PEO Risk Management & Insurance Issues

by Frank Huang | November 23, 2021

Originally published in PEO Insider (November 2021). Reproduced with permission of the National Association of Professional Employer Organizations.

At the end of 2020, management consulting firm McKinsey & Company surveyed 899 C-suite executives about the impacts of the pandemic on business operations. One of their conclusions was that the pandemic had accelerated the adoption of digital processes by three to seven years,[1] with B2B firms like those in the PEO space on the higher end of the spectrum.

I too have observed an uptick in initiatives by PEOs, with a particular focus in the area of risk evaluation for both workers’ compensation (WC) and health benefits (HB). This trend is likely driven by the desire of firms to grow their client bases while improving speed to market and confidence in expected net revenue. While there is a wide variety of potential risk evaluation models, I’d like to touch briefly on the present state, two commonly adopted approaches, and one that I affectionately call the “Holy Grail” of models. For all but the last model, I will primarily speak through a WC lens, although the concepts can be applied to other coverages.


The present approach to evaluating risk in the PEO industry is too often one of least resistance and/or legacy philosophies. In my experience, this includes approaches as simplistic as beating the price of the incumbent or incorporating claims data in a way that may have been initiated years ago by someone who is no longer with the firm (e.g., the proverbial “black box” that no one understands). While the former appears the riskiest and depends on the accuracy and relevance of the incumbent’s price, the latter may actually introduce more risk by suggesting an accuracy or rigor that is not actually there. For example, it is not uncommon to find outdated assumptions or incomplete model designs when reviewing a PEO’s current model.


One of the models currently being adopted by PEOs has actually been used by many insurance companies for decades. This model uses traditional actuarial methods to estimate ultimate losses based on the mix of historical payroll, but also actual historical claims experience when credible. The model considers a variety of factors when determining how much weight to give to historical experience, including, but not limited to, the size of the prospect and how many years of experience have been provided.


Where the actuarial-based model is arguably more intuitive but slower to produce a decision and/or risk assessment, a model driven by predictive analytics and artificial intelligence (AI) is less intuitive and quicker to produce a result. This is because such models do not need all the historical data required of an actuarial model and instead use a database of historical outcomes to predict a risk level. For example, where the traditional actuarial model may have required five years of loss runs and payroll reports in state and class code detail, a predictive model may only require more summarized attributes of the prospect to then draw upon the database of outcomes.


In my experience, many PEOs evaluate the risk of prospects and existing clients by looking at each risk area independently. For example, a prospect may appear to have low WC risk but high HB risk such that the PEO chooses not to write the account, even if the potential WC benefit offsets the HB risk.

The next generation of risk evaluation models applies the predictive/AI-driven approach from the prior section to all risk sources simultaneously, thus considering the collective risk of the prospect. From this vantage point, PEOs can potentially improve speed to market and better match their risk appetite with that of a prospect. My understanding is that very few PEOs are currently using such an approach, with the largest PEOs most likely to be closest to its adoption. Such an approach may become commonplace within the next three to five years.


While it’s simplistic to say that smaller firms are at the lower end of the spectrum of sophistication while larger firms are experts in predictive analytics, the reality is much less black and white. I’ve seen some of the largest firms use relatively rudimentary approaches and smaller firms be huge proponents and users of predictive analytics/AI.

What is becoming clearer is that those in the PEO industry have been nudged into greater adoption of analytics thanks to the COVID pandemic. Firms that adopt early will reap the benefits, with “fast followers” hoping for similar outcomes and minimal negative adverse selection.


Not surprisingly, there are a number of consulting and analytics/AI firms providing their services to those in the PEO industry, each touting a custom-tailored model specific to your firm’s needs. Take the time to vet out their claims by asking straight-forward questions. For example:

  • How well do they understand the PEO model?
  • What will their model provide?
  • How did they obtain the data that is being used to train their models?
  • How similar is that dataset to your intended use?
  • What type of predictive/AI model is being used?

On this last note, many firms will be reluctant to share such proprietary details, but it is helpful to know as much as possible what engine is under the hood of the car you are buying: a 4.0 L V8 or a 1.2 L 3-cylinder. 


Frank Huang has more than 15 years of actuarial consulting experience serving a wide range of clients, including serving as ADP’s Chief Actuary.  Learn more about our PEO consulting practice here.