Payroll Part 2: Present and possible futures, a 2024 update.

A rework of last year's payroll sermon.

This is a rework of a post I did about a year ago on payroll, I’ve added a bit more about AI, but the biggest chunk of the post held up well, so that stays.

It builds on the post from a couple of weeks ago about payroll history.

Warning, this is rather long. I thought about chopping it up, but didn't.

First up, a bit of innovation theory

The s-curve is a well established theory to illustrate technological and related innovation and disruption. Foster came up with the idea Christensen, Geroski and others expanded on it  Christensen 1992a 1992b  Geroski (1990). Geroksi's book on new markets is a joy to read. He could have written great lyrical literature, but he wrote profoundly about innovation instead. Literature's loss, business' gain.

(chart from Geroski)

I find myself thinking about s-curves and forms of innovation a lot. I blame Marc Ventresca. Turns out that moving a business from one curve to another is hard.

Managing a business that straddles multiple S-curves is really very hard. (An example of another industry challenged with this today is the automotive industry. Consider hybrid cars to be a stretching of the combustion engine s-curve).

Let’s apply s-curves to payroll for a bit

End-users are generally loathe to switch out payrolls, as time, cost and risk in replacing are high. 

This impacts both in-house payroll providers and payroll service providers.  Vendors are also loath to experiment, given the low risk tolerance of users, and the high feature complete efforts. 

Many payroll organisations and payroll vendors are running pre-cloud technologies, even if other elements of HR technology have long since switched to the SaaS / Cloud s-curve. These older payroll technologies are in stage 5 of the Geroski model, on the on-prem s-curve.

By moving existing software (with some modifications) to a cloud hosting model and perhaps refreshing the UX layer, several vendors and service providers have avoided short term obsolescence. This has the advantage of leveraging existing functional completeness, and may give off the impression of significant innovation, but the  innovation is merely incremental. This is called stretching the s-curve.  

The longer a vendor stays on the on-prem curve, the harder it will be to innovate, and jump to the cloud s-curve. At first there is a performance deficit, but over time the new technology (and accompanying business model innovation) means that the new curve wins out.

The shift to native cloud is fundamentally more disruptive for both vendors and users, but it provides the opportunity for new forms of business model and service innovation.  Native cloud isn’t the only technology innovation, but it is the prerequisite for the bundle of innovations that change how payroll is done. I consider native cloud to be a new s-curve. The innovations that cloud native enables are of a technological, market and organisational nature, and we are still relatively early along this curve for payroll, probably at prime unfolding. Adoption has been slow, because of the functional completeness demands, and risk profile of the consumer.  Most other elements of HRTECH are now much further along the curve. 

Building out a fundamentally new payroll is dependent on both technology innovation and a path to functional completeness. The quicker functional completeness is achieved, the faster the solution will be adopted.  Payroll functional completeness atrophies over time, as new laws force changes to the product. At some point the underlying technology becomes incapable of adopting efficiently to the functional requirements. This could be because there are no longer engineers that know the programming language,  or perhaps that the functional or compliance requirement exceeds the technical capabilities (for instance real time APIs). 

I’ve noted some of the areas where I’m seeing and I expect innovation and disruption below. The technology risk of cloud payroll has dropped dramatically, and we are now seeing an acceleration of functionality delivery.  Established native cloud payrolls are at point where they out-perform older products on many dimensions. The performance gap is largely closed. This is when the serious disruption starts. Plot your vendor on this chart.

(update) AI adds complexity and opportunity

A question I’ve grappled with over the last few months is does AI create a new S-Curve for payroll or not. I’m now thinking it does. Building a native cloud payroll without deep AI capabilities is not likely to be competitive. I also think retrofitting AI to ancient on-prem payrolls will stretch that S-Curve, but ultimately fail.

AI offers tantalising opportunities to rethink how to build, deploy and operate payroll technologies, I’m thinking it is at least as disruptive as the move to the cloud is.

Payroll has more patterns than a tartan factory, and AI’s ability for pattern recognition will mean all sorts of interesting opportunities. But hallucinating payroll isn’t cool, so it going to require thoughtful application. Over time AI competence is going to be a bigger determinant of payroll vendor success than cloud competence.

