I've spent the last two decades inside some of the most operationally complex companies on the planet — Apple, Johnson & Johnson, Twilio, Roku, GoPro. At each one, I was tasked with the same fundamental challenge: make the ERP work. Make it reflect reality. Make people actually use it.
After all of that, I started NexuSphere AI. Not because I saw a market opportunity in a slide deck. Because I was tired of watching smart, capable operations teams spend their best hours doing things that software should have been doing for them.
Here's what I learned.
Lesson 1: ERP projects fail at change management, not implementation
I've watched multi-million dollar ERP deployments succeed technically and fail operationally. The system is configured. The data is migrated. Go-live happens. And then, six months later, the team is still doing half their work in Excel.
The problem is almost never the software. It's that the software requires people to change how they work — but gives them no reason to. The ERP is slower than the spreadsheet for the first 90 days. It has more steps, more fields, more approvals. The person doing the data entry sees no benefit from the accuracy they're now required to maintain.
This is why NexuSphere AI is designed to eliminate data entry rather than improve it. The system should be capturing information from invoices, emails, and transactions automatically — not asking a human to type it in.
Lesson 2: The real cost of ERP is what it prevents you from seeing
At every company I worked with, there was always a version of the same conversation: "Why didn't we know about this sooner?" A vendor was overbilling us for six months. A product category was declining and nobody caught the early signal. An inventory discrepancy was sitting unresolved for a quarter.
The honest answer is always the same: the data was in the system, but nobody was looking at it in the right way at the right time. Traditional ERP is a transaction recording system. It captures what happened — but it doesn't tell you what's about to happen, or flag what shouldn't have happened at all.
After enough of these conversations, I became obsessed with one question: what if the system was always watching, always comparing, always alerting — not just recording?
That question became the ML Hub and the Exception Queue in NexuSphere AI. Four models running continuously, surfacing what matters before it becomes a problem.
Lesson 3: Finance teams are the most underserved users in the building
I've worked alongside a lot of finance teams. Without exception, they are among the most capable people in any organization — and they spend a disproportionate amount of their time on work that should not require a human at all.
Manual journal entry. Manual reconciliation. Chasing down backup for GL entries. Building the same reports in Excel at the end of every month because the ERP can't produce them fast enough or in the right format. Running the close process over two weeks because the data is only reliable once everyone has submitted their adjustments.
NexuSphere AI posts every journal entry automatically at the moment of the transaction. The close is always done. Month-end is not an event — it's a formality.
Lesson 4: The implementation timeline is the real barrier to entry
I've led NetSuite and SAP implementations. I've seen the 12-month timelines. I've been in the room when the scope increases by 40% in month three because the requirements weren't documented well enough. I've watched companies delay other strategic initiatives because IT bandwidth was consumed by an ERP project that was supposed to be done six months ago.
The 6–18 month ERP implementation is not a necessary feature of enterprise software. It's an artifact of how those systems were designed — monolithic, rigid, customization-heavy. A system designed to be modular, API-first, and opinionated about how workflows should run can go live in weeks, not months.
That's not a promise I'm making based on a theory. NexuSphere AI's core O2C and P2P is live in 2–4 weeks. The full AI platform in 4–8 weeks. Because we designed it to be deployed fast — not to require a consulting engagement to configure.
Lesson 5: AI in ERP is not a chatbot. It's a workforce.
Every major ERP vendor now has an "AI" feature. It is, almost without exception, a natural language query interface — a chatbot that can answer questions about your data. This is useful. It is not transformative.
Transformative AI in ERP means: the system acts on your behalf. It matches the invoice to the PO. It flags the anomaly. It posts the journal entry. It tells you which SKUs to reorder before you run out. These are not things that require a human to ask a question — they are things that should happen automatically, every time, without someone having to prompt the system.
When I started thinking about NexuSphere AI, I kept coming back to a phrase I heard from a CFO I worked with: "I don't want a system that tells me what happened. I want a system that prevents the bad things from happening and makes the good things happen faster."
That's the difference between a copilot and an agent. NexuSphere AI has both — but the agents are where the real leverage is.
Why I built NexuSphere AI
Somewhere around year fifteen of my ERP career, I started keeping a mental list of every problem I kept seeing at every company, in every industry. The manual AP reconciliation. The month-end fire drill. The demand forecast that was always a spreadsheet guess. The ERP that nobody trusted because the data was always three days behind.
By the time I left my last role, the list was long. And the technology to solve most of it — large language models, ML pipelines, OCR, event-driven architectures — had finally matured to the point where you could actually build something useful with it.
NexuSphere AI is that list, turned into software. Built from the inside out — by someone who has seen what goes wrong and designed around it — not by engineers who have never run a month-end close or managed a vendor payment run.
If you're running a $20M–$100M retail or DTC operation and you're tired of the manual work, the spreadsheet crutches, and the ERP that was supposed to solve this but didn't — I'd love to show you what we've built.