Skip to main content
main content, press tab to continue

Automation is expanding modeling frontiers

By Chris Bird and Joe Milicia | September 22, 2021

Automation of capital models isn’t just about doing more in less time; it’s a way of doing more that delivers genuine value to an insurance business, more efficiently.
Insurance Consulting and Technology
Insurer Solutions

If we think about the basic function and value of an insurance company’s capital model, it’s to provide a set of probability measures that help the company understand the risk in the business and potentially make adjustments in areas such as business mix or reinsurance purchase—measures which draw on and require input from the whole business such as the probability of falling below plan, being able to pay dividends, or the risk of a rating agency downgrade.

Despite rising levels of automation within the insurance industry , capital modeling has typically remained a time and people-intensive job. The amount of effort required in producing these risk metrics has dictated for many insurers that capital modeling is only carried out on an annual cycle.

But that’s becoming increasingly less the case. Automation is enabling insurers to do more useful things, more often, and capital modeling is ripe to automate for insurers of all sizes – enabling insurers to capture insights previously out of reach, derived from more frequent model runs or more sophisticated multi-year projections that reflect the potential capital impact of future development plans.

Many moving parts

A big reason capital models are often run infrequently is the number of moving parts involved. In Figure 1 below, the actual model run is the single cell in the third column of the diagram. As you can see, lots must happen first.

The starting point is all the data that’s going to be needed for the model. As any capital modeling actuary will testify, this will frequently come from different parts of the organization, at different times, in different formats, and often cover different intervals. So, it can be a major exercise to check the right files have been sent and checked for any errors.

Moving on to the second column, there’s a whole series of data manipulation, calibration, allocation and mapping activities that must take place to transform, adjust and rescale inputs.

And once the model has run, there is then a whole range of outputs and reports that are going to be required. Done manually, this is inevitably a slow, time and resource-intensive process.

This diagram shows the large number of data feeds that typically come into a capital model, along with the validation/processing stages required before parameters can be linked into the model.
Figure 1. Data feeds for a PC Capital Model


Let’s apply this structure to a typical modeling activity – a business plan update. Seemingly, this should be quite straightforward as it will only involve updates of three data files (those shown in purple in the first column of Figure 1). There’s rather more to it than that, as Figure 2 shows.

The diagram shows a typical process flow for updating the capital model for a new business plan. Most of this can be automated with only the purple boxes requiring human input.
Figure 2. Process flow for capital model business plan update


Drilling down, the capital modeling team will have to gather data from, most likely, the Finance and Pricing teams. If there are any problems, they may have to go back and request updated data, all of which, if done manually, will have to be rechecked and verified before being readied for processing.

The model run itself is no longer the slow part. With modern technology and software, along with availability of cloud computing capacity and virtual grids, it’s not uncommon for capital models to run in minutes. What takes time, and where errors can creep in, are the manual steps and interventions.

With modern technology and software, along with availability of cloud computing capacity and virtual grids, it’s not uncommon for capital models to run in minutes.”

Chris Bird | Global Proposition Leader, P&C Capital Modeling and ERM

Coming back to figure 2, the vast majority of this is mechanical – it can be codified and automated. Only the blue steps are those where humans would still be involved and can add value. Notably, this eliminates human involvement from the more mundane steps, allowing modelers to focus on review and quality of outputs. And, it is much quicker.

This example reflects four key capabilities that automation contributes to making capital modeling more efficient:

  • Process logic and management – automating the management of the process flow and logic – such as generating automated emails and ensuring the right people are doing reviews at the right stage
  • Automatic validation – carrying out all checks of data files. It’s already common for insurers to have some automation in this area, but there remains a lot of untapped potential
  • Data processing transformation and calibration – data manipulation and transformation; collecting all parameters in one place
  • Automation of the set up and run instructions of the model, including report production

Approach to automation

That doesn’t mean that automation of a complete end-to-end process, or a full set of sub-processes, should be a default strategy. The processes that are ideally suited to automation will typically be chunks of a larger whole that need to be considered at a micro level. But there are some distinct characteristics that will mark tasks out as ripe for automation.

  • Repetitive – The more frequently processes are run, the more value automation will bring - for example, a quarterly reserve process or a rate filing process done in multiple states.
  • Prone to error – If a task has a high level of manual intervention, where it’s very conceivable that someone could grab the wrong data or mis-align the location of where they’re pasting that data, then automation can help. A robot is not going to make mistakes like a human might.
  • Rules-based – For a process to be effectively automated it needs to have clear rules, logic and decision points, such that these commands can be understood by a robot. The more nebulous a process is, the harder it will be to automate.
  • Digital data - Digital data is obviously a pre-requisite for automation – there are some ways around this, like screen scraping and optical character recognition, but obviously the more information you have in a digital format, the more obvious the fit for automation.
  • Time - Where processes need to be done quickly, then leveraging automation helps meet challenging timeframes and frees up people to do other business critical work. Automated processes enable people to focus on other, more interesting, more value-add work because automation targets the independent, repetitive tasks that take up so much of an actuary’s time.

More than technology

Mindful of these points, companies also must consider how they approach transformative initiatives. Experience shows that the most successful projects are those that combine and give equal weight to three factors – 1) Technology; 2) Processes; 3) People.


Technology is normally the first thing companies think of when trying to automate aspects of their business. However, there is a tendency to over-focus on the technology side of things. Companies may purchase large and expensive systems thinking that software alone will magically solve all problems. More often though, the result is they get bogged down in the details of the software and neglect the process and people aspects of transformation. Alternatively, many companies may have incumbent automation systems which aren’t suitable for certain areas of their business but try and shoehorn them to fit anyway.

