a very key point on the decision velocity constraint. I spend my days right now more concerned about the leadership required to create the culture and systems to maximize the opportunity that emerges when you have "a country of geniuses in a data center". Building the technology to make those businesses thrive (both people and systems) will be necessary but without the leadership to forge the path to this future we must wait for "best practices" to emerge so that the cult of followership can happen. What Dave and Aneel had was a very clear understanding of the gap between the current (PSFT) and the future (Cloud) and a single clear constraint (the HRIT teams struggled to keep PSFT current and manage the operations - cloud allowed WDAY to remove that constraint). Today's new constraints and bottlenecks are shifting every week. I agree the a16z list is "fine" but not transformative. Also double vote up the snark... it speaks to me as you know.
Good fun and relevant to read your thoughts - or at least the part of your thoughts you accepted to share. Reading the piece from A16z yesterday I had some similar concerns. I've seen elsewhere "I am going to rebuild my HRIS in 24 hours with Vibe" - and either the operational part gets automated while keeping intact processes, data, payroll and compliance, or I may start getting popcorn to enjoy the show. I guess the reason I keep in this business is to see how it will evolve.
A newer stack does not matter much if HR still works the same way.
The bigger question is what becomes possible when policies, approvals, payroll impact, compliance obligations, and evidence are understood together.
That may not start by replacing Workday. It may start around it, with a layer that helps people and agents know what is allowed, what happens next, and what proof exists later.
Workday can stay as the system of record. But the real work may move to the layer that governs the context.
Aaron, yes, you are directionally right. But the question for the context governance layer is the api into the ERP. Is it up for the task, and how is it tolled?
Thomas, yes. I think the hard part is not really who owns the data.
Customers own the records, policies, decisions, and evidence. Vendors own the software, data model, and logic that turn those inputs into something useful. The valuable part sits in between.
Whether the API is up to the task really depends on the vendor. Some are improving. A lot are still too narrow or slow when customers need to check rules, approvals, and payroll impact in near real time.
Tolling is the bigger question for me. Vendors are right to protect what they built. But customers still need practical access to govern across systems, especially for compliance, audit, and agents that pull evidence from more than one place.
So it becomes an economics question. What is fair pricing, and when does it start to feel like a tax on the customer just trying to run their own business.
Agentic start ups seem to assume that access should be free for them, but they can monetise usage themselves, this is where the tension will emerge. There are also going to be interesting challenges on liability. When Agent overwrites half the ERP, who is to blame.... Interesting times indeed.
As someone who has worked a various large companies, a situation like you mention with Workday, where they should be doing their dammed best to thoughtfully and rigorously reimagine their product, rarely happens because they feel the pain of change so much later than the small guys. Transformation often arises out of disruption, and Workday might not be aware how much their world is already gone.
I used to implement a lot of SAP HCM when it was part of R/3 And ECC - it used to buddy up with Cornerstone the perfect mix of operational HCM including payroll for complex operational business’s and it still exists today as S/4 HCM, sap don’t trumpet they have a second cloud HCM solution with loads of AI.
That’s the point here SAP HCM went into decline when they purchased SF, and when they announced the ECC HCM migration path was to SF many customers went Workday instead.
Is it sexy or funky anymore maybe not but it delivers well,and its core users are highly conservative when it comes to employee data so fanciful think that major enterprise customers will consider composable agentic.
I think Workday are doing the right thing, watch and see and add bits as required.
Reinvention in this space comes from the edge products that enable workforce mgt, enable contingent worker etc
Thank you for taking the time to dismantle this A16Z piece Thomas.
I think Joe is missing the point because he's not examining why ERPs exist in the first place.
ERPs try to encode a canonical model of a company because pre-AI compute couldn't reason about meaning. Schemas, business objects, configurable workflows are all scaffolding to codify meaning into machines that had none.
AI changes the basis of meaning. An LLM carries a compressed model of how human institutions work. They have a deep knowledge of why an invoice exists, what human problem it solves and what it should contain to fulfill its purpose. Same with how to onboard an employee or cascading goals. Etc..
You don't have to encode the model in advance. You feed it context and it renders the model on demand, fitted to the case.
So the encoded ERP models become optional and probably redundant. (Same with UIs and workflows). Governance, compliance enforcement, storage and security are what is left.
Implementations are bottlenecks because they reflect a client's willingness and ability to deform themselves into canonical models. That goes away naturally when canonicals disappear.
When reading the Workday article, I found myself frowning from time to time. They make it seem easier than it is, and you point that out elegantly.
I'd love a good foundation layer for people data and the ability to easily connect stuff to it, but it's not that easy, and there will be a gazillion requests from customers around "can we add this, can we do this" even if it's clearly sold as merely a data lake of some sort.
Refreshing take on the future PeopleSoft, Thomas. So much value in funds and tech leaders that deeply know a vertical from the inside-out and outside-in.
When I read all the startling things that are going to happen in HR, I remember what I said to a meeting once, that '"HR is a very large convoy travelling at the speed of the slowest ship". In 10 years time, parts of HR will be using legacy systems, so vendors can still make money there.
The next Workday will probably be composed of 8 or so products we,'ve barely heard of linked by advanced iPaaS.
