The $500 Billion Problem: Why Direct Procurement Has Been Ignored by Software for 30 Years
Noam Shakuri
Sometime in the early 1990s, a group of software engineers at companies like Ariba and SAP made a decision — largely by default — that would shape the procurement software industry for the next three decades.
They built for accounts payable.
It was a reasonable starting point. Accounts payable meant structured data: invoices with amounts, vendors with IDs, payments with due dates. It meant high transaction volume — thousands of payments per month at large organizations — which made automation ROI easy to calculate. It meant a clear internal champion: the CFO, who wanted to close the books faster and catch duplicate payments.
The decision made sense at the time. It also meant that the $8 trillion per year of direct procurement spend — raw materials, components, manufacturing inputs that represent the majority of cost of goods sold for most manufacturers — was left to be managed by transactional ERP systems that were never designed for the relationship-intensive, operationally complex, communication-heavy reality of direct procurement management.
Thirty years later, that gap is still open. And it is enormous.
The Procurement Software Origin Story
To understand why direct procurement has been underserved, you need to understand how the procurement software industry was built.
The first wave of procurement software — Ariba, founded in 1996; Coupa, founded in 2006; Jaggaer, assembled from a series of acquisitions — was built around a specific problem: the uncontrolled, high-fragmentation spending that large enterprises made on everything except their core manufacturing inputs. Travel, office supplies, marketing agencies, software licenses, consulting services, facilities management. This category of spending, which the industry calls indirect spend, had identifiable characteristics that made it attractive to automate:
High transaction volume, low unit complexity. A company might process 50,000 purchase requisitions per year for indirect goods and services. Each transaction is relatively simple — a standard good or service, a catalog price, a straightforward approval workflow. Automation of high-volume, low-complexity transactions produces visible, measurable efficiency gains.
Clear internal pain points. Procurement teams were buried in paper-based approval processes. Finance teams couldn't close the month without chasing down purchase orders. Auditors found compliance gaps in uncontrolled spending. These pain points had named stakeholders — CFO, Chief Procurement Officer, Head of Finance — who had budgets and authority to buy software.
Favorable ROI narrative. If you could show a CFO that you would reduce maverick spend by 15% on a $100 million indirect spend base, that was $15 million in savings — a compelling, auditable ROI story. Direct procurement savings are harder to isolate because they involve supplier negotiations, quality improvements, and supply continuity benefits that don't show up cleanly in an AP report.
By the mid-2000s, the indirect spend management market was mature. SAP acquired Ariba for $4.3 billion in 2012. Coupa went public in 2016 and was eventually acquired for $8 billion. Jaggaer was valued in the billions. A dozen point solutions addressed specific sub-categories: travel management, contingent workforce, software licensing, expense management.
The market had, in the terminology of software investment, been well-served.
The Direct Procurement Gap
Direct procurement — the purchasing of raw materials, components, and manufacturing inputs that go directly into a company's products — is a different activity entirely.
The scale of it is larger. For a typical discrete manufacturer, direct procurement represents 60 to 80% of cost of goods sold. For a pharmaceutical company, it may be 70 to 85% of manufacturing cost. For an electronics contract manufacturer, it can exceed 90%. The cost base that direct procurement manages dwarfs the indirect spend that the procurement software industry spent three decades optimizing.
The operational characteristics are fundamentally different:
Supplier relationships are strategic, not transactional. A company buys office supplies from whoever has the best catalog price. A company buys specialty steel alloys, precision components, or active pharmaceutical ingredients from suppliers who may have taken years to qualify, who produce to specification, and who represent genuine supply risk if they underperform or exit the relationship. Managing these relationships requires continuity, expertise, and a quality of communication that cannot be reduced to a purchase order number.
BOMs create complex dependency structures. A finished product may incorporate 500 distinct components from 200 different suppliers. A delay or quality issue with any one of them can halt production of the entire assembly. Understanding this dependency — knowing which suppliers are on the critical path, which components have single-source risk, which lead times are tightening — requires real-time visibility into supplier communication that no ERP was designed to provide.
