Support, CX and overall CRM processes automation and optimisation.

Customer Support automation. Practical perspectives.

Executive Summary

This article is for managers and management consultants who are aiming to optimise their Customer Support and Customer Experience processes. Its core idea is to help Support professionals to avoid common pitfalls and misconceptions in transforming Support and CX departments. To do this I discuss below most ROI-efficient optimisation directions and share practical aspects of process automation in Customer Support environments.

In a ‘Short Introduction’ section I set the scene by quickly covering the following topics:

• What is Customer Support?
• Why do Support processes have big potential for automation?
• What is automation for Support?
• How is automation transforming Support organisation?

Aiming to be as practical as possible, I look at key problematic areas of Support processes and show how they can be optimised and automated, pointing out major possible pitfalls in their automation journey. This is section ‘Support optimisation directions’ which focuses on the following:

• Productivity
• Demand forecasting
• Workload Backlog
• Workload distribution
• Knowledge as a reusable resource
• Automation as a job killer

A dedicated attention is given to benchmarking automation and outsourcing. That is explored in sections ‘Automation or outsourcing?’ where different implication of Offshoring and Outsourcing of Support processes are looked at with comparison to automation as an alternative.

Key outcomes and aspects of automation at Support are summarised in ‘Conclusions’ section at the end.

Please feel free to reach out to me with any ideas or feedback at max.somov@gmail.com

1. Short introduction

1.1 What is Customer Support?

Customer Support is a big part of CRM operation of any business. It can be described from three co-existing perspectives:

• internal business process: It is creative, knowledge-based, problem-solving and consulting-type of business process.
• external service to customers: It is iterative human-to-human type of service, where customers are full co-creators of the end-results.
• sales and marketing tool: It is effective after-sales and customer-retention tool. It is also personalised and data-rich marketing channel. It is key building block of CX paradigm.

1.2 Why do Support processes have big potential for automation?

Customer Support is a rewarding area for automation and optimisation within organisation`s value chain landscape. It has rich hidden sources of comparatively cheap efficiency gains and cost savings. Optimising and automating Customer Support can bring visible return on investment (ROI) for the following four reasons:

• It is labour-intensive process, which includes a lot of administrative activities and cognitive efforts. Humans (experts) are the main assets and productive units at Customer Support. Many people are required to work at Support by nature of the service and in many companies Customer Support is one of the biggest departments by headcount dragging on excessive HR-related overhead. Here automation may help to reduce headcount. This is bare truth, but also a somewhat old-school view on automation. We will further explore it in a more modern business-oriented and operational way in this article.
• Humans are costly, or more precisely, they have significant alternative costs. Meaning that if you employ this human on low-output, low-ROI activity, you are losing (or not receiving) a potential benefit of assigning same human to a more effective activity (with higher ROI). Thus, if you optimise this assignment mechanism, direct gains and cost-savings will become visible. We look at this point in details further on.
• Human-driven support productivity is limitedly scalable. Experience shows that productivity can be boosted by motivation, but its effect is transient, and, in routine processes, there is always a natural normality level of maximum output to which individuals and teams tend to revert. Same time, demand for support is often seasonal and fluctuating. It is external and out-of-control factor for Support. Volumes of demand are usually hard to foresee or predict. Limited scalability and fluctuating demand together define a classical dilemma of capacity management at Support, where over-capacity is expensive in labour costs, while under-capacity is expensive in its possible overall impact on business. We explore optimisation and automation options for capacity management further in this article.
• Quality and effectiveness of provided support are the first influencers for customer experience and satisfaction. Improvements in customer experience bring tangible benefits in stronger customer retention, loyalty, higher sales and upselling leads. Better Support gives extra strength to many other competitive advantages of the company. However, a great care must be taken before ‘optimising’ Support experience, as it may also lead to costly failures in the face of the customer community. We look at the main pitfalls of automation along with value-points further too.

1.3 What is automation for Support?

Support business process automation is an optimisation exercise, and it comes with two main focuses. We call them HOW and WHAT for simplicity:

• Focus on “HOW”: meaning how workers are involved into the process. Or in general terms — how the process looks like and operates. ‘HOW’ represents intertwined proportion of humans and machines engaged in the sequence of process steps.

‘HOW’ focus is covered by so called replacing automation: some steps or human-led effort will not be needed anymore as it will be automated (replaced by machines and software). Here automation will reduce man-hours needed to facilitate the process resulting in labour savings.

