By: Andrew McKinzie, Tower Research Ventures | February 24, 2025

Framing the Opportunity

Artificial intelligence (AI) enables businesses to take a fresh perspective on how and what they procure.

It’s difficult to imagine a business that does not care about what they buy. A manufacturer counts on the correct, reliable component to manufacture their product. Without it, the product does not work. They care a lot about the components they procure. Many times, the end customer cares too. To use a simple example, you might notice when your favorite restaurant switches a key ingredient supplier for your favorite dish, or when your preferred hotel chain changes its pillows – these items matter to you, as they influence your experience of the meal and the stay. The restaurant or hotel might even lose you as a loyal customer following a supplier change.

And yet, many businesses of different scales view procurement as a non-strategic service to the business in pursuit of higher gross margins. Procurement motions can seem like a race to the bottom on price across suppliers. A small business may prioritize cheaper suppliers to proactively manage their cash needs, a mid-size business looking to grow may prioritize trimming supplies to free up additional cash to reinvest in growth, and a large business may aim to consolidate suppliers and leverage its scale to exercise pricing power. Keeping margins in check is important across businesses.

The AI Opportunity

As we think about the evolution of procurement technology pre-AI, major developments focused on business communication, data unification, and matching suppliers and buyers. Electronic Data Interchange (EDI) enabled more efficient communication with suppliers. Material Requirement Planning systems (MRPs) synced production schedules with procurement schedules. Enterprise Resource Planning systems (ERPs) brought other business departments into the fold, creating a system of record and master source of truth for transactions within a company. The internet enabled B2B marketplace platforms for suppliers and buyers with product catalogs and RFQs. SaaS cloud-based solutions reduced the cost of tech implementation and focused on dedicated procurement tools.

The AI opportunity leverages that data and those systems to actually do the procurement transaction workflows. If the ERP infrastructure developed over the last four decades were thought of as a car chassis, AI systems today can act as the car engine, moving procurement processes forward and performing the work procurement teams do.

A Brief Overview of The Procurement Workflow

The work procurement teams do focuses on identifying a business’s needs and then finding, engaging, and managing the suppliers who can meet those needs.

Needs Assessment

The first step is for the procurement team to communicate with various business unit leads to determine what goods and services they need, when they are needed, and the budget available for them. This is referred to as a “needs assessment” or “requisition process,” and ultimately results in a document that outlines details like specifications, quantities, and timelines.

Supplier Discovery, Validation & Selection

Procurement teams take this form and engage in a process to discover potentially relevant suppliers, validate those suppliers could be a potential fit, and ultimately select a supplier. Depending on the size of a procurement need, procurement teams might decide to bid the contract out to various suppliers, in which case the process might formalize into a “request for proposal” process. Procurement teams will then evaluate suppliers based on price, quality, capability, reputation, and availability among a host of factors important to the business. This process involves lots of communication and back-and-forth between the procurement team and suppliers.

Supplier Management

Once a supplier is approved and signed-on, the procurement team is then responsible for managing the relationship with that supplier, including regular communications regarding shipments, inspections, and any problems and potential resolutions.

Supplier Payment & Accounting

Managing the day-to-day procurement needs of the business across all suppliers can involve many tasks involved with data entry, including raising purchase orders, reconciling invoices, checking to make sure orders are fulfilled according to the supplier’s contracts, and communicating across the business to make sure various departments are satisfied. There can be frequent discussions between the procurement team and the accounts payable and accounting departments to ensure suppliers are paid according to payment terms and to ensure the general ledger of the business accurately reflects which items were procured and paid for according to the company’s expense classifications.

These procurement workflow steps are outlined below:

• Needs Assessment
• Supplier Discovery
• Supplier Validation
• Supplier RFP and Selection
• Supplier Onboarding and Integration
• Supplier Communication
• Purchasing
• Approvals, Limits, Permissions
• Product Inspections
• Receiving
• 3-Way Match (PO, Invoice, Proof of Receipt)
• Documentation and System of Record Update
• Closing Out Purchase Orders
• Payment Execution and Working Capital Management
• Updating Inventory Balances
• General Ledger Accounting Entries

Practical AI Applications in the Workflow

Summary

Across each step of the procurement workflow, there are opportunities to use AI to streamline processes, augment workflows, and drive efficiency. AI can be used to find and communicate with suppliers, price compare items across suppliers and website domains through LLM-based search, negotiate with suppliers, summarize contracts and suggest contract language improvements, access systems of record to analyze supplier performance, process purchase orders and invoices in the accounts payable process, automate ERP data entry, analyze enormous sums of data, and generate sophisticated analyses and forecasts from it, among other use cases.

