NetSuite customers, take heart

The magic moment

I knew Splotch was doing something important when we ran our first workshop with a NetSuite customer, Paul. He zoomed in and out of business areas on the huge graph (what we call "the universe") that contains every object and reference in his instance. "It's very cool to see my instance like this, but what do I DO with it?" The "universe" was far more data than a human could ever use or understand. "What do you wish worked better?" I asked, knowing we were on this call because he had spent more than a million dollars on his instance, yet like most NetSuite instances, it remained both a mystery and a mess.

"Invoicing," he immediately responded. "When we moved to NetSuite, we hired ACS for the implementation. It took them nine months to launch, and when we went live, invoicing didn't work. For three months." We looked at custom invoicing scripts. Most were patches for previous edge cases and problems. "What part of invoicing is most broken?" I asked. As Splotch went to work, a small smile appeared. Relief. "This," he said as we dragged each of the components in the blast radius of his project into a workstream, "is how an ERP should actually work."

We resolved that issue on the call. I sent him a sales order for Splotch and he signed it the same day, without redlines.

First into the cloud

Did you know that NetSuite, founded in 1998, is the world's first cloud computing company? When Evan Goldberg, the founder, got on the phone with Larry Ellison, this was his pitch:

"Manage your entire business operation from one unified platform, using a simple web browser, with real-time access to your data from anywhere."

I imagine Evan looking at the current state of play — spreadsheets for small companies, on prem legacy spaghetti systems for big ones — and shaking his head: "THIS is how people are running their businesses?" Seeing what he saw, he must have felt the looming, fluttery feeling when you know you can make a dent in the world.

I know the feeling. But before I tell you why, we need to cover the rest of the NetSuite story.

NetSuite built the core accounting features first, then came inventory, order management and HR. As NetSuite grew into a single platform across departments, customers needed a lot of custom business logic. No problem: write it in SuiteScript, based on JS, the core language of the web. Now customers had an out of the box core with an infinitely extensible, customizable horizon. Hooray!

Fast forward twenty-plus years, and I'm talking to Paul. When we looked at his entire instance we found 13,200 objects total, and 12,409 of them were custom. Those consultants did deliver a lot of work — fixing the fixes. "That's hilarious," Paul said, incredulous. "You can use that as the case study of what not to do with NetSuite," he said looking at a million dollars of consulting fees. We all chuckled mournfully: this is exactly where NetSuite customers end up.

An Innovator's Dilemma

Oracle was always the mothership and original backer, and it bought NetSuite eventually. Every company must navigate how to balance explore versus exploit; Oracle is on the far side of that spectrum, farther out than other big tech companies. I doubt that Evan has much room to ship explorations these days. NextSuite Next, started as the new version of NetSuite, was gutted to protect the cash cow legacy product. Now NextSuite Next is a grab bag of AI features that Claude can do better.

But from Oracle's perspective, why should they explore? They have the world's best lock-in. They run the beating heart of the enterprise. Paul told us flat out: the switch to NetSuite was so painful and expensive that he will likely never switch again. "We'll just build around it; our goal is to never interact directly with NetSuite." Now that's lock-in — no wonder NetSuite doesn't feel it needs to rethink its product. It can coast.

I was on the phone with a CFO earlier this year who said, "We buy a company every six months. What ERP do we use? Three separate NetSuites a handful of others! We cobble it all together. Switching ERPs is open heart surgery. We'll manage as best we can."

This is happening all over legacy SaaS: these companies have the data and the business logic. They're battle tested. The legacy players are not going away. The profit margins are good. The SaaS index is up. Salesforce is outpacing NetSuite with its headless strategy, an admission that AI is the new UI. But now it's a headless mess. These systems need a new interface. Marc Benioff announced that projects will be delivered twice as fast using AI, but my conversations with Salesforce SIs indicate that, as with many of his announcements, this is aspirational.

Just use AI!

Despite this enormous switching cost, VCs have placed equally enormous bets on AI-native ERPs. Are they better architected for today's world? Absolutely. But I've talked to the founders of these companies. They're going after QuickBooks, not NetSuite. Even lavish offers to buy out existing ERP contracts don't work for mature companies. Open. Heart. Surgery.

We're not the only ones to notice that you can use AI to investigate and work with legacy systems. The other attempts I see are a mile wide and an inch deep. They're taking on Databricks, MS Dynamics, Salesforce, NetSuite, SAP… In other words, they're going to get eaten in a year or two by Claude.