• Focus on “WHAT”: meaning what kind of actions workers do throughout the process. Or in general terms — what are the inputs to the process required for its smooth and efficient functioning. This covers tools and data available to workers. It also includes workers` skills and knowledge required to successfully operate in the process.

‘WHAT’ focus is optimised by so called assisting automation: some human-led process steps will be assisted by machines. Here automation will augment workers` skills with ready-to-use data and sophisticated tools resulting in more productive labour.

Optimisation and automation of Support will drive efficiency gains on both these ‘focuses’. Improved skills, automated tools and optimised process steps together will lead to more scalable and performant operation. HOW and WHAT will deliver best results but only in combination with each other. Thus, it is beneficial for the organisation to keep balanced and consistent approach to automation. Advancing on these two directions in parallel will allow to obtain an amplified effect of HOW and WHAT.

Here comes main pitfall: over-automation. Support has clear automation boundaries due to its human-to-human nature (explored in mode details in main part of the article). It is very important to know these boundaries to be realistically pessimistic in your expectations and optimistically challenging in your goals regarding Support automation.

1.4 Automation is transformational

Automation brings changes and these changes will be transforming the organisation.

Transformation will come in three key aspects:

• Processes: sequence of steps will change, steps will change, human involvement will change, process landscape will change.
• Digital transparency: more and wider data will be generated and flowing through the organisation. Inevitably this data will be available to broader internal audience. As the result, there will be fewer grey areas to hide inefficiencies. Such transparency will demand change in participants` mindset.
• Management culture: roles and practices, decision-making criteria, evaluations and success drivers will change.

As Support is labour-intensive environment, any change in it will affect lot of people and will be felt and seen throughout the whole organisation and outside of it: employees, management and customers will all be affected. Organisations must be aware of such transformational changes and need to manage them upfront while embarking on their automation journey.

People organisations are naturally change-resistant, and automation will not solve this. However, success of Support automation and optimisation critically depends on the behavioural and mental change of employees and especially management. Simple awareness of the changes will be a good place to start, but organisations should not stop there. Significant benefits will be realised out of wider employee involvement in co-creating this change. Thus, it is important for the organisations to target broader consultations and participation of employees and team-management in these transformations.

Here comes another pitfall: change momentum. There is clear and practically proven gap between the speed of changes brought in by automation and how fast organisation itself can catch up to these changes. It is important to proactively work on this momentum gap and not to address it in hindsight.

2 Support optimisation directions

2.1 Productivity

Automation is a productivity tool of business. The purpose and the end-result of automation is that organisation can do more output with less input resources. This is then a basis of ROI of automation: value of increased output versus costs and efforts of automation project.

There will be a range of IFs and WHENs, myriad of details and practicalities, hours of analytics and strategizing, vision and mission discussions, lines of RFPs and vendor comparisons, gating and approval chains, versions of MVPs and prototypes and so on… But the true underlying reason of it all should remain clear: you invest in automation expecting end-results.

You can be trapped by similar talks on importance of intangible aspects of ROI: qualitative improvements can be counted-in, strategic goals can be weighted-in, customer and employee retention, brand image and recognition, market share, upselling leads, R&D feedback loops etc. It all certainly can be and should be considered, but again, the essence will remain — returns versus investments.

For Customer Support business this all will mean that you target higher number of resolved customer issues with either same or reduced headcount. Why so simple? Because headcount is main productive unit at Support; and resolved customer issues are main output of Support. And again, you can come with any visionary and business-critical justifications for automation (or against it), but the essence will remain — more resolutions (output) with less human effort (input). Or in other words, you probably don`t want to spend money on automating to eventually end up in a situation with lower number of resolved issues, which will require more human effort.

Automation in Customer Support environment will bring four especially important gains:

• Quicker resolutions: average case is resolved faster. Less time is taken between case creation by customer and case resolution by Support.
• More scalable capacity: your overall capacity increases and especially with regard to ‘easy’ and straightforward cases. Capacity becomes more elastic, bigger parts of sudden workload volumes can be addressed ‘automatically’.
• Higher customers satisfaction: customers become happier, as effort required from customer side decreases (on top of already mentioned decrease in resolution speed).
• Higher employee satisfaction: effort required per case by Support employee decreases. Employees will have bigger value-added time proportion in building up the case towards its closure.

