Steel is one of the least profitable sectors in manufacturing and the overall economy. For firms producing steel on a large scale, procurement expenses are closely tied to the cost margin and hence serve as a value lever.
With a cost margin of 45 to 55%, the slightest improvement in the procurement process can translate into significant savings. While the primary component, which is the raw material, may not be structurally easy to transform into AI-led buying, low-cost consumable items can be practically re-engineered for this purpose.
For a typical large Indian steel company, the total cost component of such items can be as high as USD 3.5 billion. Even a 1% savings will translate into USD 35 million, whereas the cost of implementing AI-driven procurement will be far less.
As raw materials, energy, and supply risk dominate the cost structure and margin pressure, the procurement function must evolve. The rise of agentic artificial intelligence (AI)—systems that act autonomously, ingest context, make decisions, and execute tasks —marks a tipping point. In this article, I argue that in a large-scale steel setting, the traditional procurement specialist role is poised to be replaced (or radically reskilled) due to four forces: cost leverage, process automation, risk and resilience demands, and the availability of data, as well as AI platforms. I present key metrics, realistic impact scenarios, and a roadmap for steel-manufacturing firms.
In steelmaking, raw materials and fuels typically account for 60-80% of the total production cost, as per publicly available financial data. For example, one industry summary states, “Raw material and fuel costs typically make up 60-80 % of production costs” in the steel industry.
This means that procurement of iron ore, scrap, metallurgical coal, alloying elements, and fuel isn’t a support activity—it drives margin. Consequently, incremental improvements in procurement (price, timing, supplier risk) yield outsized returns.
In this context, the procurement specialist (category manager, buyer, or supplier risk analyst) traditionally manages: sourcing strategy, RFx and tendering, supplier negotiation, PO issuance, contract enforcement, supplier performance monitoring, and risk and compliance. Given the scale of spend and complexity of supply chains in steel (e.g., global ore flows, volatile scrap markets, energy and CO₂ regulation), it is a high-stakes role.
Factor A: Cost efficiency and automation
A large-scale steel producer aims to reduce procurement spend by 5% by implementing Agentic AI and eliminating human interactions in low-value, high-volume purchases in a phased manner. This could lead to a 20% increase in the availability of the full-time equivalent (FTE) workforce.
A comprehensive process reengineering can save 10% or more of the total procurement spend.
In another case of automation, a simple RPA reduced a two-hour task to eight minutes. Implementing Agentic AI flow will further reduce it by three more minutes, freeing up another FTE from the following operations.
For large steel manufacturers with high procurement FTE costs and transaction volumes, reducing manual tasks by 50-80% means fewer specialists are needed—or those specialists can shift to strategic oversight, while AI handles tactical work.
Factor B: Process standardisation and digitisation
A significant amount of time is lost when a newer associate replaces a procurement specialist. There is a considerable loss in resignations, and another one joins. Digitising and standardising the process will result in fewer procurement FTEs and significantly higher revenue contribution per employee.
For IT organisations supporting the procurement function of large enterprises, the essential key metrics to be tracked should be:
1. Total cost to perform the procurement process per US$1000 purchase;
2. PO cycle time;
3. Number of POs processed per procurement employee.
When processes (such as PO issuance, supplier onboarding, and invoice matching) are automated and data-rich, there is far less need for “eyes-on-tasks.” Agentic AI can act on triggers (such as supplier risk changes, spend category triggers, or contract expiries) without human involvement.
Factor C: Supply-risk, complexity, and strategic sourcing demands
Steel supply chains are vulnerable, with one estimate indicating that over 60% of raw material supply sources are located in geopolitically unstable regions. Procurement is not just about cost control; it’s about resilience and risk management.
Agentic AI brings:
Real-time supplier risk monitoring (financial health, logistics risk, geostrategic events)
Predictive ordering (lead time variability, inventory re-order triggers)
Autonomous contract enforcement (maverick spend detection, on-contract routing)
Thus, procurement specialists whose value lies in transaction management are increasingly redundant; their role must shift to strategic oversight unless replaced by agentic systems.
