The Benefits and Pitfalls of Intelligent Process Automation – Part 2 of 2

Jun 5, 2020 | Delivery & Operations, Process and Change, Skills and Knowledge, Technology and Partner, Vision and Strategy

Intelligent Process Automation (IPA) has the potential to transform how an organisation operates, delivering multiple benefits to the organisation, its customers, suppliers and partners. Here I consider these potential benefits, discussing how automation can deliver each benefit but also why many organisations fail to achieve these benefits at scale.

Part 1 of this blog series introduced these benefits and the related pitfalls for the first five of the benefits listed below. This blog focusses on the second five. Please read part 1 here if you missed it.

  1. Transforming the customer experience
  2. Improving products and service delivery
  3. Reducing process lifecycle time and increasing throughput
  4. Removing mundane work, freeing people time for higher value tasks
  5. Increasing employee satisfaction
  6. Improving accuracy, quality, consistency and reliability of processes and data
  7. Expanding management and business information and insight
  8. Reducing risk, enhancing compliance and improving resilience
  9. Improved security
  10. Significantly reducing operating costs

Improving accuracy, quality, consistency and reliability of processes and data

Repetitive tasks, when performed by people, such as taking information from one system and entering it into another are prone to human error. These errors will affect the quality of information held within systems and introduce inconsistencies between systems. Well implemented automations can remove this risk of error.

A team of people who have been trained to perform the same task, including following a consistent set of steps, applying a set of documented business rules and performing the same operations on business systems will often not deliver exactly the same outcome. People can interpret and apply rules differently, they can make their own decisions on what parts of a process are important and what are not, they can get distracted and errors can occur ultimately leading to variable outcomes. An equivalent team of bots programmed (correctly) to perform the same task will always achieve the same outcome irrespective of how large the bot team may be.

Additionally, bots don’t get tired, don’t need breaks and don’t get ill.

This all means automation can improve accuracy, quality, consistency and reliability in the execution of the process, the associated data, and ultimately the outcomes achieved.

There is though a big dependency on how well the automations have been designed, implemented, optimised and maintained:

  • The design needs to handle scenarios where, for example, input data is incomplete. People are naturally good at inferring from the data that is available and bridging the gap caused by missing data allowing the process to complete rather than be aborted. Equivalent rules need to be factored into the automated process to achieve high levels of performance and reliability.
  • Business systems often aren’t always on and, when they are, may not be 100% reliable. Automations need to be able to cope with this so that they can provide consistency even when the underlying systems aren’t particularly performant.
  • Process optimisation is a key consideration for the automation lifecycle. It is unlikely that process mining and design will capture all possible scenarios (and spending too much time trying to do this can often be a false economy). So, performing incremental delivery of functionality and factoring in optimisation post go-live is a good approach, as long as unsupported scenarios are handled gracefully by implementations and don’t introduce errors into business systems.
  • Solutions need to be implemented with maintainability in mind, this needs to include gathering extensive performance information and handling exceptions well. Changes can occur to input data over time and business systems can change unexpectedly (even when upgrades are managed). Reactive support needs to be combined with proactive maintenance to maintained levels of accuracy, quality, consistency and reliability.

There are other factors to consider, such as reuse, modularity, implementation consistency, testing, change management etc all of which can impact how well automations perform. Some of these are covered in my blog series on intelligent process automation, the first blog can be found here.

Expanding management and business information and insight

Business systems maintain their own operational and performance data, but processes often involve multiple systems. Understanding process performance has therefore traditionally been a challenge as information needs to be extracted from multiple systems to get a full picture. Automation provides an ideal opportunity to gather this information and to make it readily available to deliver greater insight on the performance of the organisation and, ideally, to identify other opportunities for process/business improvement.

The automation platforms typically provide a capability to monitor the performance of the automation workforce. Providing a view of the utilisation of this workforce and the volume of automations executed, the level of success and the time it took. This information is useful but does not provide information that will deliver insight into the specifics of, for example, the reasons the process is running, the level of service being provided to an end customer, the decisions made within the automation etc. A MI/BI gathering and associated implementation strategy needs to be identified early in the automation journey to ensure a consistent approach across automations and ideally to drive the creation of an actual framework for gathering and recording useful business/process information for all implementations to follow. This should also be linked to a strategy on how this information will be used whether via inbuilt platform capabilities or external analytics and visualisation tools.

Note, GDPR regulations also need to be considered if customer data is being recorded by the automation platform (see below).

Reducing risk, enhancing compliance and improving resilience

There are many types of risk relevant to automation including security risks (see next benefit), delivery risks, reputational risk, risks to customer satisfaction, financial risks etc. Various risks are reduced by improving the accuracy, quality, consistency and reliability of processes and data as discussed earlier in this blog. Other example opportunities to reduce non-security risk, enhance compliance and improve resilience include:

  • Implementation of automations that specifically monitor for risks and then proactively do something to reduce the risk. For example, checking that products/services will be delivered on time and, if not, alerting another automation or even escalating to a person. Another automation could be used to check system availability across the enterprise architecture, monitoring performance and proactively responding to degradations in system performance that could negatively impact automations or the human workforce.
  • Automations can be implemented to enforce compliance, encapsulating the necessary business rules to enforce compliance to legislation and industry regulations. Automations will consistently apply these rules, whereas people are prone to being inconsistent both individually across a working day and across team members.
  • The automation workforce can be scaled up (and down) much more quickly than the human workforce. Providing the ability to react quickly to changing demand and unexpected impacts on the business (for example, a pandemic!). The automation workforce is dependent on the IT infrastructure used to support it. Typically, this is now cloud based, benefiting from the resilience provided by cloud deployments. The automation workforce is also not impacted by office problems, illness and other interruptions that can impact the human workforce.