Here comes a pitfall: automation is panacea (it will be new magical world where all problems will be resolved). It is not (and to resolve all problems we, humans, and managers, still have to work a bit more). To get positive ROI organisations must be very prudent on what exactly they are automating and why. For example: you do not automate a process, you automate a step. And within that step you automate this operation because it creates added value. Efficient worlds are efficient in the details.

So how exactly automation can increase productivity at Support? How to be prudent? All good outcomes start with the right approach. What is Customer Support again? — it is a call and problem management process (CPM), it is a predictable sequence of rational steps. Automation loves sequences. And this is the key in the Support automation approach. Going through your CPM process steps, analysing them, validating their costs\benefits — is the first thing to do. If you are lucky, some steps would not hold to their purpose value and thus can be eliminated (although it does not happen that often). Those steps which will survive this analytical purge, will each have their own optimisation potential. Your organisation`s automation capabilities will be your main source of inspiration at this stage.

Besides sequences, automation also loves everything digital. Thus, more digital data points you can have in your steps — better you can teach your algorithms. Better and more sophisticated algorithms can replace more human operations (see above — replacing automation) or can better assist your employees (see above — assisting automation).

Savings in effort or in man-hours received out of the optimised and automated steps will sum-up into overall productivity gains. These gains will shine nicely in your ROI formula together with all improved qualitative aspects (do not forget to count in here increased workforce happiness and engagement, if any). More creativity is better, but you also must be very prudent still and especially in quantifying those ‘qualitative’ variables.

Automation itself is a journey which starts with the best intentions and courage. It also usually starts with absolute lack of realistic assessments on what is good to automate (or to tackle first) and what is actually feasible to automate based on your organisation`s capabilities. To help to be realistic, we proceed further by looking at best automatable areas in Customer Support.

Most of automation activity, especially on its initial phases, will happen in ‘motion’-type of steps of the CPM process. For example — evidence collection and filtering-out the errors or important events. As organisation will mature in the automation journey, more operations of cognitive manner will be possible to automate. For example — detecting root cause event; as now your algorithms will be clever enough to find important events, with time (and effort for teaching the algorithms) they will be able to pin-point triggering patterns and thus root issues causing the whole case.

Below are the key areas of CPM process where automation can bring benefits:

• Information collection and preparation: RPA, chatbots, remote logging, feedback loops — are the technologies to look for here. That includes symptoms elicitation from the customer too. This is mainly replacing automation. Goal is to eliminate (replace with machines) employee`s effort for gathering the right information. It is often estimated as 30–40% of the time spent by Support — simply to collect right, consistent logs and data around the issue.
• Actual investigation and root cause analysis: this is a lot of human cognitive activity. Replacing automation will be a hi-end futuristic goal here, thus it is reasonable to target assisting automation options at first. Realistic estimations on what and how can be automated here is a very good input for feasible ROI in the end. Text-mining and pattern recognition AI technologies are to look for in this area.
• Knowledge management and reusable common solutions catalogues: this is where the cases are resolved during their creation phase (an effective self-help phase), or where employees are reusing already existing template scenarios for resolving cases. Thus, it is mix of replacing and assisting automation where search engines, knowledge bases, a bit of AI and human centric knowledge recording processes are put in place together. KCS methodology will commonly cover the process organisation part, while leading vendors can offer effective text indexation and matching algorithms here.

Success of above will heavily depend on several parallel areas. Key of them are covered further below. For example: data meta-level in organisation; changes in managerial roles and priorities. A broader topic of capacity management at Customer Support, represented by demand forecasting, backlog management and workload distribution, together with other organisational aspects are also covered further in this article.

Here comes another pitfall: automation will leave my employees with only very complex cases. At first, it is good to realise that customers will not start suddenly to create more complex cases knowing that you started an automation journey. Thus, actual complexity of your cases will not be changed. What will, in essence, change is that your workforce will have more time now to address complex cases. Again, those ‘remaining’ and predominantly ‘complex’ cases will not take longer to get resolve than before.

Although you still may rightfully think that, as remaining cases are only ‘complex’ cases now, then average resolution time or similar legacy KPIs will get worse — this is exactly the momentum pitfall (see ‘Automation is Transformational’ section above).

