Beyond ERP: Will AI Agents Replace the Digital Backbone of Manufacturing?
Why Calin Drimbau’s “Death of ERP” provoked one of the most heated debates in enterprise tech
Calin Drimbau, CEO at Broadn, recently ignited a lively debate on LinkedIn by declaring that “ERPs will slowly die.” His argument is simple but radical: manufacturers are exhausted by expensive ERP implementations, endless consultant fees, and systems that are outdated before they go live. Instead of yet another ERP upgrade, Drimbau suggests that companies want “agents on top of their data.”
The vision is clear. Data is the real asset. The next competitive edge lies in orchestration layers and adaptive workflows powered by AI agents that can run operations end-to-end. For Drimbau, this is more than an efficiency play. It’s a generational shift: “SaaS, as we knew it, is dying. This is the warm-up act for agents running operations end-to-end. Start planning your data + agents. Don’t wait until you’re forced.”
That post struck a nerve. Hundreds of professionals weighed in, with reactions ranging from enthusiastic agreement to outright dismissal. The debate reveals deep tensions in how we think about enterprise technology. Are ERPs obsolete? Or are AI agents just the latest shiny object?
What follows is a synthesis of the most compelling viewpoints from that thread, interwoven with analysis to help us think critically about where the enterprise stack is headed.
The Case Against ERP: A Tired Model
Several voices echoed Drimbau’s frustration with ERP’s complexity and cost. Brandon Card, Founder and CEO at TERZO, argued that “The ERP is missing the most critical info—the document metadata.” In his view, core transaction systems fail to capture the rich data businesses actually need to make decisions.
Rod Bishti, CEO at Bosbec, added: “Adaptive layers that orchestrate workflows instead of locking companies into one structure will change the game.” Here the critique is not only about cost but about rigidity. ERP’s promise of integration has too often become a straitjacket.
Others went further. Vish Puttagunta, CEO in food & beverage, bluntly declared: “Consulting based ERP companies will certainly die. Nobody’s got time and money for that. If you can’t deploy within 2–3 months and show at least 3% in recovered margins, it’s not worth doing.” This sentiment reflects a broader impatience among operators: time-to-value has become the ultimate metric, and ERPs rarely deliver it.
But the harshest critique came from Rodrigo Ramírez Riquelme, a seasoned supply chain manager: “I’ve seen too many companies celebrate the ‘death of ERP’ every five years. The reality? ERP doesn’t die... it just mutates. The failure is not the system itself, but the obsession with long, expensive, big-bang projects that collapse under their own weight.” His warning is that technology is not the root problem—governance, processes, and discipline are. Without fixing those, AI agents will only orchestrate chaos faster.
The Data Question: Where Does It Live?
Perhaps the sharpest counter to Drimbau’s vision came from those who asked: if ERP dies, where does the data go?
Sam Graham, an ERP author and long-time observer, put it bluntly: “I believe that AI will make accessing and analysing the data in ERP systems easier; but where is that data going to sit if companies don’t have an ERP system? In unconnected silos and spreadsheets?”
Nick Van Maele, a senior project manager, expanded this point: “When presented with the same input, ERP needs to generate the same output every time—i.e., be deterministic. Probabilistic agents cannot guarantee that. As a comparison: radio was invented more than a century ago and it still has its place. Same for ERP and relational databases. They’re not going to disappear any time soon.”
This deterministic argument is crucial. Compliance, accounting, and tax reporting require systems of record. AI agents, by contrast, are probabilistic. As Vi Jayakumar, SAP consultant, stressed: “ERP systems like SAP are very powerful and integrated when implemented correctly. AI should be leveraged, but data integrity and governance remain non-negotiable.”
The debate here reveals a foundational split. AI proponents emphasize agility, orchestration, and adaptability. ERP defenders emphasize stability, determinism, and auditability. Both are right, but they point to different risk tolerances.
Agents as a Layer, Not a Replacement
A large number of participants took a middle ground: agents will not replace ERP, but enhance it.
Mateus Parra Sgarbosa suggested: “I tend to think the ERP as the system of record will stick around, but the way work is done around it will be completely redefined. Over time, the ERP may look more like a ‘dumb database,’ with orchestration, intelligence, and workflows happening in the agent layer.”
Melvi Pais drew a parallel with malls and e-commerce: “You could think of agents as extensions or enhancements of ERP, not replacements. When e-commerce came to life, malls were supposed to go extinct. They didn’t. Coexistence is the more likely scenario.”