Cloud and AI become, in Geroski-speak, a bundle.

You have earned a break. Have a cup of tea now that you are through the theory bit.

The s-curve shift implications

Figuring out the tax

Payroll still has the primary function of accurately calculating tax obligations, and tax has become more complex since the days of the Lyons Electronic Office (LEO). Analysing, assessing and transposing law into code has been very challenging, requiring deep country specific payroll taxation expertise. This is one of the reasons why very few vendors have been able to build payrolls to cope with multiple countries. Building a reliable gross to net engine that can cope with multiple country rules remains a daunting challenge, and the ongoing effort to maintain and update the payroll requires significant discipline and investment.

Using software to interpret regulations has been a topic of academic research for decades. Massive improvements in AI, machine learning and especially natural language processing mean that turning regulations into machine readable schematics is within our grasp. The research in academia is promising. I expect widespread adoption of these technologies to interpret tax rules, and identify legal changes. While this will not replace the need for strong local payroll management skills, that role will change dramatically, to focus on exceptions, regulatory liaison and rollout.  Over the last year or so, there has been great progress using Gen AI to do this.

This will open up all sorts of new payroll service models, and will lower the barrier to entry for vendors with strong HR and generic payroll capabilities to enter new markets. 

ML will also make it easier to migrate from one vendor to another, both in terms of structuring the employee data, but more interestingly in migrating rules from one vendor to another. 

Sometimes a regulatory change will create significant market disruption. For instance, the changes in tax-withholding in France created an opportunity for Payfit to disrupt small company payroll, and grow remarkably quickly.  The changes in time and attendance recording across Europe will spur time and attendance and payroll investment. Similarly, changes in employment law in Mexico will drive SME payroll adoption, one of the reasons we invested in Worky. Several sub-saharan African countries have implemented or are planning to implement e-filing, which we believe will lead to explosive growth of payroll tech in Africa, which is why we invested in Workpay. Similarly with Brio in S.E.Asia, and Palm.hr in MENA.

Low code/no code/rules engines

If all that payroll did was calculate the tax that governments demand, every company could run the same standard process. But almost every company has rules that are either specific to itself, a union agreement, or to an industry. The complexity is not always in the gross to net, but in actually assembling the gross pay components coherently, and then communicating with various parties that require the output from the calculations, especially governments.  This is why the ITSG process in Germany is significant.

Most payroll tools have always had some form of scripting language that enables customers or consultants to set up, modify and maintain payroll rules. In one sense, these are a form of low code tooling, but historically almost all of them have had steep learning curves and many are decades old. The massive improvement in usability and flexibility of modern rules engines provides an opportunity to empower business users to build payroll rules. We are not yet at the point of full natural language processing for payroll rules, but it is on the horizon. I’m impressed with what Datascaler HR is up to in this regard.

Many governments have embraced automated filing processes, not just for tax, but for healthcare and pensions too. Some governments now regularly modernise their side of the integration. Over time, as more countries do this, payroll reporting will become easier and considerably faster. It will also enable governments to invent new forms of taxation and payroll reporting. Vendors will be required to innovate their integration technologies to perform government reporting, and some vendors will find this harder than others, especially those on-prem s-curve.

Easier to run

While making payroll completely automated is unlikely to happen in my lifetime, the tools for payroll managers to predict and resolve payroll issues are improving dramatically. Exception predictions, data quality assessments are becoming better, in part due to advances in ML. ML should help in picking up issues such as over and underpayments, or even payroll fraud. Cloud and AI enables the software vendor to do a far deeper analysis of what is working and what isn’t working in the product, real time. In theory, this should dramatically increase service and support quality. 

Vendors are also able to deliver legal changes far more effectively and efficiently. this is really big deal in terms of operating costs and service quality. the process of getting the legal change into the end users payroll has been almost trivial compared with traditional tenancy models.

New ways of work demand new forms of payroll

Remote work, flexible work, four day weeks, gigs, multiple jobs, employee v contractor. Work is taking on new forms, and the classic five day/eight hour model is looking less and less like the default. Legal models of work are changing too, witness the rather messy IR35 saga in the UK. How and where work is done is changing, and this is changing how payroll will be done. 