2.2 Automation or outsourcing?

While covering automation, it is appropriate to devote some attention to Outsourcing in Support as well. Why? — Because organisation will naturally seek out easier ways to resolve “do more with less” problem. And for many organisations, who are facing this problem, outsourcing and offshoring (O&O) are becoming go-to easy alternatives to automation. This common managerial way of thinking has good practical reasons behind because of the below O&O properties:

• Straightforward to implement.
• Within managerial control.
• Can be governed by directive decisions of management.
• Do not require deep analysis and venture-type of investments, thus seemingly less risky.
• Create sort of importance illusion — more headcount, international-team flavour etc.

All this facilitates popularity of O&O as a much more management-friendly, easier and more reliable alternative to automation. Thus, to be more exhaustive in our analysis, it is practically necessary to reflect on such views. This will also help to present automation in more realistic terms by benchmarking it with O&O*.

*O&O is studied here only with regard to automation, as alternative tool to reduce labour costs. Obviously, O&O may have its own strong business reasons to exist. For example, when you do need to deliver follow-the-sun 24/7 customer support, or you need localised support for your customer community in particular location. This aspect is outside of benchmarking automation to O&O here.

2.2.1 Offshoring

Offshoring in Customer Support is a longer-term alternative. It involves setting up a remote cheaper labour costs location and then ‘educating’ it up to the level of performance and quality, at which CPM process is provided at head quarter. The question “To offshore or not?” then can be rephrased as “Expensive automation or extensive globalisation?”. Offshoring of a labour-intensive Support process will have clear disadvantages and a somewhat peculiar advantage.

Let`s start with disadvantages of offshoring in Customer Support:

• High staff turnover (developing markets: you teach them and then they leave for higher salaries).
• High investment in educating each employee (costly knowledge and expertise transfer).
• Span of managerial control, time-zone spread and thus required 24x7 management.
• Demanding integration of people and processes (cultural differences, islander mentality etc.).

Advantage of offshoring lays in coexistence with automation. There will be two options here:

• You offshore what is technologically hard to automate.
• You offshore what is expensive to automate.

Both options have own complications though, but they both mean, that organisation first need to study what is automatable and what is offshorable.

Pitfall: lack of root cause analysis. In practice, it often leads to an opposite picture: organisation will offshore simple mechanical tasks, which are easy to automate. It is justified by straightforward ‘logic’: I have many mechanical tasks; it means, that I need many people to do it. Many people will cost me a lot; thus, I need to find a cheaper location.

Alternative logic here can be as follows: many mechanical tasks, means many similar simple algorithm-like tasks, meaning I have my first business case for automation project and my solution here will be — RPA + chatbots — assisting or replacing automation.

While seemingly inexpensive at first sight, offshoring can be a good partial alternative to automation. However, it is important for organisation to make weighted decision in this question of “Expensive automation or extensive globalisation?”. It very-well can be, that sunk costs of doing due-diligence on this decision with internal or external analytical unit can in the end pay back.

2.2.2 Outsourcing

Full clean outsourcing of Customer Support may only make sense for the very new business lines. For legacy organisations, who historically already have in-house Customer Support teams, much more adaptable option will be temporary outsourcing. By foreseeing lack of capacity for a coming peak period, organisation can consider ‘seasonal’, temporary outsourcing to cover the peak.

Majority of the disadvantages of seasonal outsourcing are like the ones mentioned above in Offshoring chapter, but some will bring extra difficulties:

• You need an already up-to speed and knowledgeable workforce on the market, as you basically do not have time to educate them. Or, otherwise, learning curves should be too steep to be realistic if you want to avoid a drop in service quality.
• Shorter-term engagement of already knowledgeable experts will cost extra.

Main, and possibly the only one, advantage of the seasonal outsourcing is a so called ‘capacity on demand’. You arrange extra capacity by outsourcing in peak periods and this allows you to have no over-capacity during the rest of the year. However, this advantage is based only on the standpoint that having over-capacity is bad and expensive. This point is questionable for customer support environments.

Pitfall: the belief that we need to avoid over-capacity. Problem here is that over-capacity is usually measured in idle-time or the customer issues which could have been resolved in this period (if there would be any). This view is alike business myopia. As per our definition of Support, support experts are the main productive unit of the whole process. Their productivity is, in the end, a function of their knowledge. In short term, knowledge is endless, and it will not hurt if the over-capacity would be reinvested into further development of this knowledge. Payback on such over-capacity reinvestment is increased workforce expertise and thus higher productivity at peak periods.