Quality requirements drive ongoing documentation needs. Direct procurement is not just purchasing — it is also qualification, certification, deviation management, and change control. A supplier changing a manufacturing process must notify customers and obtain approval before the change takes effect. These interactions generate continuous documentation requirements that live at the intersection of procurement and quality management.
Lead times require proactive management. Indirect procurement lead times are generally measured in days. Direct procurement lead times — particularly in electronics, chemicals, and specialty materials — are measured in weeks to months, with significant variability. Managing these lead times proactively requires tracking confirmation status across hundreds of open purchase orders simultaneously, identifying exceptions early, and acting before delays cascade into production disruptions.
None of the procurement software platforms that were built for indirect spend were designed to handle these characteristics. Coupa's spend management suite handles catalog-based purchasing and approval workflows. SAP Ariba's sourcing modules handle strategic sourcing events and contract management. These are excellent tools for what they were designed to do. They are the wrong tools for the daily operational reality of direct procurement.
The ERP Illusion
If the indirect spend platforms don't solve direct procurement, what does? The answer most manufacturers would give is: the ERP.
This answer is both correct and deeply misleading.
ERPs — SAP S/4HANA, Oracle Manufacturing Cloud, Infor, NetSuite — do manage direct procurement in the sense that purchase orders live in the ERP, vendor master data lives in the ERP, and goods receipts are recorded in the ERP. The ERP is the system of record for what happened in procurement.
It is not, and was never designed to be, a system for managing what is happening.
The actual work of direct procurement — the work that determines whether production lines run, whether suppliers deliver on time, whether quality issues are caught before they enter the assembly process — happens outside the ERP entirely. It happens in email. In phone calls. In spreadsheets that buyers maintain because the ERP doesn't surface the right information at the right time. In the institutional memory of experienced buyers who know that Supplier X always delivers three days late and Supplier Y needs a follow-up call two weeks after PO issuance or nothing happens.
Consider what happens when an ERP-centric direct procurement operation receives a purchase order acknowledgment from a supplier. The supplier emails the buyer. The buyer reads the email. The buyer extracts the confirmed delivery date. The buyer navigates to the ERP and manually updates the PO line with the confirmed date. The ERP now has current data — but only because a human translated an unstructured email into a structured ERP record.
If the buyer is busy that day — managing 40 other supplier emails, attending meetings, covering for a colleague — the update might happen tomorrow, or the day after. In the meantime, the production planning system is making decisions based on the requested delivery date, not the confirmed one. The ERP appears to have the data. The data is stale.
This is not an ERP failure. ERPs were designed to manage structured, transactional data at scale. They were not designed to read emails, understand context, handle exceptions, or manage the dynamic, relationship-intensive communication layer that connects manufacturers to their direct supply base.
That layer has been left to people. And the people are overwhelmed.
The Scale of the Opportunity
The investment community has begun to recognize the magnitude of the direct procurement software gap. The calculation is worth building explicitly.
Total global manufacturing output is approximately $16 trillion per year, according to World Bank and UNIDO data.
Direct procurement as a share of manufacturing cost ranges from 50 to 85% depending on the industry. A conservative estimate of 60% implies approximately $9.6 trillion in annual direct procurement spend globally.
Software spend as a percentage of procurement-managed spend is a well-established benchmarking metric. The Hackett Group and Gartner both estimate that organizations spend approximately 0.3 to 0.8% of total procurement spend on procurement technology (software, implementation, and support). Applying the lower bound of 0.3% to the $9.6 trillion direct procurement base implies a theoretical software spend opportunity of approximately $29 billion per year.
In practice, current direct procurement software penetration is far below this theoretical ceiling. Most manufacturers' direct procurement technology stack consists of an ERP (already budgeted as core infrastructure, not specialized procurement software) and spreadsheets. The delta between current spend and addressable spend represents the market opportunity.
For context: Coupa was acquired for $8 billion in 2023. Coupa's primary addressable market is indirect spend — a market that represents a small fraction of total procurement spend. The indirect spend software market has generated multiple billion-dollar outcomes precisely by serving the smaller, less complex portion of the procurement universe.