The last few years have seen an explosion of pay on demand services. The need for payroll flexibility will grow, be it pay frequency, geographic location, or distribution method. Companies whose payroll lacks the capabilities to deliver flexibility will be disadvantaged in workforce attraction and retention. Organisations that have batch only payrolls will need to modernise, especially in industries where flex pay will be in high demand, such as retail and hospitality.  Traditional BPO players will also be under pressure to modernise their service offerings and ultimately, the underlying technologies. Remote work vendors (global Employer of Record EOR) may well eat into the traditional BPO market, especially for smaller, fast growing, multi-country businesses.  Global EOR has come of age. See the rapid growth of companies such as OysterHR, Deel and Workmotion, and the transformation of SD Worx.

Better APIs and aggregated data are creating a new category of companies: Fintech has landed in payroll

No sooner is the exploitation of the labourer by the manufacturer, so far, at an end, that he receives his wages in cash, than he is set upon by the other portions of the bourgeoisie, the landlord, the shopkeeper, the pawnbroker, etc. Karl Marx CM.

If Marx lived today, he would probably replace pawnbroker with databroker. Being able to tap directly into the pay history of workers is very valuable for those seeking to sell to the worker, whether selling products, or perhaps assessing the creditworthiness of the worker.. 

Over the last few years we have seen a massive growth in services that tap into the payroll data of the employee, in order to provide some sort of value add service for the individual. The development of “early wage access” solutions are, at one level, a welcome mechanism to reduce the need for extortionate short term loans. Their very existence though, does raise questions about pay rates for many jobs. 

The ability to simply and safely connect systems together has helped to spawn new financial services businesses. Regulatory changes have helped too, see for instance PSD2. Traditional retail banks have been forced to open up access. This direct access is coming to the payroll world. Rather than paying the salary into a current or chequing account, and then disbursing it, new solutions enable service providers to directly tap into the pay at source. Since the invention of the current account, traditional banks have been the main distributor of the employee’s earnings to investments, pensions, debt payments and so on. The opening up of the payroll threatens to massively disrupt this.  

This places new demands on payroll in terms of openness, security and data protection/privacy.

“Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness. Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery” Charles Dickens, David Copperfield.

It is in the interest of employers that employees are less stressed about pensions and financial wellness. Vendors are now offering solutions to employers that purport to help employees better manage their outgoings and prepare for retirement etc. There has been a strong flow of VC funding to financial well-being services, and this is likely to continue. Some of the use cases are very impressive. For instance, I’ve seen a service for migrant workers to distribute earnings cross-border.  Business models will change too. Free payroll is on cards (or your card).

Better service for employees 

The payslip (paycheck) has remained relatively unchanged since LEO. For all the talk of employee engagement, for a lot of us, the payslip is the most important and meaningful communications opportunity, yet it is poorly leveraged. The opportunity to make the payslip more intelligent and engaging is  long overdue attention.

Helping employees understand payroll output, for instance tax, overtime etc, through a conversational experience is now technically feasible. Employees should be able to do their own real-time simulations, answering questions like how much overtime will I need to do over the next three months to increase my net take home pay by $1400?

Will this be done by the existing payroll providers, or will it be done by new fintech/payroll players? 

Data quality and better processes, even beyond HR

The first payroll run at Lyons had one error, it was due to some missing time and attendance data. Unlike a lot of organisational data, payroll data is almost always accurate. Integrating payroll data into other HR processes improves the data quality of those other processes. It is one of the most powerful reasons to have a suite. HR and payroll working closely together keeps the HR data up to date. 

When most organisations do financial planning, they rely, at best, on average payroll costs for budget and planning simulation. Modern, real time payroll would enable a much more accurate approach by bringing actual payroll simulation data into the planning process. This would be so much better for determining cash flow demands, planning for forex exposure, leave liability, and social insurance obligations. Better APIs would facilitate this too.  I expect the large enterprise suite vendors to do a better job of articulating and delivering the benefits of integrated payroll, finance and HR.