Once managed correctly, over-capacity phenomenon can resolve the overall need in O&O expensive options at all.

Rather one-sided negative perception of over-capacity is common among management. It is basically the default way of approaching the problem. Mantra “Over-capacity is dangerous” does have raison d`etre. Natural tendency of a worker is to have a ‘rest’ period after a ‘peak’ period. This resting may practically end up in doing nothing. Then this ‘doing nothing’ is very visible on the shop floor and it is immediately taken as a clear signal of mismanagement.

Reinvesting over-capacity into knowledge build-up is rewarding, but more complex managerial puzzle to solve. Instead of clear administrative decisions, it will require more motivational type of team-management approach. It takes greater effort and time, while the next workload peak might very well be just around the corner. So, we need to propose more realistic plan of action here.

Main practical dilemma for business strategy here is: scaling by outsourcing or investing into scaling by automation? Seasonal outsourcing is not the only option to get capacity slack, because automation can also bring required elasticity in capacity. For example: an algorithm can close 10 cases and 1000 cases of similar type, and possibly somewhat higher electricity bill will be the only price for it.

As Support is a very important channel of improvement ideas for the processes and products of organisation, once outsourced, it will be difficult to maintain this improvement-feedback-loop culture for the periods of outsourcing. Negative effect of outsourcing may get multiplied because outsourcing periods will most likely be the peak periods, and the peak periods are most fruitful time to discover and reflect on new types of issues and bottlenecks in your processes and products. In other words, classical outsourcing may have a serious drawback for Support (if you believe in CI). Organisations should be aware of this negative effect and thus must take it into account once venturing into O&O strategies.

Another point to keep in mind on outsourcing at Support is that even the most advanced subject matter experts from outside of the organisation will still need to be emerged and trained into internal CPM-culture and customs. Those most advanced experts are usually a rare thing on the labour market and are already employed. Thus, realistically, you need to count in a noticeable effort to teach your temporary new hires to perform CPM. This teaching often implies sharing your build-up internal know-how. Then, once the peak period is over and you will be stopping your outsourcing, this newly trained staff will be released back into the market. From the one side it is capacity on demand, but from the other side it is an outflow of local expertise and know-how. On top, the effort spent teaching them will be your sunk costs or alternative costs comparing to automation trajectory.

In general, seasonal outsourcing has several disadvantages to be considered a good long-term alternative to automation. However, in short-term, as a temporary solution — it can survive the critique.

There is also a pitfall: predictable seasonal peak times. We consider outsourcing for a certain coming future period. Meaning we anticipate new workload, meaning that we may roughly know what will be causing this workload. If we have clues over time- and root-cause, we can try to get prepared upfront with assisting optimisation or automation, can`t we?

2.2.3 Automation or outsourcing? — Organisation to decide

In principle O&O are the same operational business tools as automation. There are some differences in their usability though:

• Automation is longer time-to-efficiency, but also longer-lasting benefits once implemented.
• Outsourcing is comparatively quick to implement but not scalable and not flexible. More short-term option. And it implies legacy business process stiffness and HR-overhead.
• Automation is more alike an R&D process: you venture into ‘research’ for the best-possible solution fit for your process and, strictly speaking, efficiency of the end-results is not guaranteed. There are practical risks involved in automation and the outcomes are very much dependent on the automation experience of the organisation (what is usually has not yet been built up).
• O&O is more a command-driven administrative type of solution. It is more under managerial expertise, which is usually not a scarce resource inside organisation. It means, naturally, that from a management perspective there are less risks and clearer path to go.

Despite above differences O&O and automation serve the same purpose for customer support environments — to fight the lack of capacity. Same time, while the purpose is similar, these are two implementation-wise very different ways forward for organisation. Current analysis shows, that on weighted average automation gives comparatively more benefits to the organisation, if not even adding a very valuable transformational experience and confidence. These bigger benefits, though, come with somewhat bigger practical risks.

The decision on automation or outsourcing is obviously context specific. Organisation needs to do its own comparative analysis of the options at hand if there is a problem of capacity. And possibly the main criterion in this analysis will be forecasted behaviour of demand volumes. If it is long-term steady growth, then you need scalability of automation. If it is few ‘waves’ of demand inflow only, then seasonal outsourcing is quicker (but dirtier) solution. We look at demand forecasting in next chapter.