A “plug and play” approach is the way forward; our experience shows that it is best to select specialist support tools that are scalable for specific functions within the business, but which can integrate with each other. This is crucial to capital modeling function when there are inputs coming from reserving teams, pricing teams, management teams, finance teams, and so on.

Further, technology that comes integrated with the tools used in your process out-of-box helps to make automation more accessible to the business user and decreases dependencies on external vendors and even internal IT. The rise of technology targeted for insurers and insurance-specific technologies is enabling a democratization of automation and empowering business users to own, maintain and enhance automated processes that historically could only be altered by requesting IT support and waiting in IT queues.


Often, process engineering and automation initiatives are viewed as two separate tools that can be deployed to increase efficiency and efficacy. While benefits can be derived from each approach independently, those benefits pale in comparison to those that can be realized by automating an optimized process. Automating a broken process may save some time and effort but will not help to achieve broader business objectives. Similarly, large tech spends on data marts and new finance or administration systems have rarely delivered the promised value when the rest of the business process has not been reengineered to best take advantage of new technologies. The other side of this coin, where companies have employed process engineering to eliminate waste without bringing technology or domain expertise or have tried to redesign processes without having the right technology to enable the process, have also often not delivered on the intended promise.

Processes are just as important as technology because it is key to first understand exactly the purpose and desired outcomes of automation rather than just automating a process as-is and blindly throwing technology at the problem. When reengineering processes, spending a good amount of time up front designing how the newly transformed end-state architecture should look and feel and ensuring that it is consistent with the target operating model is time well-spent.

There’s also a cost-benefit exercise to consider in terms of how much of the process to automate now and improve the quality of that automation later, versus re-engineering first and then more comprehensively automating afterwards (i.e., the “shift and fix” versus “fix and shift” argument).

And finally, when automating a process, a question that can’t be ignored is: “What if something goes wrong?” What if a user does x or y; what if data doesn’t arrive on time; what if the software fails – automation is only as successful as its weakest point, so it’s important to take the time up front to design and consider what process should be followed when things go wrong.


People are such an important aspect of ensuring that any automated solution is successfully adopted, but it is often the most neglected aspect of the three factors. Perhaps more critically than any other consideration in a transformation is what should the end-user experience be? Neglecting this or failing to focus on how the process impacts our workforce early enough in the design is likely to lead to challenges in the future.

For change to take effect, a clear vision of what that change will need to achieve has to be established and upheld throughout all levels of the business. People need to feel empowered by the change that is happening; they need to feel like the culture around them is actively encouraging this change and that automation is at the forefront of the company’s strategy.

Don’t forget also that people’s roles will change, new teams will emerge, and other teams will cease to be needed. Being cognizant of this change, designing the structure around this change and being transparent with employees early will pay off.

A transformation program with this mindset will be far more successful than one where automation is considered a side activity, and in our experience automation done well leads to a more engaged workforce due to being able to focus on the more interesting and value-add parts of their jobs.

People need to feel empowered by the change that is happening; they need to feel like the culture around them is actively encouraging this change and that automation is at the forefront of the company’s strategy.”

Joe Milicia | Global Proposition Leader, Business Process Excellence

Implementation best practices

In addition to striking a good balance among these three aspects of the automation approach, we also recommend some practical considerations for implementation:

  • Use a phased implementation approach – This allows you to strategically target pain points in the current process and minimize disruption to business as usual. Implementing automation in smaller chunks retains the ability to stay agile in the face of inevitably changing requirements.
  • Maximize early delivery of benefits – A phased approach allows incremental and quantifiable benefits to be achieved quickly and continuously. Also, people freed up from the earlier phases can then be redeployed to delivery of the later phases, providing on-the-job training and better equipping the team to take ownership of any future developments.
  • Collaboration is key – Client and vendor teams working closely together throughout all stages of the implementation enables maximum knowledge transfer and ensures the in-house team is more likely to buy in to solutions and is equipped to own the new processes going forward.
  • Focus is necessary to overcome inertia – Clients can deliver on transformation without external help, but only if resources are dedicated to that transformation and these initiatives are top priority for those employees. Too often, our day jobs interfere with achieving our transformative initiatives and the inertia of what needs to be done now, prevents the investment required to deliver what will unlock the future success of our business.

Expanding possibilities

With many insurance companies continuing to upgrade levels of automation within their operations, the market and regulatory environment would suggest no slowdown is likely.

From a market perspective, competition and the economic environment are driving efforts and the need to reduce expense ratios. On regulation, and virtually irrespective of geography, insurers have seen huge change, particularly in the last decade. The changes have substantially increased the work involved, often with a shorter time period in which to do it. Put these together, and time is a hugely valuable commodity to insurers that puts added emphasis on efficiency and using that time, together with technology and expertise, wisely.

Beyond core capital modeling applications, developments in and understanding of what automation can deliver is working in insurers’ favor and expanding the range of applications it can support. Examples of new capabilities brought about by automation, which time limitations would otherwise have made impossible, include projecting and tracking key capital measures a year ahead each quarter and multi-year projections that can enable an insurer to assess whether capital resources are sufficient to support its mid-term business plans and where strains on risk appetite capital buffers might arise.

For the many insurers that are still just scratching the surface of automation of modeling activity, a brave new world is out there and waiting to be explored and exploited.


Global Proposition Leader, P&C Capital Modeling and ERM
Insurance Consulting and Technology
email Email

Global Proposition Leader, Business Process Excellence
Insurance Consulting and Technology

Contact us