Automation does though introduce new risks which must be managed by the automation strategy, for example:

  • If the automation platform fails (at a process or system level) resource needs to be immediately available to resolve the issue. This is likely to be a new 24×7 support capability as automations are often expected to not be constrained by human working hours. Where an automation has been put in place, it is typical that the human workforce which previously performed the process manually is no longer available, at scale, as a backup. A failure of automation can therefore have a major impact on the business. A successful automation programme will rapidly mean the automation platform becomes a business-critical system.
  • If an automation isn’t implemented well and, for example, has an error in a business rule it can very quickly apply the error in thousands of instances of the business process (much quicker than an individual human worker could). This requires a proper testing and rollout strategy to minimise the risk of this occurring.
  • The automation platform will likely process sensitive data, including people, service and commercially sensitive information. There is a significant risk that automation developers will be tempted to record this information within the platform (to aid debugging or as a valid part of the implementation). This needs to be controlled by policies and review. It is easy, for example, to become non-compliant to the GDPR ‘right to be forgotten’ if the automation platform has recorded an uncontrolled copy of customer data.

Improved security

Automations are often replacing the need for people to interact with multiple systems, often using many different system credentials. Unfortunately, people aren’t good at remembering multiple credentials and will have a tendency to write them down or use easily guessable passwords which pose security risks. Automations need the same credentials, but these can be maintained within encrypted databases that are only accessible to the automation account (e.g. bot) and approved administrators. Initial administrator assigned passwords can be changed by the bots to complex randomly generated passwords when appropriate automations are put in place.

Many automations run on centralised environments (e.g. unattended bots). This centralised approach provides a single controllable environment to secure, reducing the dependency on distributed local environments still used by many human workforces. Attended automations will still often run local to the user but under their control and through their access permissions.

In addition to providing more management information, automations can be configured to provide enhanced security information. Providing greater traceability and transparency in when a system was accessed, what information was changed and by who.

Best practice security approaches have to be followed when deploying and utilising an automation platform otherwise it can introduce additional security issues. For example, dev/test/live environments should be separated (with different accounts/permissions for each), live access should be limited to only those needing to provide support and maintenance, sensitive data and automations should be held in encrypted format, communication between automation components and business systems must be secure, user accounts given to bots should be assigned roles/permissions appropriate to the operations they need to perform on each system and not, for example, admin access.

Significantly reducing operating costs

Operational cost reduction is typically linked to a reduction in the size of the human workforce and the associated saving of direct (salary, bonus etc) and indirect (IT, management, office space, etc) employment costs. Many automations, particularly where acting on medium to high volume processes, will deliver significant human time savings providing the organisation the opportunity to cut costs or alternatively invest the new time available to the human workforce in supporting business growth.

When looking at cost reduction it is important to consider the following key factors:

  • Automations cost money to design/implement and test, the cost of the automation platform (infrastructure, setup, licensing) is significant, support and maintenance costs can be high (particularly in changing environments) and the resource needed for the programme can be difficult to find and keep. Automation resources will often cost a lot more than the operation resources automations may be looking to replace. Automations which deliver a small time saving may therefore not deliver a positive return based on employee cost savings (see point two below though). It may also be difficult to realise an employee cost saving if the saving is just a small fraction of the work performed by an individual or a small team. This highlights the need for the automation programme to be efficient and also for other benefits to form part of the driver for automation.
  • Ultimately, there are 10 benefit areas in this benefits list (not just 1 i.e. cost reduction) with each having value to the business. To build initial business cases and to demonstrate the value an automation is delivering it’s wise to consider all the benefits it will provide and if necessary (for the finance team!) to associate a financial benefit against each. For example, improving the customer experience can lead to increase sales, improving employee satisfaction can reduce staff turnover and increase productivity. Both examples deliver a financial benefit to the organisation and (like all the other benefits) are good reasons to automate beyond just cost reduction.


I’ll end this blog with highlighting again the need for intelligent process automation to be considered as an innovative transformational activity and not just about automating existing as-is tasks. The former and not the latter is where large scale, maintainable business benefits can really be maximised.


We regularly publish blogs (like this one), articles and news about AI, Automation, Analytics and related technologies. If you subscribe to our newsletter below or follow us on LinkedIn you’ll get notified when these are published

If you wish to understand more about any of the aspects covered by this blog, want to provide feedback or are interested in how we can help you avoid some of these pitfalls then please contact us via


About the author

This article was written by Dr Steve Sheppard, CEO of Combined Intelligence. Steve has many years of experience in leading the construction and delivery of transformational digital, automation and AI solutions. Steve is happy to connect via LinkedIn with others working with or just interested in intelligent technologies.