The direct procurement software market — serving the larger, more complex, more strategically important portion — has produced no equivalent at-scale outcome. Not because the problem is less valuable to solve. Because the problem was, until recently, harder to solve than the tools available could handle.
Why AI Agents Are the Unlock
The direct procurement problem was not solvable by the first two generations of enterprise software for a specific reason: the problem is inherently unstructured.
Rules-based automation — the architecture that powered the first generation of procurement software — works when inputs are standardized and exceptions are rare. Build a workflow. Define the rules. Handle the common case automatically; route the exceptions to humans.
Direct procurement fails this model at every step. Supplier communications arrive in hundreds of different formats. Confirmations contain ambiguous language that requires interpretation. Exceptions are not rare — a meaningful percentage of every PO batch has some form of deviation from plan. The "rules" for handling each deviation depend on inventory levels, production schedules, supplier history, and business context that changes daily.
Rules-based automation produced partial solutions for structured parts of the direct procurement workflow: EDI for the largest, most standardized supplier relationships; basic ERP alerts when delivery dates pass. It produced nothing for the communication-heavy, judgment-intensive operational layer that constitutes most of the actual work.
Large language model-powered AI agents change this calculus in three specific ways:
Unstructured input comprehension. LLMs can read a supplier's email confirmation — regardless of format, language, or phrasing — and extract the same structured information that a buyer would extract. The model understands that "we can deliver the balance of your order by the 15th" means a partial confirmation with a specific delivery date, even without a structured data template. This capability eliminates the primary bottleneck in direct procurement automation: the requirement for structured input.
Contextual decision-making. Beyond reading supplier communications, agents can evaluate them against business context. Is a 3-day delay acceptable given current inventory? Is this supplier's confirmation rate historically reliable? Does this component sit on the critical path for next month's production build? These are judgments that require context — and agents connected to ERP data can make them at the same quality level as a buyer, for routine cases, at unlimited scale.
Edge case handling. The proportion of direct procurement interactions that fall outside a simple confirmation-or-delay binary is significant. Partial shipments, pricing disputes, capacity constraints, material substitution proposals, force majeure notices — each represents an "edge case" that rules-based automation routes to a human by default. LLM-powered agents can interpret and appropriately handle or escalate these situations, not because they were explicitly programmed for each case, but because they understand language and context at a level sufficient to recognize what kind of situation they're dealing with.
The combination of these three capabilities means that the full supplier communication loop — read inbound, understand context, decide action, execute response, update systems — can be automated for the majority of direct procurement interactions, not just the structured minority.
Rules-based automation could automate approximately 20 to 30% of direct procurement communication. AI agents can automate 80 to 90%. That difference is the difference between a tool that helps and a system that transforms.
What the Next Five Years Look Like
The operational model of procurement departments at manufacturing companies is about to undergo a structural shift. Understanding what it looks like on the other side matters for both the leaders managing through the transition and the investors and executives thinking about where value will accrue.
The procurement org of 2030
In five years, the procurement organization at a well-run manufacturer will look structurally different from today's. The operational headcount — the buyers who spend their days reading supplier emails, updating delivery dates in the ERP, sending confirmation requests, and logging exceptions — will be substantially smaller. Not because companies cut investment in procurement, but because the operational execution layer that occupied that headcount will be running autonomously.
What grows is the strategic layer: supplier development managers who invest in deepening relationships with the most critical supply partners. Category strategists who design the sourcing approach for complex spend categories. Supply chain risk managers who monitor geopolitical, quality, and financial risk across the supply base. And a small group of operations specialists who own the configuration and continuous improvement of the AI systems doing the execution work.
This is not a hypothetical. RH Electronics, managing 1,000+ suppliers and 25,000 active purchase order lines with a team of 35, already operates with a procurement model where AI handles the execution layer. Their buyers manage exceptions and relationships. The AI manages the volume.
