way the back-office teams were doing their work and were able to quickly address some root causes and prevent the calls and emails.

In another company, there was effective reporting of demand, but misaligned incentives so senior executives saw the information but had no reason to act on it. In our book, The Frictionless Organisation (see www.frictionlessorg.com ) we describe the way AI can be used to improve reporting and help prioritise opportunities to change demand.  However, we also describe the processes that will get value from that information. 

Summary

The examples we have shown here are just four ways in which organisations can make customer experience better with or without AI.  We hope we have illustrated how AI can help but how it is at best, part of the answer. We’re always happy to provide further information on the ideas so please get in touch with David or Peter (see www.limberidge.com.au/ourteam) or email to info@limebridge.com.au or call 0438 652 396.