NTT Data has released a report entitled “Why automation fails but doesn’t have to.” It focused on the London Market and lessons that NTT Data has learned from projects it has been involved in. It set out to address questions such as, what is the right way to start your automation project? What are the pitfalls?
Kim Gray is head of insurance and head of diversity and inclusion in NTT DATA in the UK. On a webinar, Gray says: “We’re going to touch on some of the things that have meant that they’re slow to take off or they’ve been stalled, or they’ve failed to meet their objectives. It’s certainly not because the technology doesn’t work.”
Automation works if you choose the right processes
Ask any programmer, and they will tell you that choosing the right processes to automate is critical. Simple processes are easy to automate and code. This is because each process has a very limited set of outcomes. By comparison, larger and more complex processes are hard to automate due to the multiplicity of potential outcomes.
That does not mean that you cannot automate a complex process. The solution is to break it down into multiple smaller processes that can be chained together. In doing so, it makes it easier to understand what the process does. Additionally, debugging the process is easier because the decision tree is easier to understand.
Matt Kearney, senior strategy director at NTT DATA, makes this point saying: “One of the first hurdles companies face on their automation journey, is deciding exactly which processes to begin with. There can be a tendency to choose processes that employees don’t like, as opposed to the processes that have the largest value for the business. This is a common pitfall.”
Which London Market insurance processes are ripe for automation?
Kearney addresses five different processes that can be readily automated to deliver benefits. Some are obvious, scale easily and are used in other industries outside of insurance. All are manual intensive processes. Yet it seems that they are not as widely automated across the insurance market as they are in other industries.
- Broker: Automatically collecting claims history from multiple sources and collating into a single view for the broker. Kearney calls out that this is time-consuming, and it makes no sense that the insurance industry is still doing this manually. Many industries, such as mobile operators, have been doing this for over a decade.
- Underwriting: This process is about adding new risks to a policy. The white paper cites flood rating by postcode. Again, this is done by many other systems that take publicly available data and use it to enhance their systems.
- Finance: This includes collating Bordereaux information into reports and dashboard. It also covers extracting and recording invoice information using OCR for multiple invoice formats. In January, Carl-Petter Udvang, Product Manager at Lowell Norge explained to Enterprise Times how it had automated OCR to improve speed and consistency.
- Claims: Chatbot technology has improved considerably in recent years. It is now being used to get the initial claims information from customers through self-service solutions. It also reduces pressure on contact centres.
- Human Resources: This is an area that is long overdue for automation. Kearney talks about its potential impact on dealing with starters and leavers. He is not alone. HR knows when people join and leave, but often the process between HR and IT is broken. As Cezanne HR discovered in its recent report, first-day impressions where employees can’t get onto systems is a problem.
Common pitfalls that lead to failure
The paper calls out five common pitfalls. Not all are obvious at first.
- Choose the right process to automate: Automation should be business-led and demonstrate clear value. It is easy to allow supporters of automation to drive the agenda without setting clear targets that must be reached. If automating a process cannot be proven to deliver benefits, move on.
- Robots aren’t always the answer: NTT Data gives the example of a specialist insurance company who wanted to automate a process. They believed automation would deliver a 34% reduction in time and effort. Re-engineering the process, not automating it, delivered an 88% benefit without automation.
- Start small and build up to an end-to-end process: Choose simple processes, not complex processes. Automate single steps and then chain those together to create a largely automated process. Often steps can be reused across other processes. It saves time, money and also delivers wider benefits.
- Don’t forget to engage the team: If employees are not convinced that bots will help them, they won’t work with them. Lowell Norge did not use OCR to replace staff. Those staff moved to other positions within the company. Assure staff that bots are not making them redundant but opening up new opportunities in the business.
- Avoid overcommitting: Set clear KPIs but be flexible. Set too high an expectation and the entire approach is set up for failure. Consider if building automation in-house makes sense. Are there external services that could be used at a lower cost? In a regulated market, this means additional time should be spent proving the security of those services.
Enterprise Times: What does this mean
Is automation the magic bullet that will make your business effective? What should you automate? Where do you start? Is automation another of those mercurial technologies where every year is going to be its year until it isn’t? Judging by the amount of time technology vendors spend talking about the benefits of automation, you’d think everyone was already deep into automation projects.
The reality for automation is that like any other technology when used correctly, it delivers benefits to the business. The problem is that too many organisations are still struggling to understand what ‘correctly’ means.
Getting automation right is not rocket science, but it does require its own process and a structured approach. Rushing in will almost certainly guarantee, if not failure, a disappointing solution. It does not mean spending a large amount of time repeating proof of concepts just to be sure. At some point, the pin has to be pulled and an automation project allowed to run.
Too many reports gloss over the operational challenges to getting automation right. This one gives a more realistic view of what can be achieved if automation is thought out correctly. It also leaves open the question as to why the London Market for insurance still struggling with things like automating the collation of data.