Detail

Procurement teams during the requisition and sourcing process often find suppliers can be slow to react and negotiate with, prior to entering decision-making approval workflows. This back-and-forth communication can delay sourcing activities and increase procurement cycle times. Today, large language models (LLMs) can automate the requisition approval process by using data-driven rules, and LLM chat bots with guardrails can generate communications with suppliers, recommend price negotiations, and even execute contracts autonomously in some cases.

Once suppliers are vetted and signed, procurement teams must manage the contract lifecycle. Contracts often live in long PDFs and are rarely revisited or analyzed until renewal. Traditional procurement systems used to file and store contracts lack the ability to scan these documents and extract the key terms a human would read for, such as specific amounts, dates, renewal opportunities, and nuance in the fine print. Teams must spend hours combing through supplier contracts to pull that information out today. LLMs can analyze contract terms and deadlines, automate reminders for renewals, and even assist in drafting new contracts. LLMs can also flag risks hidden within contract language or terms that could negatively affect a company’s supply chain or cost structure.

AI can:

• Requisition & Sourcing – Generate communications with suppliers, recommend price negotiations, execute contracts autonomously in some cases
  
  
• Contract Lifecycle – Analyze contract terms and deadlines, automate reminders for renewals, assist in drafting new contracts, flag hidden risks
  
  
• Supplier Management – Analyze supplier contracts, extract key rebate program terms so procurement teams can optimize their spending with the supplier

Procurement teams have traditionally struggled with effectively managing every supplier relationship, tracking supplier performance, product availability and back stock, tracking spend against rebate programs, and ensuring compliance with regulations. These are text-heavy, manual, paper-based workflows, and are often done reactively when supplier disruptions are identified. LLMs can quickly help analyze supplier contracts for items like quantity of product promised, search through them to identify whether a supplier is responsible for a specific good experiencing a shortage, and extract key rebate program terms so procurement teams can optimize their spending with the supplier. For example, during COVID-era supply stockouts, many retailers of goods such as toilet paper and baby formula were forced to explore alternative suppliers. LLM-based systems could have enabled procurement teams to process invoices for those items from existing suppliers, quickly reconcile them against the quantities specified in the supplier contracts, and proactively identify potential stockouts before they occurred. This would enable procurement teams to stay in front of the issue while exploring alternative suppliers. LLMs could also enable procurement teams to parse supplier qualification documents for supplies such as masks and swabs and reconcile them against safety standards documentation to ensure the supplies met CDC standards prior to purchasing them. The result of this LLM-enabled functionality is a procurement team able to quickly process documents, respond proactively to supply shocks, and act as a strategic partner to the business.

Despite the document organization that current procurement systems provide, they still require significant manual data entry. Entire teams are tasked with uploading purchase orders, invoices, supplier contracts, and supporting details in systems and manually entering and checking information. This information can be distributed across emails, WhatsApp, paper notes, sticky notes, handwritten records, and Excel sheets, among other locations. Documenting this information for supplier relationships in a system is a genuine challenge to do efficiently, with significant risk of human error. However, AI-based systems can be trained to automate these repetitive tasks. OCR and computer vision models can be used to extract key information from invoices and take unstructured data and parse and structure it into ERP fields, while LLMs can generate purchase orders with prompts and input data into systems with instructions. They can read and process supplier emails, invoices, and contracts and populate ERPs with information residing in unstructured formats. The work it takes to enter information into ERPs and maintain it in an accurate and timely way can be done in a fraction of the time. As a result, businesses can benefit from a more fluid, up-to-date picture of their procure to pay workflow.

ERPs are designed to connect information across the departments of a company, but often that data resides in silos. Procurement, Finance, Operations, and other departments may store and manage their data separately across drives, folders, and systems. The dispersion of data and different data formats can make it difficult to process, analyze, and transform into actionable information. Today, manually stitching together this fragmented data to arrive at actionable information can take a long time, which limits the potential for real-time visibility into company performance and procurement needs. However, LLMs can help effectively parse this data in different formats and locations and unify it into a centralized data lake so that the data is usable for AI-based analysis and data automations. For instance, a restaurant chain could sync data from the finance department like online order data, point-of-sale data, historical sales, and forecasted sales, with operations data like foot traffic and occupancy data in a data lake that can be queried by LLMs in order to measure which supplies and ingredients at specific locations might require more or less supply. Or, the Operations department of a retail company could relay information about an unforeseen inventory shrinkage event at a store location, and the company’s Procurement team could leverage LLMs to efficiently search for the suppliers of the missing inventory items that they need to contact to order the replacement items. By enabling these kinds of actions, LLMs can transform static and siloed data in an ERP into a live, unified, and insight-rich action system.