This coexistence thesis is perhaps the most pragmatic. AI agents can free ERP from being the interface for every workflow. Instead, ERP remains the backbone while agents handle the orchestration and adaptive decision-making.
The UX and People Factor
One striking omission in Drimbau’s original post was any mention of users. Mohamed Marwan, an ERP veteran, flagged this: “As long as companies have users who struggle with ERPs, they will continue to struggle with agents (unless agents will completely replace all users and processes, and then we have a scary version of a DAO). Agents add layers of complexity, not necessarily simplicity.”
Calin Drimbau responded: “I think UX will evolve as we move away from user interface = an interface to control what actions a user wants to take, to user interface = an interface to monitor and respond to agentic systems.”
This exchange highlights a central challenge: agents may shift the role of users from operators to supervisors. But will that make adoption easier or harder? If ERP already suffers from poor user adoption, layering agents risks creating a more sophisticated mess unless UX design is prioritized.
AI as Revenue Engine, Not Cost Cutter
Several comments reframed the debate entirely. Jason Ling, AI product executive, argued that most companies are misusing AI by treating it like ERP: “Everyone’s treating AI like a cost center. Congrats, you just bought the world’s most expensive calculator. The CEOs who get it? They’re using AI to attack revenue: predictive upsells, dynamic pricing, product features that ship themselves. AI isn’t here to make your bad business model 20% more efficient. It’s here to build a different business model entirely.”
This perspective shifts the debate from ERP vs. AI to how companies think about technology altogether. If AI is merely deployed to reduce ERP costs, it risks missing the transformational potential. If it is used to reimagine the business model, it could become the growth engine that ERP never was.
The Evolutionary Path of ERP
Others reminded us that ERP has been declared dead many times before. Chris Andrews, Partner at EY, recalled: “I read the same 15 years ago when SaaS boomed. ERP’s are still here. Mature companies with 40+ years of ERP experience think in terms of multi-tiered architecture—systems of record, differentiation, innovation, and orchestration. The split varies, but ERP remains core.”
Ketharaman Swaminathan was even more blunt: “LOL, I heard ‘RIP ERP, Hello eRP’ in 1999. ERP has not only survived for 26 years but flourished. So it will for the next 26 years, AI or no AI.”
History suggests that ERP adapts. Cloud, SaaS, mobile, and now AI have all been layered into ERP. The big vendors—SAP, Oracle, Microsoft—are already embedding AI copilots. The likeliest outcome is not death, but metamorphosis.
Governance and Risk: Who Owns the Failure?
Daryl Hewison raised a fundamental question: “How does the data get there, that the AI agents are on top of? Are other AI agents creating it? I have a feeling there are some colossal disasters on the way rooted in confusion about what business data even is, and how it’s generated.”
This is not a trivial concern. AI hallucinations are one thing. AI errors in financial reporting, compliance, or safety are quite another. As Sébastien Troszczynski noted: “Do you want a shift in the validation workflow because the agent thought of something nice? In some businesses, you still need this ERP-style determinism.”
If ERP has been criticized for over-engineering, AI agents may risk under-engineering. Governance and accountability will be decisive in determining whether agents truly replace ERP or simply complement it.
So, Will ERP Really Die?
The answer depends on what we mean by “ERP.” If by ERP we mean bloated, consultant-heavy, decade-long projects, then yes, that model is dying. If we mean deterministic systems of record that provide compliance, auditability, and stability, then no—they remain indispensable.
The real shift may be semantic. ERP as we know it may fade into the background, its transactional core buried beneath orchestration layers of AI agents. To the user, the “ERP” may disappear. But underneath, something still books the purchase orders and reconciles the accounts.
As Nick Van Maele wisely concluded: “At this time, we can state that agents can control ERP functions or augment them, but replace them they do not. Not yet, anyway.”
Final Reflection
Calin Drimbau’s provocation is valuable precisely because it challenges us to confront ERP fatigue. But the LinkedIn debate makes clear that replacing ERP is not as simple as layering agents on raw data. ERP persists because businesses need deterministic backbones. AI agents will matter most not as replacements, but as orchestrators, enhancers, and perhaps one day, disruptors.
The deeper question is not whether ERP will die, but whether enterprises will use AI to finally move beyond incremental optimization. Will AI be the world’s most expensive calculator—or will it become the infrastructure for new business models?
That is the conversation worth having.
 | A guest post by
|
I mean AI is basically doing a lot of things for supply chains in a good regard