2.3 Demand forecasting

Majority of Support centres are ‘reactive’ to demand waves. They conventionally see the demand (basically number of tickets or issues coming in) as external factor with often sporadic and random behaviour. Such legacy approaches will be first to change once organisation will tap into the flows of data around its processes. Usual path to follow here, which is most frequently chosen by incumbent organisations, is to start with variation analysis led by adepts of Six Sigma and Continuous Improvement. This is certainly a very good step to do, but it contains another pitfall.

Pitfall: demand is uncontrollable. The very notion of demand for Support is the key here. As long as demand is treated as external factor to adapt to (via forecasting or via sustaining over-capacity, as we discussed earlier) — organisation is deemed to bear often unnecessary costs related to workload management.

A crucial step towards whole Support optimisation will be to rethink the demand and not to be afraid to challenge its ‘externality’ and ‘out-of-control’ manner. And automation is here to help, as it can eliminate this managerial prejudice of unknown workload volumes. It can help in two directions:

• First, it allows deeper data analysis. Once processes become more data-transparent, forecasting can benefit just by this, as obviously more patterns now can be auto-recognised and thus can be better predicted.
• Second, it reveals hidden drivers of ‘unknown’ demand volumes. And here management can start to work their way up from being completely ‘reactive’ to become more ‘proactive’ regarding demand volumes.

Such proactivity, enriched with analytics and insights about demand drivers, tools delivered by automation and data out of digital process transparency, in the end, will demystify and tame the wildness of demand. Support management will feel the power to adopt the demand to internal schedules (instead of always adapting to it). Workload volumes and demand for support will get under organisation`s control allowing visible cost savings by eliminating unwanted over-capacity and bringing optimised backlog management.

Demand analysis and forecasting is a rich and rewarding topic which deserves a dedicated discussion. We leave it for now to be addressed separately in full details later, so that we can instead focus on workload backlog.

2.4 Workload Backlog

Backlog is a natural result of capacity mismatch. It is a buffer between comparatively permanent capacity and changing seasonal fluctuating demand. Digitisation of monitoring and estimating capacity lag or under-utilisation is another example of automation as a cost savings mechanism.

Backlog, besides direct `waiting` costs for customers, has another dangerous effect — a never-ending frustration among workers, as they will feel the pressure from the mountain of yet-to-be-resolved cases in front of them. Practical experiments show that such pressure has visible demotivating influence on employees and leads to downward productivity. Although it is often perceived as a good sign by line management (as there are “we are busy” looks and talks on the floor), it still does not cover real negative losses in motivation and productivity.

As zero-backlog environment will, most probably, remain more as a futuristic organisational dream than tangible reality, automation of backlog should be considered as viable business case with the following advantages:

• Digitized backlog monitoring and analytics will allow more operational flexibility for line management.
• Intra-day insights on backlog build-ups and bottlenecks out of CRM data processing algorithms will facilitate effective coordination of the teams.
• Data-based coordinating management actions will increase Support reaction agility towards customers.

And another pitfall of automation is to expect that algorithms will solve it all with no human involvement. Backlog automation here is a good example. Algorithms will show the bottlenecks but the actions to tackle them will remain on human shoulders. Realistic view on what is being automated here and what are the qualitative improvements of it will strengthen ROI numbers and add more credibility to your automation project.

2.5 Workload distribution

There are many examples when Support teams are performing very-well in distributing their incoming work manually. Most of the time it happens in environments dealing with comparatively homogeneous types of incoming issues and with a designated coordinator roles. Such human-led distribution is usually efficient only in small, coherent and motivated teams glued by shared KPIs. As soon as team dynamics changes, or if you get bigger teams, or there is 24/7 multi-location setup — manual workload distribution loses its charm and either gets too unreliable or too expensive (as too many FTEs get involved in its coordination).

In general, manual setups are prone to hiccups, are very dependent on individuals, can cause intra-team tensions, may be unstable in longer-term and take costly FTE resources. On top, they are non-transparent in measuring efficiency. You can see the end-result (say, all issues get assigned), but you don’t have insights on how efficient these assignments were in the end.

Automation is having a good business case here by addressing those weak points and delivering more stable, transparent and scalable distribution systems. Best efficiency in this use case is gained by automating not just the actual assignment actions, but by maximising team productivity through automatic algorithmic matching between currently available employees` expertise and predicted cases` complexity. Simply saying: by assigning complex cases to experts and easy cases to learners. Thus, you maximise total output of the available team capacity and, same time, minimize waiting and resolution time for the customer.