Factor D: Agentic systems themselves, the “intelligent buyer”
Agentic AI refers to systems that can ingest context (specs, spend history, supplier database, market intelligence), reason (evaluate alternatives, risk vs. cost), act (issue RFx, negotiate via defined rules, issue POs), and learn (through feedback loops). For example, a GenAI engine was built by a client to issue RFPs, evaluate supplier proposals, and generate category strategies in manufacturing, delivering high accuracy. The Agent also maintains extensive records of past negotiations and determines the best applicable price for the product.
In short, the technology exists to automate a significant portion of procurement specialist tasks. When scaled, the human role becomes exception handling, strategic supplier partnership, rather than issuing POs.
Automation often hits resistance. Success requires change management, governance, and a culture shift.
AI impact metrics for steel procurement
| Metric | Example (Current vs AI) | Impact / Benefit |
| Spend under management (%) | 70% → 80–90% | More spend under contracts → automated savings |
| Purchase Price Variance (PPV) | +2% → 0% or negative | AI sourcing achieves market-best or better pricing |
| Cost per PO / Invoice | $80 → $30 | Automation reduces cost-to-serve |
| PO Cycle Time (days) | 5 → 2–3 | Faster cycles → less delay in raw materials |
| Automation Rate (STP%) | 40% → >80% | Less human intervention → fewer errors, headcount impact |
| Maverick Spend (%) | 12% → <5% | Reduces off contract buys → protects negotiated savings |
| Supplier Lead Time Variability | ±22% → ±10% | Less variability → lower buffer inventory |
| Procurement Cost as % of Spend | 0.7% → 0.4–0.5% | Direct proxy for efficiency improvement |
| Procurement FTEs per $1B Spend | 45 → 25–30 | Headcount impact and automation ROI |
| Spend Reduction from AI (%) | — → 5–15% | Real cost savings figure |
Tracking these metrics month over month allows you to build a business case: fewer people doing more value-added work, cost savings, risk reduction, and better uptime.
For a procurement specialist, the current role means handling RFx and tenders, supplier negotiation, PO issuance, invoice matching, supplier performance monitoring, and risk flagging.
With the Agentic AI workflow in place, the role will change.
Tomorrow’s role with agentic AI
The procurement specialist will now focus on strategic supplier partnerships, innovation in raw material sourcing (e.g., scrap circular supply, green inputs).
The new role will involve overseeing agentic systems, including monitoring exceptions and reviewing dashboards of events flagged by AI.
The person will now drive category strategy, risk-scenario planning, ESG, and sustainability sourcing, and ultimately be the process architect, defining rules, workflows, training data for AI, managing change, and ensuring data integrity.
Role eliminated (or sharply reduced): Repetitive tasks — PO issuance, manual invoice matching, routine sourcing, commodity price benchmarking, risk-alerts triage.
Implications for staffing: Procurement departments will shift from large, transactional teams to smaller, strategic teams. Many specialists will either be reskilled or replaced by AI.
To capture value and enable this shift, a steel company should follow these phases:
1. Foundation – Data and process
Clean up the supplier master, spend-cube, contract database, PO, and invoice history. Note: Poor data or broken processes can reduce the benefits of AI.
Standardise the PO process, inventory triggers, supplier qualification, and performance monitoring.
2. Pilot targeted use-case
Pick a high-volume, low-complexity category (e.g., consumables, refractory, scrap sourcing).
Implement agentic AI to automate sourcing and PO for that category, measuring: cost per PO, cycle time, cost savings, and FTE hours saved.
3. Scale broadly
Extend agentic AI to strategic categories (such as raw materials and fuel) with more complex rules and risk analytics.
4. Build dashboards for key metrics (see above).
5. Redeploy procurement specialists to strategic roles.
6. Recognise risks: AI only works well if data, processes, and governance are solid. Without that, you may amplify errors.