To survive the advance of models, I believe you need deep insight into a system so that you understand all 12,409 custom objects running and how they interact with objects all over the system. You have to help the agent see deeply, in a way a human can follow along and correct. When we started out pointing LLMs at problems, they hallucinated elements into existence, out of existence and misunderstood dependencies. Only a deep knowledge graph tailored to the quirks of a specific system will produce trustworthy results. (And yield 10-20x compute cost savings.) We recently went up against one of these companies — a serious, Sequoia-backed crew. In practice, it was just AI Jira. We won hands down.

Just use AI (services)!

What about AI-enabled services? Have we not all heard tales of the vaunted FDE? VCs used to sneer at services — no longer! There is no shortage of tokens thrown at black box ERP and CRMs, but the core bottleneck remains: human experts have to double-check everything. Of course services will get better. But that's just whipping the horse harder. A horse is not a car.

Yes, AI enabled services are important — we will always need humans in the loop exercising judgment, holding responsibility. But is that really the best we can come up with? Evan didn't start a cloud-based consulting company to help with spreadsheets and on-prem systems. He built a new platform that could handle things in a different way, and customers made it the market standard. So what are customers faced with the spaghetti of the current cloud SaaS core systems to do?

Garden or graph?

Imagine if we said to your engineering team: "When you're ready to ship code, just paste it into this doc and ask your team to leave comments." That's where we are with core systems. There's no Github: no source of truth, no way to diff changes, no version control, no visual layer.

AI enabled services is another way of saying "Let's keep the customers, the people who are actually responsible for the business outcomes that these ERPs and CRMs are supposed to deliver, let's just keep them in an endless loop of meetings and guesswork — but faster and better than before."

No. The AI era is fundamentally different. Software doesn't follow the same rules. There are two kinds of software now:

Gardens: Non-essential software that you grow like a garden with AI. You can change it on the fly, make it different for every user. The user can actually change it. Let a thousand probabilistic flowers bloom.

Graphs: Core software that MUST work. You cannot just let the LLM skip about this garden making merry. You must define exactly what happens at each step. It's essentially a giant flowchart, scaled to millions of nodes, pointing in many directions. This second kind of software — ERPs, CRMs — require guardrails for AI agents in the form of a knowledge graph. And that knowledge graph must allow humans to follow along.

Enter the Mastergraph

This era requires a new category, just like the cloud era did. We need a Github for core systems — with a new UI that helps humans follow along. This "mastergraph" should span all your core systems, allowing you to query, update and port logic how and wherever you want. The core of this product is a deep knowledge graph used as guardrails, so changes can be made quickly and safely. It should include:

1. A knowledge graph of what is actually there and the relationship between each object.
2. A "blast radius" view of the relevant business logic and affected components for any change you want to make.
3. Branching and diffing functions that allow agents to propose changes that must adhere to the guardrails in the knowledge graph.
4. Governance features like version control, rollback and human-segmented comments and approvals to understand when, why and how changes were made. Changing the system is easy. Safe change that you understand and can follow, is hard.

I'm not going to pretend this is all figured out. Most companies are still asking for a faster horse; they can't imagine riding in this car. But a few companies that are so fed up with spending inordinate amounts of money for so little return will get curious about this new machine. We already have the roads; the LLMs built them. Splotch is building the vehicle to hold these systems and navigate the AI era — without requiring open heart surgery.

Over the past months we've had conversations and design partnerships with the top system integrators. Most of them have no real interest in changing their business model except at the edges. A few of them, like Plative, whose CEO is heading an agentic task force and asked me questions I had never considered, are going to make the very hard move to the new model, like Netflix did from DVDs to streaming. They're helping us build the platform that will run core systems going forward and figuring out how services will look in that new world — because humans will always have to exercise their real-world judgment and accountability.

Launching Splotch for companies

This week we're launching open access to companies. Our offer is simple: bring us your top 3 NetSuite pain points and we'll scan your instance for free, open up the black box to show you the "universe" of what's inside and generate an action plan to address them now. You see and control your own instance: make small and medium changes safely, or work with one of our partner SIs on big projects.

This week an exceptional solution architect who has been immersed in NetSuite for 15 years, began doing things with Splotch I had never considered. "The next year and a half are going to be really fun," she said with that same little smile I'd seen before.

It's no fun to be a NetSuite customer right now. But it could be.

p.s. Salesforce customers, stay tuned…