From data science perspective this is not the most trivial, but still rather straightforward and solvable task. First, you need to predict ‘complexity’ of the incoming issue based on very initial amounts of data about it; and second, you need to be algorithmically aware of which employees have best capabilities and skills to tackle this issue in current moment, so that you can proceed with automatic assignment decision-making.

With adding extra machine learning capabilities this algorithm may facilitate employees’ progression: somewhat more complex cases in some proportion should go to less experienced employees to push their learning curve; or to give some easy cases to experts for a break. With time you can build up your algorithm`s artificial intelligence to count-in many other behavioural and individual factors.

From business perspective this intelligent automation will drive team performance and productivity keeping your staff engaged. And it is all beautiful in theory, but not easy to implement. Feasibility of such automation is dependent on the level of digital maturity of the organisation: how streamlined are the processes in general? Are there enough data points inside CPM process for the algorithm to become self-learning and surpass human read-and-see experience-driven coordinating skill?

Here again, avoiding automation pitfalls, knowing clearly what is being automated and why — will define the success. No doubt, that these methods and models bring value, but to be able to materialise this value organisation will first struggle through implementation process. And, although there are many similarities in Support processes among different organisation, in this particular use case, closer you will stay to the specifics of your own CPM process — more efficiency you will get out of its automation.

Besides implementation complexity, there is also a positive side to it: every organisation has invaluable but often overlooked internal source of practical ideas and quick reality checks — this is voice of the employee.

Here comes common pitfall: trust your employees. Most of the time, task assignment is automated ex-ante, meaning somehow everybody set themselves to automate a new just coming-in task. Certainly a good exercise, but here is the pitfall: do not automate what humans can still do better than machines. If you trust your employees (as you trusted them before you decided to automate) — leave ex-ante assignments to ‘free-market no-rules’ manner. Workers will do their best. Plenty of motivational drivers will exist for them to take the right task for themselves. Automate, instead, a coordinating role of assigning not-taken remaining tasks — ex-post automation. Best benefits are there.

In principle, what will be automated here is a set of unwritten rules of the teams and workers on how the cases are flowing through differences in expertise and experience of the team members. Even if at first sight it might look un-automatable, it still can be put into algorithms, but then the main question will be — how realistic such algorithms would be. Realistic means how truly they would be able to reflect these established customs of the teams in assigning cases. Therefore, key success factor in implementing this use case is to consider wider input from your employees throughout all phases of the project and especially in design and testing. This is where the automation should be at its maximum of sophistication, as what you will be automating is indeed very context-specific human decision making in matching foreseen issue`s complexity to specialization and expertise levels of individuals. Do not leave this ‘magic’ to data-scientists alone, mix them with humans and ask for frequent reality-checks.

Although this business case is not easy to implement, it has bright future worth fighting for and its impact can be very significant for your overall Support efficiency and its departmental structure. Even before the actual possibility of its automation came to existence, people were trying to solve it administratively by layering customer support: level one, level two, centre of expertise etc. Now such administrative overhead can be leaned and CI-ed with new, simpler, cheaper, scalable and efficient agile Support frameworks.

2.6 Knowledge as a reusable resource

Can knowledge be reusable in Customer Support? Practice shows that it can be, but partially.

Knowledge by itself is hardly productive. Its effectiveness clearly depends on how easily it can be applied and to whom it is available (how widely). And, of course, at first, this knowledge must be recorded (to become accessible to others). We leave quality and motivational aspects of knowledge recording aside, only mentioning, that more practical and to-the-point the knowledge is recorded — easier it can then be applied. As soon as we have nicely recorded widely available (shared) and applicable (right for the context) knowledge at Support, it indeed becomes a fully reusable productive knowledge. At the same moment, such knowledge turns out to be very ‘basic’ and ‘common-sense’ as now everybody ‘knows’ it. It also means that the issues it can be applied to immediately become simple to resolve as well.

The theoretical framework described above is impossible to implement without a knowledge automation system. Searching the knowledge, linking it to the issue, providing the resolution message to the customer out of it — all these and many more alike practical steps are time consuming and thus FTE-expensive. Automation is the only possible realistic way to provide these foundation pillars of reusable knowledge with reasonable costs.