The competitive divide
The organizations that build operational AI infrastructure for direct procurement in the next two to three years will have structural cost and capacity advantages that will be difficult for competitors to close. The advantages compound:
- Lower operational cost per PO processed
- Higher data quality in the ERP, enabling better demand planning and inventory management
- Earlier exception visibility, enabling faster response and lower production disruption costs
- Larger effective supplier base management capacity without headcount growth
The organizations that delay will continue operating a manual model against competitors running an automated one. The gap between those two operational cost structures grows with every year of delay.
The software market implications
For the software market, the direct procurement opportunity represents the next major investment cycle in enterprise procurement technology. The indirect spend wave generated public companies, major acquisitions, and significant investor returns between 1995 and 2023. The direct procurement wave is beginning now, at a larger total addressable market, with a more capable technology substrate.
The platforms that emerge from this cycle will look different from Coupa and Ariba. They will be agent-native rather than workflow-native. They will operate at the email layer rather than the portal layer. They will integrate with ERPs rather than compete with them. And they will be judged not by user adoption metrics but by the percentage of supplier communications they handle autonomously without human intervention.
Evolinq's Position
Evolinq was built from the first line of code for the direct procurement gap.
Not for spend analytics. Not for strategic sourcing events. Not for contract management or supplier qualification workflows — though those are valuable problems. For the operational execution layer: the daily, high-volume, unstructured supplier communication that moves between manufacturers and their direct supply base and that has never had a purpose-built software system to handle it.
The architecture reflects this focus. Evolinq operates at the email layer — the universal supplier communication channel — and handles supplier responses in any language, format, or structure. It connects to ERP environments (SAP, NetSuite, Infor) not as a separate system that requires integration projects, but as an operational extension that reads and writes to specific data objects through standard interfaces. It deploys in days, not months, because it does not require suppliers to change anything about how they communicate.
The decision logic runs on top of real ERP data — actual inventory levels, actual production schedules, actual supplier performance history — so that the agent's determination of whether a delay is acceptable or critical is the same determination a buyer with full system access would make, not a rule applied without context.
Every action the agent takes is logged with full traceability: which supplier communication triggered the action, what the agent understood from it, what decision logic was applied, what was written to the ERP, and what was communicated back to the supplier. For regulated industries — pharmaceutical, food, aerospace — this traceability is a compliance asset. For all manufacturers, it is the audit infrastructure that makes autonomous execution trustworthy.
The problem Evolinq addresses is not a new problem. It is a problem that manufacturers have been managing manually for decades because no software solution existed that could handle the unstructured, judgment-intensive, relationship-sensitive nature of direct supplier communication at scale. LLM-powered agents are the technology that finally makes the problem addressable.
The Conclusion Procurement Leaders Need to Hear
The procurement software industry spent thirty years generating enormous value for the minority of procurement spend that was easiest to digitize. Indirect spend — travel, office supplies, contingent labor, marketing services — represents perhaps 20 to 40% of total procurement for a typical manufacturer, and it has been served by a multi-billion-dollar software ecosystem.
The majority of procurement spend — the raw materials, components, and manufacturing inputs that constitute the core cost structure of manufacturing — has been served by ERP systems that record transactions and email inboxes that manage relationships. The humans bridging that gap are talented, experienced procurement professionals who are spending most of their time on work that should not require talent or experience.
The technology to close this gap now exists. The architecture — AI agents operating at the email layer, connected to ERP data, deploying without supplier change management — solves the problems that made direct procurement automation intractable for three decades.
The manufacturers who deploy it in the next two years will build operational advantages that their competitors will spend five years trying to match. The CPOs who understand this market shift now will be the ones speaking at the conferences explaining how they made it happen.
The $500 billion problem is not waiting for a better solution to be invented. It is waiting for organizations to recognize that the solution already exists.
We're building the operational layer that direct procurement has been missing. See it in action.
Market data referenced: World Bank and UNIDO global manufacturing output estimates; Hackett Group and Gartner procurement technology spend benchmarks; SAP-Ariba acquisition ($4.3B, 2012); Coupa acquisition by Thoma Bravo ($8B, 2023); industry COGS composition benchmarks (electronics, pharmaceutical, discrete manufacturing).