Moving data and money

Payroll involves moving data and money though multiple parties. Improvements in currency transfer mechanisms, bank transfers, data integration, and workflow are changing the nature of international payroll. For multinational organisations it is so much easier to manage global payroll today than it was a decade ago. Organisations have more choice now, with better payroll aggregation vendors offering, remote work solutions and better local payroll solutions.  Meaningful interest rates mean that managing payroll float becomes cool again.

SME has led the way so far, but likely to remain national and regional.

New national champions are rapidly emerging to serve SME companies. Payfit has grown rapidly in France, Pento in the UK, Gusto and Rippling in the US, Worky has launched in Mexico, Brio in Malaysia. SME has also led the way with innovative sales models, such as product led growth.

After a long gestation period, Personio has finally got its German payroll approved by the ITSG. I’m watching this one closely. If they have got this right, it will significantly disrupt the DATEV Hegemony in Germany. More on ITSG and German payroll here.

HiBoB has acquired Pento in the UK. Deel acquired Payspace (African payroll play) recently. Anita notes some of the recent deals here. I’m expecting to see more acquisitions over the next few months.

The European large enterprise market will become more competitive

Large employers that operate in multiple countries have largely moved or plan to move their core HR processing to the cloud native s-curve. 

Workday was the first mover with cloud native enterprise core HR, initially successful in North America, but now very competitive in Europe. Its US payroll has a high attach rate. It has announced German payroll, and I expect many Workday German customers will adopt this over time as it matures. Workday’s UK payroll has an increasing attach rate, but I would not yet call it a success.

Oracle has also strengthened its payroll offering, recently launching in France. SAP extended the s-curve of the on-premise payroll through ECP,  but at the same time it has been working on a next-generation cloud payroll solution, now in testing in the UK. Ceridian continues to strengthen its international capabilities. The next few years will see many large organizations rethink their payroll strategies.  This will be in part a push of obsolescence, but increasingly it will be a pull from products offering lower long term risk, lower operating costs, more flexibility, and empowering new ways of paying people. 

Traditional outsourcers: consolidation and innovation

The shift to the new s-curve means that outsourcers that have built their businesses offering services on top of the previous s-curve technologies will become less competitive. I expect considerable market consolidation, with PE backed funding. At the same time, these outsourcers will seek new engine providers, or even develop their own. Payroll outsourcing will remain attractive, but it will need to adapt to new customer demands. Outsourcers will drive more partnership with broader HCM services, offering more end-to end services.

Update: Last year we saw UKG acquire Immedis (I was on the board of Immedis).

Crypto / distributed ledger

To date, almost all crypto applications in the HR context are solutions looking for a problem. They use block chain to solve something that can be adequately solved without it.  Paying employees in crypto has a novelty element, and but we have not seen employees or employers seeking to set salaries in crypto given its volatility.  Over time we expect distributed ledger technologies to play a role in payroll and related financial processes. For instance, we have invested in start-up looking to disrupt how stock options are managed.- Futurz. For more on web3, make sure you read Anita Lettink. 

Some parting thoughts

Risk avoidance versus the cost of doing nothing

Few applications have the longevity of a payroll. Many companies and outsourcers run payrolls that are 30 years old or older. While stable, eventually the age of these solutions itself becomes a risk, in that the underlying technologies are no longer supported, fewer and fewer engineers are prepared to maintain these solutions, and finally, adapting them to changing legislation becomes very expensive.

I expect to see a gradual but accelerating replacement of older payroll engines, with cloud native AI enriched payroll engines. These engines easily support real time, constant payroll calculation, are API first, make effective use of AI for error handling, and can cope with multi-country legal requirements.  The leading international payroll vendors either have, or are close to releasing these sort of solutions. Establishing stable and rich localizations does take time, but the wave of conversions has begun. This engine change will impact both in-house and outsourced payroll provision. this has been brewing for a while. 

The battle for payroll services has only just begun. New business models and new services will change how employees are paid. The line between financial services and payroll will blur. Some of these solutions will benefit the employee, others may not.  One thing is clear: the value of payroll data is massive, and it will not remain buried for much longer.

There will be more change in how payroll is processed in the next 10 years than there has been in the last 30.  If you are doing something cool in payroll, let's talk.

If you are a founder applying AI in a novel way to payroll, or in the upstream or downstream processes around payroll, stop what you are doing and call me or Jason.