Although different approaches exist for reusable knowledge (KCS is possibly the most known one), it is important to realise, that effectiveness of even the most automated knowledge will have limits. Complex cases, non-standard cases, exotic contexts and use-cases will exist without commonly accessible reusable knowledge (as such ‘special’ knowledge has not been yet recorded). Thus, automated knowledge platforms will be most efficient for low-complexity high-volume tasks.

As soon as you want to be more performant in higher-complexity (but luckily low-volume) cases, your main productive resource will be individuals, powered with their proprietary personal and not yet sharable cognitive abilities, expertise and knowledge. Namely those brave enough employees who will assign themselves to these complex unseen before types of issues.

For modern Customer Support environments reusable knowledge frameworks are more a license to operate than a power-tool. You will suffer without them (and your customers will hate you too) and, same time, having them in place will not give you exceptional ‘competitive’ advantage. They are a ‘must-have’ and they must be automated.

The last aspect worth mentioning on reusable knowledge, is that, in the end, it will make your workforce more expensive. Once you automate simple tasks, your need for ‘simple’ workers will decrease. You still will need them for human-to-human relationship management with customers, but more as an axillary resource than a core productive unit. Your main battle now will be to retain highly skilled Support experts (to be able to cover not yet created reusable knowledge and respective complex cases). This task will come laying on management shoulders among other changes to their role brought in by automation.

2.7 Automation as a job killer

Is automation a job killer at Customer Support? — Absolutely not. And to explain why not, we need to see what can and what cannot be automated in general.

Three main directions which can be automated:

• Production (basically, the process of resolving cases).
• Coordination (management roles as they are mostly perceived by the industry now).
• Delivery channels (these are self-support, omni-channel approach, chats and bots etc.).

The only thing which cannot be automated is the human-to-human relationship. No matter how fragmented and low-margin your customer base will be, there will always be more important customer groups which you want to retain or upsell. And to do this you will always need a workforce who can understand your customers first-hand. On top, in emergency situations (what is, basically, a norm at support) only human experts can perform ‘crisis management’ with customer on the line.

This is rather obvious common sense, however, it is outstandingly applicable to customer support realities: all man-hours released out of automation of those three above-mentioned general directions will certainly find a good use in this last one non-automatable human-to-human context.

More focus on relationship and last mile human touch together with more proactivity throughout CPM process are always among top Support differentiators in the polls and industry reports. However, in practice they are very rarely met. For example, a simple proactive call to a customer can be a huge investment in loyalty resulting immediately in higher customer satisfaction. But usually, due to constant lack of capacity or high backlog pressure, there is simply no time for such activity. The weight of not-yet-automated workload is pushing support departments into ‘reactive’ mode of working and proactive support remains unreachable or too expensive dream.

Rarity and exclusivity of proactive support makes it indeed a decisive competitive advantage in the industry now. And automation can become the source of skilled context-aware human expertise and effort needed to pave the way to higher efficiency and proactivity of Support.

Conclusions

We have been very successful so far in looking at what are the pitfalls and rewarding possibilities of automation at Customer Support.

We have identified key aspects of automation as a transformational project for the organisation:

• Assisting and replacing automation (two focuses of automation).
• Shifts in process transparency and in management culture.
• Stepping from reactive to proactive support.

We went through the most impactful and same time most painful areas of Support automation:

• Demand forecasting.
• Backlog management and workload distribution.
• Productivity
• Reusable knowledge.

We looked at how organisation should prepare itself to gain maximum ROI out of Support automation:

• Continues root-cause and costs\benefits analysis (What and Why of automation).
• Benchmarking of automation to outsourcing and offshoring.
• Reinvesting resources and over-capacity problem.
• Legacy KPIs and momentum of change.

We shared main practical pitfalls of Support automation:

• Over-automation and almighty algorithms (AI and ML including)
• Late employees’ involvement.
• High proportion of complex cases after automation.
• Uncontrollable demand and predictable seasonal peaks at Support.

Let`s us here conclude our analysis by reinforcing the key practical observation on Customer Support transformation. CRM, CX and Customer Support will be getting more and more automated and digitally rich areas of organisations, but they will still remain human-driven and human-centric activities. And the main question will thus be whether these humans, both customers and employees, will become happier by interacting through these activities.

Automation today is certainly a very mighty tool to deliver added-value and added-happiness into your employees` and customer journeys.