AI comes for the most painful part of ERP
Inventory is where digital finance meets physical reality. When that link breaks, brands end up with stockouts, write‑offs and very expensive consultants. The fresh $55 million round for Doss is not just another AI startup story – it is a bet that the next decade of enterprise software will be won in this messy intersection between warehouses and ledgers. In this piece we look at what Doss is actually doing, why AI inventory suddenly attracts serious money, how this reshapes the ERP market, and what it means for European mid‑market companies still living inside SAP and Microsoft Dynamics.
The news in brief
According to TechCrunch, U.S.-based startup Doss has raised a $55 million Series B round. The round is co-led by Madrona and Premji Invest, with participation from Intuit and existing investors including Theory Ventures, General Catalyst, Contrary Capital and Greyhound Capital.
Doss, founded in 2023, started as an AI-native accounting platform but pivoted to focus on what it describes as an AI-first inventory management layer. Instead of replacing ERP systems, Doss plugs into existing accounting and ERP software – from legacy platforms like NetSuite to newer AI-native systems such as Rillet and Campfire – to handle procurement, inventory and supply-chain traceability.
The company mainly targets mid-market consumer brands with annual revenues between roughly $20 million and $250 million. One public customer is Verve Coffee Roasters. Doss sees itself competing with traditional ERPs and newer “agentic” procurement startups like Didero, while also partnering with accounting providers including Intuit’s QuickBooks.
Why this matters
For all the talk about AI copilots in email and slide decks, the real economic value of AI lies in moving atoms: deciding what to buy, when to ship and where to store goods. That requires an accurate, real-time picture of inventory and commitments – precisely the area most legacy ERPs handle poorly and most AI-native ERPs have underinvested in.
Doss is effectively saying: inventory is the missing infrastructure layer for AI-first finance systems. Instead of building another monolithic ERP, it wants to be the connective tissue that keeps physical goods, procurement and accounting in sync, and that is architected from day one to be “legible” to AI agents.
If this works, several groups benefit:
- AI ERP startups like Rillet and Campfire gain a way to match legacy feature coverage without sinking years into building deep inventory capabilities.
- Mid-market brands finally get modern, AI-powered workflows around purchasing and stock without committing to a multi-year ERP migration.
- Investors get exposure to a horizontal infrastructure play rather than a single-vertical SaaS niche.
But there are losers, too. Traditional ERP vendors risk being pushed down into commodity accounting, while their higher-margin implementation partners may see parts of their manual customization work automated away. And customers face a new kind of complexity: running what is, in practice, two ERP cores (one for finance, one for inventory) and hoping the integration is as seamless as promised.
The bigger picture: the unbundling of ERP
ERP has always promised a single system of record. In reality, mid-sized companies live with a patchwork: ERP for finance, spreadsheets for planning, Shopify or marketplaces for sales, and warehouse systems held together with custom scripts. Doss fits into a broader trend: critical ERP functions are being unbundled, rebuilt with AI and then re-aggregated around new data models.
We are seeing this pattern everywhere:
- In HR, Rippling and others combine system of record with automation and agents.
- In spend and finance, companies like Ramp treat accounting data as a substrate for autonomous workflows.
- In supply chain, a new wave of startups promises “self-driving operations” if you let them sit between your ERP and your logistics stack.
On the incumbent side, NetSuite has announced AI upgrades, SAP is pushing its Joule assistant, and Microsoft is infusing Copilot into Dynamics 365. Yet most of these efforts bolt AI on top of 20‑year‑old data architectures designed for batch processing and human operators, not for autonomous agents making continuous micro-decisions.
This is what Doss’s CEO is hinting at when he talks about architectures that are “most usable for agents.” AI agents need consistent schemas, clear object lifecycles and machine-readable traceability across purchase orders, shipments, returns and financial entries. Retro‑fitting that into an old ERP is much harder than building it into a new, inventory-centric layer from scratch.
If Doss and similar players succeed, the ERP of the 2030s may look less like a monolith and more like a network of specialized, AI-native services stitched together by agents. The system of record becomes more distributed; the value shifts to whoever owns the cleanest data model and the tightest real-time loop between the books and the warehouse.
The European / regional angle
Europe is ground zero for this shift. The continent’s economic backbone is the Mittelstand: thousands of mid-sized manufacturing and consumer-goods companies that run on SAP, Microsoft Dynamics, Infor and a long tail of local ERPs. Many of these systems are a decade or more old, heavily customized and extremely risky to replace.
For a German automotive supplier, an Italian fashion brand or a Slovenian food producer, the idea of ripping out SAP is close to unthinkable. But adding an AI-native inventory and procurement layer that plugs into the existing general ledger? That is far more realistic – especially if it can be piloted within one plant or product line.
Regulation adds another twist. Under GDPR and the upcoming EU AI Act, training agents on ERP data and letting them place purchase orders or adjust forecasts triggers strict requirements around data governance, explainability and human oversight. An inventory layer built explicitly with agent use in mind could embed those controls more cleanly than legacy customization.
There is also a competitive angle. Europe already has strong players adjacent to this space – from Odoo in Belgium to process-mining champion Celonis in Germany. Both have deep visibility into operational data but do not yet position themselves as AI inventory layers sitting across ERPs. That gap is now visible, and Doss’s funding will not go unnoticed in Berlin, Munich or Barcelona.
For European SaaS founders, Doss is a blueprint: you do not have to attack SAP head‑on. Instead, you can take one painful horizontal function – inventory, pricing, rebates – and rebuild it as an AI-native network service that integrates with the incumbents while gradually commoditising them.
Looking ahead: what to watch in the next 24 months
The Doss round is early evidence of where the next systems war will be fought. Several things are worth watching:
Depth of integrations. It is one thing to plug into QuickBooks and AI-native ERPs. It is another to offer bulletproof, certified connectors into SAP, Dynamics, Oracle and regional ERPs common in Europe and Latin America. The breadth – and reliability – of those integrations will determine whether Doss stays a U.S. mid-market tool or becomes a global infrastructure layer.
Agent-first workflows. Many vendors now market “AI agents,” but most are glorified macros. The real test will be: can Doss support autonomous reordering within policy constraints, multi-supplier sourcing and continuous reconciliation between physical counts and the ledger – with humans supervising exceptions instead of every transaction?
Incumbent reaction. If NetSuite, SAP or Microsoft decide to truly rebuild inventory for the agent era, they can bundle it with existing licenses and undercut standalone players. On the other hand, their internal politics and technical debt may slow them down enough for a specialist to entrench itself.
Implementation partners. European system integrators and local consultancies could either suffer margin erosion as automation spreads, or become powerful distribution channels for AI layers like Doss, packaging them into modernisation projects for the Mittelstand and mid-market retailers.
Expect the most visible traction not in giant corporations but in brands between €20–250 million revenue: large enough to feel the pain of broken inventory processes, small enough to make bold architectural bets.
The bottom line
Doss is wagering that the hardest part of ERP – inventory and procurement tied tightly to the ledger – is exactly where an AI-native specialist can beat both incumbents and new all‑in‑one ERPs. The bet makes sense: agents are useless without clean, traceable operational data. The open question for European and global companies is timing. Do you embrace a dedicated AI inventory layer now and accept the complexity of “two ERPs”, or wait and hope your existing vendor successfully rewires its stack for the agent age?