AI can:

• Data Entry – Read and process supplier emails, invoices, and contracts, and populate ERPs with information residing in unstructured formats
  
  
• Data Analysis – Transform static and siloed data in an ERP into a live, unified, and insight-rich action system
  
  
• Price Shopping – Extract categorical-level detail buried in item descriptions in supplier catalogs and identify identical item substitutes with respective prices across suppliers
  
  
• Forecasting – Analyze seasonal trends and produce accurate and dynamic forecasts

While ERP and procurement systems offer basic spend tracking and controls, they often lack advanced analytics to optimize savings opportunities across categories and suppliers and lack the control to prevent specific ingredients or items from being procured. LLM-powered procurement platforms can enable powerful price comparison shopping by extracting category-level detail buried in item descriptions across supplier catalogs to identify identical item substitutes with their respective prices across suppliers. For example, an automotive servicing chain could use LLM-based search to identify the same lubricant if currently uses, but from a new supplier for a much cheaper price. Or, a medical clinic could identify cheaper substitutes for soft goods like gauze, cotton balls, and ointments from alternative suppliers they have not previously procured those items from. Companies can also institute semantic rules-based systems to prevent procurement teams from searching for or procuring certain types of items that are off-limits. Maybe the same medical clinic wants to avoid latex products. The clinic procurement team could specify “no latex” and an LLM-based search system would systematically exclude items containing latex from search results across all types of products.

Businesses rarely if ever trust the business forecasts generated by ERP and procurement systems and instead build independent financial forecasts. This is because the forecasting functionality in ERPs relies on static data – especially for inventory – which can result in over-stocking or under-stocking supplies. By contrast, machine learning models can incorporate real-time data with respect to historical data in an ERP and produce more sophisticated forecasts. For example, businesses that must respond to seasonality can take into account the current season’s trends, analyze them with respect to prior season trends, and produce more accurate and dynamic needs assessments. One such case in the medical field relates to stocking certain types of seasonal allergy medications.While a doctor knows that in the spring they might typically stock up on certain allergy medications to meet demand for pollen-related allergies, they likely don’t have a strong sense of the exact timing and don’t want to mistakenly order too early or too late. If they order too early, they bear the inventory carrying cost of unused drugs. If they order too late, some medicines may be backordered due to demand spikes. Machine learning models can learn the clinic’s seasonal demand and offer precise predictions for when the clinic should order inventory to best fit the needs of the business.

Conclusion

There is an opportunity to improve the procurement process across the business landscape. From automating internal and external communications, to augmenting supplier search and negotiation, to extracting insights from burdensome contracts, to even improving forecasting capabilities, artificial intelligence can enable more efficiency and better outcomes.

Given the ocean of possibilities, how should one decide what workflow to innovate, which product to develop, which industries to build for, and what business model to use? Tower Research Ventures has developed insights and frameworks surrounding these questions from hundreds of conversations with industry experts and entrepreneurs. If you are building in this space, we’d love to chat. Please reach out to Tower Research Ventures at ventures@tower-research.com.

Some of TRV’s procurement-related investments to-date:

About Procurement Sciences

Procurement Sciences AI helps government contractors discover the right contracts for their business with advanced AI tools that match opportunities based on unique capabilities and past successes. The platform enables customers to make informed decisions with AI-powered opportunity analysis, offers bid/no-bid calculators, and agency/competitor analysis. It streamlines the RFP/proposal process with AI tools that guide customers in creating compelling win strategies, compliant proposals, and effective risk assessments, enhances proposals with AI tools that review and refine drafts, perform compliance checks, and provide score enhancement recommendations. Post-proposal, Procurement Sciences helps customers learn and grow with post-award AI tools that analyze debriefs, evaluate protest grounds, and track key metrics, trends, and patterns.

About Mentium

Mentium is an AI-powered payment automation platform for shippers, brokers, and factoring companies. Their technology captures and centralizes documents across multiple channels: emails, ERPs, TMSs, EDIs, etc., understands documents and structures data using a proprietary large language model, automates manual workflows including audit, data entry, sending alerts, requesting approvals, etc., and features self learning to continuously improve accuracy and automation based on user actions and data. The platform also unlocks insights, identifying anomalies, payment patterns, and payment-related issues.

About Assembli

Assembli is developing AI-native products that enhance the efficiency and profitability of construction businesses.

---

The views expressed herein are solely the views of the author(s), are as of the date they were originally posted, and are not necessarily the views of Tower Research Ventures LLC, or any of its affiliates. They are not intended to provide, and should not be relied upon for, investment advice, nor is any information herein any offer to buy or sell any security or intended as the basis for the purchase or sale of any investment. The information herein has not been and will not be updated or otherwise revised to reflect information that subsequently becomes available, or circumstances existing or changes occurring after the date of preparation. Certain information contained herein is based on published and unpublished sources. The information has not been independently verified by TRV or its representatives, and the accuracy or completeness of such information is not guaranteed. Your linking to or use of any third-party websites is at your own risk. Tower Research Ventures disclaims any responsibility for the products or services offered or the information contained on any third-party websites.