M4P: the AI agent suite for Procurement

Imagine a platform where every stage of the Procure to Pay (P2P) process—every document and every decision—is supported and enhanced by Artificial Intelligence. This platform exists: it's called Mashfrog for Procurement. Born from the synergy between Mashfrog's expertise in procurement and GenAI/XAI, M4P transforms every phase of the P2P process, making operations faster, more accurate, and smarter. We discuss this with Veronica Rao, Managing Director of the Spend Management area and Product Manager of M4P.

Veronica, in a few words, what is Mashfrog for Procurement?

Mashfrog for Procurement is the fastest, most accurate, and up-to-date colleague a buyer could have. It’s a suite of multi-agent artificial intelligence components specialized in procurement language, integrated with spend management systems (Ariba, Jaggaer), infoproviders, and SAP S/4HANA. It helps companies accelerate purchasing processes, improve decision-making, and reduce operating costs.

What makes M4P different from a traditional automation tool?

It’s not just a tool that automates repetitive tasks, but an intelligent extension of the procurement team. Our AI agents understand natural language requests, analyze data from multiple enterprise sources, and deliver ready-to-act responses—drastically reducing processing times and minimizing the risk of errors.

What are the main components that make up the suite?

• ProcAssistant AI: A module that, through interaction between multiple agents, answers complex questions by aggregating data from Ariba, S/4HANA, and external infoproviders.
• SmartClause AI: An advanced component that analyzes contracts and clauses, automatically comparing them with the company’s clause library.
• TenderAI Assistant: A vertical solution composed of multiple agents dedicated to drafting tender specifications and building technical evaluation grids.
• Vendor Validator AI: Performs in-depth checks on supplier information by cross-referencing corporate, financial, and reputational data.
• DocSum AI: A solution that transforms complex documents into clear, actionable summaries.

Can you share a real-world example of how ProcAssistant AI has accelerated a process?

Imagine a buyer who needs to answer the question: "Which suppliers in category X worked with us last year, have an Open.es score above Y, and at least 100 employees?" Normally, this would require manually extracting and cross-referencing data from different systems. With ProcAssistant AI, all it takes is to ask the question. In seconds, the application aggregates data from Ariba, S/4HANA, and external infoproviders, delivering a complete answer—ready to support purchasing decisions.

Companies often underestimate how much time they spend managing tenders. Can you tell us how TenderAI Assistant truly transforms this process?

Absolutely. Tenders are complex processes, often slowed down by manual tasks that consume valuable time. With TenderAI Assistant, the generation of RFP documents and the technical evaluation grid is supported by artificial intelligence, ensuring information is complete, consistent, and error-free.
But the real game-changer happens when the bids come in: the application, thanks to its multiple agents, automatically analyzes the submitted documents, extracts relevant data, identifies non-compliances, and provides a detailed comparison of proposals based on the generated evaluation grid.
One of our clients reduced the time spent analyzing bids from three weeks to just three days—freeing up time and energy to focus on the strategic decision of selecting the best supplier.

The application of AI to procurement can raise concerns about data security. How do you address this issue?

Data security is a top priority. The M4P suite clearly separates the data management component from the processing components. Data can be stored within a private infrastructure—either cloud-based or on-premises—managed directly by the client. The processing components run on Mashfrog’s centralized cloud infrastructure, while still ensuring the required security levels and a clear separation for each client, thanks to Docker-based container technology. Additionally, any calls to commercial LLMs are made with explicit instructions not to use the data for model training, and sensitive data is masked when necessary—fully complying with corporate data governance policies and GDPR regulations.

What is the reaction of procurement teams when they start using M4P?

At first, there's surprise—they don't expect such fast and complete responses. Then they realize that AI isn't replacing them, but empowering them—freeing them from manual tasks so they can focus on strategic decisions and high-value negotiations.

How is M4P implemented within companies?

Through a model we call "50Days", because in less than three months companies can start benefiting from intelligent automation in procurement. It’s a fast and structured process that enables the activation of AI multi-agent components already tailored to the client’s processes and data—ensuring a quick and secure launch with support from our dedicated team.

In this way, artificial intelligence becomes a daily operational resource, accelerating procurement processes and freeing up valuable time for strategic decision-making.

In which direction is AI in procurement heading, in your view?

AI will become a standard—not just to save time, but to improve decision quality and reduce risks. With M4P, our goal is to make procurement faster, more transparent, and more sustainable, by equipping people with tools that help them make better decisions.

In three words, what does M4P represent for procurement professionals?

Speed, accuracy, security.