- The Angle
- Posts
- How the system of record dies
How the system of record dies
David Peterson
There’s a popular belief that systems of record are on their way out.
The argument goes something like this: AI agents are going to do more and more of our work. They’ll handle outreach, follow-ups, research, and internal updates. And as they do, they’ll generate structured data automatically. If the agents are doing the work and generating the data as a byproduct, why do you need the system of record at all?
In that world, the system of record becomes redundant, recreated passively. Or, in my favorite version of the argument from Benn Stancil, the system of record is just abstracted away entirely because the specific content, structure and interface no longer matter.
This is something I want to believe. I hate logging activities in Salesforce as much as anyone. But I’m not sure I buy it. Or, to be more specific, I think that vision might eventually prove correct, but in the medium-term, systems of record are going to become more important than ever.
It’s dangerous for investors to make predictions. Especially now, with AI shuffling the pieces on the chessboard before our very eyes. But I thought it’d be fun to try to follow this one through.
So here’s how I see it going down.
If agents are going to do real work on your behalf, they need constraints. They need to know what data they’re allowed to use, what actions they’re allowed to take, and where to record their outputs. And users need to be able to govern all of that. The system of record is the most natural place to do it! It’s well structured. It has permissioning already. It’s legible to both humans and machines. It’s more or less up-to-date. In a world where agents are doing more, that kind of constraint layer becomes essential.
For an example of how this might work, look at ServiceNow’s AI Control Tower, which is their solution for “overseeing and coordinating both first-party and third-party AI agents and workflows,” and appears to be growing like a weed.
If you’re an incumbent system of record - well, great, right? Problem solved! The problem, though, is pricing power. Right now, systems of record are the central node and, as a result, can extract a pretty penny. But how will that premium pricing hold up against the agents who are actually doing the work? My guess is that over time that pricing power erodes. Companies will want to bring their own agents. Companies will be increasingly comfortable paying for work, and increasingly unwilling to pay for data. And incumbent systems of record will react by trying to maintain revenue (and influence) by locking things down. They’ll throttle APIs and make it harder for agents to interface with their data. Customers will balk at being forced to pay extra to get their own data out. Some early adopters will churn entirely, recreating their own single source of truth elsewhere.
And eventually, the systems of record will realize the real opportunity is not in locking things down, but in strategically opening up. The systems of record will reinvent themselves as control layers. They’ll offer agent-aware permissioning models, filtered access to data, and interfaces built for observability and audit, not data entry. They’ll stop trying to be the place where work happens and start trying to be the place where agent activity gets governed.
This is where my predictive powers begin to falter.
I can imagine a world where the systems of record, now somewhat hollowed out, slowly die off.
I can imagine another world where some of the more agile systems of record internalize much of the “agentic” activity and evolve to finally become true “systems of intelligence.”
I can imagine yet another world where an AI-first challenger leverages a wedge product to get access to novel data, circumvent the existing system of record entirely, and use that as an opportunity to build a ground-up “system of intelligence” with none of the system of record baggage.
I can’t say for sure. But if I were managing a legacy system of record today, I’d be nervous. Not because I’m going to be replaced tomorrow. But because the most interesting thing about my product, the thing users actually care about, is increasingly happening somewhere else.
What a great time to build something new.
FROM THE BLOG
No More Painting by Numbers
It’s the end of the “SaaS playbook.
The Age of Artisanal Software May Finally be Over
Every wave of technological innovation has been catalyzed by the cost of something expensive trending to zero. Now that’s happening to software.
Founders as Experiment Designers
David on why founders should run everything as an experiment.
When Growth Stalls
Or why to kickstart growth you should narrow your ICP.
WORTH READING
ENTERPRISE/TECH NEWS
Going Rogue - In his weekly newsletter Ed Sim, founder of Boldstart, writes about how we are now in a new phase of AI deployment where the focus shifts from capability to control. Through a series of recent tweets (including Jason Lemkin’s dramatic posts about how Replit went ‘rogue during a code freeze and shut down’ and Replit CEO’s subsequent response) Sim’s makes the case: “we’ve all heard about and seen enough cases that it’s clear that we’ve entered a new era, the era of guardrails, policy, and control.”
HOW TO STARTUP
Getting Noticed by AI - Another seminal post by Kyle Poyer’s Growth Unhinged, about how to get noticed by the ‘influential gatekeepers’ like Perplexity, ChatGPT, Google Gemini etc. He shares nine practical takeaways, including getting mentioned on Reddit, which is now the most cited domain within ChatGPT and Perplexity. Ed Elson tweeted that 10 years ago job postings for professional Redditors would have been called ‘Brand Communication specialists’, to which Will O’Brien (founder of Ulysses Inc) responded that Ramp (who listed the job post in question) had ‘missed a trick not calling the role Apex Redditor’.
Courting Controversy - Astromer’s marketing team needs a raise, as they signed up Coldplay’s ex Gwyneth Paltrow as their ‘temporary spokesperson’, in a clever video they posted this week in the wake of their recent controversy.
Building with Resilience - Ctech spotlights nineteen Israeli companies that have passed the $250m revenue point, including Snyk, Forter, Redis, Wiz and others. “Interestingly, most of the companies on this list are not recent startups. Coincidentally, or perhaps not ,the majority were founded around 2015, making them about a decade old. Many have also faced serious challenges along the way. Rapyd saw its valuation fall from over $12 billion to $4.5 billion today. Via abandoned its consumer transportation model and pivoted to enterprise software, becoming a major provider for public transit systems. Navan had to reinvent its business entirely after the COVID-19 pandemic decimated business travel, and it is now the closest among Israeli unicorns to achieving an IPO.”
HOW TO VENTURE
Playing the Field - Charles Hudson, founder of Precursor Ventures, writes about how as venture funds grow, they may be forced to reconsider policies around investing in competitors of their portfolio companies. "As venture fund sizes keep getting larger, I do think that tradition of venture firms having a norm (if not a stated policy) to not invest in competitive companies is likely to go away. This is simply a function of the fact that as venture funds have grown larger, it has become increasingly essential for those firms to be associated with the biggest and most important companies. The larger the fund, the more important it is to be an investor in the companies that are true outliers; there is no way to make the fund math work if you are not in those companies unless you are in other, similarly-situated companies. My sense is that there is more money chasing outliers at the moment than there are outlier companies to fund."
PORTFOLIO NEWS
Reco features in two Gartner Hype Cycles, for cyber risk management and network security.
FalkorDB has been integrated as a database backend support by Graphiti.
PORTFOLIO JOBS
Reply