I've found some already...give me a bit more time!
a very key point on the decision velocity constraint. I spend my days right now more concerned about the leadership required to create the culture and systems to maximize the opportunity that emerges when you have "a country of geniuses in a data center". Building the technology to make those businesses thrive (both people and systems) will be necessary but without the leadership to forge the path to this future we must wait for "best practices" to emerge so that the cult of followership can happen. What Dave and Aneel had was a very clear understanding of the gap between the current (PSFT) and the future (Cloud) and a single clear constraint (the HRIT teams struggled to keep PSFT current and manage the operations - cloud allowed WDAY to remove that constraint). Today's new constraints and bottlenecks are shifting every week. I agree the a16z list is "fine" but not transformative. Also double vote up the snark... it speaks to me as you know.
If there’s 1 thing I’ll definitely spend my free retirement time on, then it’ll be reading the fruits of your keyboard, dear Thomas.
Rudi, as long as you keep reading, I’ll keep writing.
Good fun and relevant to read your thoughts - or at least the part of your thoughts you accepted to share. Reading the piece from A16z yesterday I had some similar concerns. I've seen elsewhere "I am going to rebuild my HRIS in 24 hours with Vibe" - and either the operational part gets automated while keeping intact processes, data, payroll and compliance, or I may start getting popcorn to enjoy the show. I guess the reason I keep in this business is to see how it will evolve.
A newer stack does not matter much if HR still works the same way.
The bigger question is what becomes possible when policies, approvals, payroll impact, compliance obligations, and evidence are understood together.
That may not start by replacing Workday. It may start around it, with a layer that helps people and agents know what is allowed, what happens next, and what proof exists later.
Workday can stay as the system of record. But the real work may move to the layer that governs the context.
Aaron, yes, you are directionally right. But the question for the context governance layer is the api into the ERP. Is it up for the task, and how is it tolled?
Thomas, yes. I think the hard part is not really who owns the data.
Customers own the records, policies, decisions, and evidence. Vendors own the software, data model, and logic that turn those inputs into something useful. The valuable part sits in between.
Whether the API is up to the task really depends on the vendor. Some are improving. A lot are still too narrow or slow when customers need to check rules, approvals, and payroll impact in near real time.
Tolling is the bigger question for me. Vendors are right to protect what they built. But customers still need practical access to govern across systems, especially for compliance, audit, and agents that pull evidence from more than one place.
So it becomes an economics question. What is fair pricing, and when does it start to feel like a tax on the customer just trying to run their own business.
Keen to hear your view on this.
Agentic start ups seem to assume that access should be free for them, but they can monetise usage themselves, this is where the tension will emerge. There are also going to be interesting challenges on liability. When Agent overwrites half the ERP, who is to blame.... Interesting times indeed.
As someone who has worked a various large companies, a situation like you mention with Workday, where they should be doing their dammed best to thoughtfully and rigorously reimagine their product, rarely happens because they feel the pain of change so much later than the small guys. Transformation often arises out of disruption, and Workday might not be aware how much their world is already gone.
Brilliant "The only people asking for an AI native version of this are people who have a substack and an early soho house membership"
Absolutely love this post.
I used to implement a lot of SAP HCM when it was part of R/3 And ECC - it used to buddy up with Cornerstone the perfect mix of operational HCM including payroll for complex operational business’s and it still exists today as S/4 HCM, sap don’t trumpet they have a second cloud HCM solution with loads of AI.
That’s the point here SAP HCM went into decline when they purchased SF, and when they announced the ECC HCM migration path was to SF many customers went Workday instead.
Is it sexy or funky anymore maybe not but it delivers well,and its core users are highly conservative when it comes to employee data so fanciful think that major enterprise customers will consider composable agentic.
I think Workday are doing the right thing, watch and see and add bits as required.
Reinvention in this space comes from the edge products that enable workforce mgt, enable contingent worker etc
Not gimmicks
Great post
Thank you for taking the time to dismantle this A16Z piece Thomas.
I think Joe is missing the point because he's not examining why ERPs exist in the first place.
ERPs try to encode a canonical model of a company because pre-AI compute couldn't reason about meaning. Schemas, business objects, configurable workflows are all scaffolding to codify meaning into machines that had none.
AI changes the basis of meaning. An LLM carries a compressed model of how human institutions work. They have a deep knowledge of why an invoice exists, what human problem it solves and what it should contain to fulfill its purpose. Same with how to onboard an employee or cascading goals. Etc..
You don't have to encode the model in advance. You feed it context and it renders the model on demand, fitted to the case.
So the encoded ERP models become optional and probably redundant. (Same with UIs and workflows). Governance, compliance enforcement, storage and security are what is left.
Implementations are bottlenecks because they reflect a client's willingness and ability to deform themselves into canonical models. That goes away naturally when canonicals disappear.
When reading the Workday article, I found myself frowning from time to time. They make it seem easier than it is, and you point that out elegantly.
I'd love a good foundation layer for people data and the ability to easily connect stuff to it, but it's not that easy, and there will be a gazillion requests from customers around "can we add this, can we do this" even if it's clearly sold as merely a data lake of some sort.
Refreshing take on the future PeopleSoft, Thomas. So much value in funds and tech leaders that deeply know a vertical from the inside-out and outside-in.
When I read all the startling things that are going to happen in HR, I remember what I said to a meeting once, that '"HR is a very large convoy travelling at the speed of the slowest ship". In 10 years time, parts of HR will be using legacy systems, so vendors can still make money there.
The next Workday will probably be composed of 8 or so products we,'ve barely heard of linked by advanced iPaaS.
I've found some already...give me a bit more time!
Like Rudi, I'm going to enjoy these themes!