Making Better Enterprise Software with Vcola
More on our latest investment.
I’ll pause the Acadian Summit posts, to bring you some deal news. Do check the press release, coverage and our deal announcement on the Acadian Substack.
In my polite critiques of A16Z posts on SAP and Workday, I argued that they underestimate the short term defensibility of the incumbents (the straits of Hormuz metaphor). I also noted that A16Z weren’t able to coherently articulate what could potentially replace them, given the stack fallacy challenge for the frontier models. It is a bit of paradox.
Our first investment from Fund III is our bet on what shapes the next generation ERP (or whatever it may become) enterprise stack. It is called Vcola.
Better Enterprise Software in the early era of AI generated code.
How do we build significantly better enterprise software than we did in the past, or we do it today? We can build software far faster than we used to do, but that is only part of the answer.
First, what is better software? I’m going to steal two concepts from statistics for a moment.
The first one, validity. validity is the degree to which a test or measure actually measures what it claims to measure and that the inferences drawn from its scores are appropriate, meaningful, and supported by evidence
If we apply this to enterprise software, validity is building the right thing. Have we taken a set of requirements, genuinely understood them, and built something that exactly meets those requirements (I realise there is a whole heap of requirements research in computer science, but bear with me).
The second one, reliability is the degree to which a measure produces consistent, stable, and reproducible scores - free from random measurement error.
If we apply this to enterprise software, does the product give the same answer to the same question every time? Does it work in production, not just in a demo? If we alter a part of the software, can we be sure other bits don’t break?
Today’s vibe coding tools provide us with massive velocity improvements. We are able to write far more code than we could have imagined even just a couple of years ago. But more code in the enterprise doesn’t necessarily mean better code, or better outcomes, but it can do.
For the last 40 years or so, enterprise software has been about accepting a standard “best practice”, and awkwardly modifying it to suit your own needs, often with the help of consultants. The alternative of building it yourself, once you really looked at what’s needed, was simply too complex. Vibe coding has created the tantalising opportunity to rethink this compromise, but in and of itself it is not yet capable of the providing genuine progress in either validity or reliability.
I love going back into the history of enterprise software, as regular readers will know. At one level not much has changed since Major Thompson, Mary Blood and others built the first business software in 1951 at Lyon’s Tea Room. Here’s a comment from Thompson.
In our new brave world of AI-assisted coding, what we have not yet done is created the right mechanisms to build a far more robust specification. We have also not yet created the mechanisms to take that specification and build the same application, every time. And in we are still waiting for the run-time environment that ensures graceful and secure upgrades.
Over several generations, companies like SAP, PeopleSoft, Oracle and Workday built proprietary frameworks, platforms, tools and languages that abstracted applications developers away from reinventing role based authorizations, rules engines, effective dates, audit, country level configurations, performance tuning, database choice, consistent UX controls and more. These frameworks enabled these companies to build the enterprise software that runs the world today.
Some them rather painfully transitioned these tools from on-premise to the cloud, but we don’t think any of them can morph what they have today to truly leverage the latest advances in AI programming or the emergence of agents as first class users. The dissonance is too great, the chassis is compromised.
Imagine building enterprise software leveraging an enterprise specific ontology, with object models that have learnt from the successes and failures of previous generations of enterprise software, but then learns from every new deployment, and your specific needs Imagine a specification so precise that it would produce the same product every time you asked it to. Imagine a system designed with agents as first class contributors to processes. Imagine a runtime version with rock solid performance, foundation model independence and graceful upgrades.
Imagine what you could build.
A dangerous meander into a philosophical metaphor.
I tried to read a bit of Wittgenstein at university 35 years ago, I pretended to understand it, but his ladder concept stuck with me. I recently had a chat with Claude about it, and I think I can use it as a metaphor in the enterprise software world. This will probably irk a few philosophers, vibe-coders and incumbent vendors, and bore a lot of others.
“climb the ladder, then kick it away” Tractatus 6.54.
Having used logical propositions to show what cannot be said, Wittgenstein instructs the reader to kick the ladder away once they've climbed the wall - the scaffold of preliminary concepts becomes not just unnecessary but actively misleading once genuine understanding is reached. The propositions themselves are unsinnig (nonsense) in the strict sense; their value was purely transitional. (nothing like squeezing in a German word every now and then, thanks Claude).
I imagine the progress of enterprise software to be a multi-storey wall, requiring multiple ladders to scale it. But to scale the next storey, one needs to leave the previous ladder behind.
At one level Workday has been very successful, in part because the founders’ exit from PeopleSoft gave them the opportunity to leverage what they had learnt and created at PeopleSoft, but freed them from the obligation to defend and maintain it. They had a freedom to challenge assumptions from a place of knowledge. SAP and others did not have this opportunity to kick the ladder away.
Vcola was founded by ex-Workday executives Annrai O’Toole, David Clarke, and Kashif Qayyum. They built the frameworks at IONA and Cape Clear (acquired by Workday) that power 100s of millions of users, Workday and a massive ecosystem. They have climbed the ladders of successful enterprise software. They have the medals and scars.
Read about their Irish Payroll adventures too.
In my post on A16Z and SAP I said:
For now, the ERP vendors aren’t going away. A16Z are indeed right to note that the context layer will become vital, agents will do more of the system interaction, extensibility will become significantly easier, and system migration and support will be significantly improved with AI tools. I hold the ERP vendors have more fight in them than A16Z reckon, but a new wave of vendors are going to do very well on top of the existing systems of record.
But I’m also confident someone will come up with a 100x better way of doing systems of record (and the context layer) than we have today. I am hoping that we fund this vendor.
When I wrote this, we had already made the decision to invest in Vcola, but we had to keep it under wraps.
Perhaps best way to advance enterprise software is to understand what has gone before, and then kick the ladder away. We think that’s what the team at Vcola are doing.
We are also pleased to be investing with Base10, Frontline and Booom.
As usual, here’s a tune. It is a banger. Best played very loud. Helicopters optional. I could have gone with John Lennon’s Imagine, but I think this fits better. Annrai and David will know why.