The Evolution of Agentic AI in Procurement

Procurement has always involved more than just buying something.

A single request triggers supplier research, approvals, negotiations, contract reviews, compliance checks and ongoing supplier management across different teams and systems.

Automation has taken on parts of this process through digital workflows and automation.

AI then made it possible to analyse information, identify patterns and support procurement decisions.

Now, agentic AI in procurement is taking the next step: AI systems can work towards a defined goal, coordinate multiple tasks and take authorised actions rather than simply responding to individual instructions.

That changes what procurement technology can do.

What is agentic AI in procurement?

Agentic AI is a form of artificial intelligence that can take over certain procurement tasks with a degree of autonomy such as evaluating suppliers, routing purchase requests, or preparing renewals.

Instead of waiting for instructions for every individual task, an AI agent can interpret what you are trying to achieve, determine the steps needed, use relevant data and systems, and take authorised actions along the way.

A request for new software typically involves several steps, like understanding the requirement, checking existing suppliers and contracts, assessing procurement policy, obtaining approvals, comparing options, negotiating terms and tracking the agreement afterwards. Agents will be able to take action across every steps without necessarily requiring constant intervention.

An AI assistant can help with individual parts of that process.

You could ask it to summarise a supplier proposal, compare contracts or draft a negotiation email.

An AI agent goes further by connecting those tasks.

Given a defined objective, it can gather the relevant information, determine what needs to happen next and coordinate the process, while bringing you in when a decision requires human judgement or approval.

For an AI system to be considered agentic, it generally needs several capabilities working together: it must be able to understand a goal, reason about the steps required, access relevant context and tools, take actions, and respond when circumstances change.

In procurement, those actions can include routing requests, checking supplier or contract information, gathering missing data, preparing sourcing activities or flagging an upcoming renewal.

So, agentic AI in procurement means using AI agents to plan, coordinate and execute parts of procurement processes towards a defined business outcome.

Different technologies have their own tasks within the procurement process:

Agentic AI vs traditional procurement automation

The important distinction is between following a process and working towards an outcome.

Traditional automation is most effective when you already know what should happen. You define the rules and sequence, and the software follows them consistently.

For example, a procurement workflow might automatically send a purchase request above a certain value to a specific approver.

Agentic AI can work with more variable situations. Instead of simply following a fixed sequence, it can assess the context and determine which actions are relevant to the goal.

For example, rather than simply reminding you that a contract expires in 90 days, an agent could review the contract, examine relevant supplier and spend information, identify what needs attention and prepare the renewal process for the appropriate stakeholders.

Some other differences include:

How has AI evolved in procurement?

The evolution of AI in procurement is really the story of how much of the procurement process technology can take on.

What began as a largely manual, people-led process has gradually become more connected, automated and capable of acting on its own.

In traditional procurement, most of the work happened between people. An employee raised a request, procurement gathered the requirements, buyers searched for suppliers and quotes, stakeholders exchanged information by email, and approvals were handled separately.

Contracts and renewal dates often needed to be monitored manually, while reporting meant bringing information together from different sources.

The first change was digitalisation.

Procurement software brought purchase requests, approvals, supplier records, purchase orders, contracts and spend data into structured systems. Essentially, all teams could work from a shared source of information.

The next shift was automation.

Routine steps could happen without someone manually initiating them: a request could be routed to the right approver, a notification could be triggered when a contract approached renewal, or information could move between systems according to predefined rules.

Procurement became faster and more consistent, but people still had to determine what should happen when a situation fell outside those rules.

AI began changing that by making procurement systems better at understanding and analysing information. Instead of simply storing a contract, for example, AI could extract its key terms. Instead of showing a list of suppliers, it could help compare them. Spend could be classified automatically, risks could be flagged and large volumes of procurement information could be analysed.

The latest shift is towards agentic AI, where the technology can take a broader objective and help carry the process through multiple stages. An agent can potentially work through the information needed to prepare that renewal and determine what needs to happen next.

For example, given the goal of renewing a SaaS contract while reducing cost and maintaining the functionality your business needs, an agentic system could:

1. Identify the contract and renewal deadline
2. Analyse current spend and licence usage
3. Review previous negotiations and supplier information
4. Check the existing contract and relevant terms
5. Identify potential savings opportunities
6. Gather the information needed for the renewal
7. Recommend a negotiation strategy
8. Route the renewal to the right stakeholders
9. Support supplier interactions
10. Track progress and flag anything that needs attention
11. Escalate important decisions or request human approval where necessary

Each stage has moved more of the process from manual execution towards intelligent assistance and, increasingly, authorised action.

However, let’s not forget that procurement professionals are still responsible for the decisions that require commercial judgement. What changes is how much of the work surrounding those decisions can be handled by technology.

How agentic AI impacts procurement ROI

The value of agentic AI in procurement is not limited to how many tasks it can automate. Its impact on ROI comes from improving what procurement teams do with their time, spend, information and capacity to act.

We can see these outcomes across four areas:

1. Time savings

Procurement teams often spend considerable time on coordination: gathering information, chasing approvals, comparing documents, preparing renewals and following up with stakeholders.

Agentic AI can take on more of this work, helping you:

• reduce manual administrative tasks
• shorten sourcing cycles
• spend less time chasing approvals and information
• review contracts and supplier information faster
• reduce the time needed to prepare for negotiations

The ROI comes from putting those hours back into the procurement team’s capacity rather than simply making individual tasks faster.

2. Cost savings

Better procurement decisions can have a direct effect on spend.

By bringing more information into the decision-making process, agentic AI can help identify opportunities that are difficult to spot when data is scattered across contracts, suppliers and purchasing systems.

This can include:

• stronger negotiation preparation
• more informed supplier comparisons
• identifying unused or unnecessary software licences
• reducing purchases outside preferred suppliers
• identifying opportunities to consolidate suppliers
• starting renewal discussions before deadlines limit your negotiating position

For example, understanding licence usage before a SaaS renewal gives you a stronger basis for deciding what you actually need before entering a negotiation.

3. Risk reduction

Not all procurement ROI appears as immediate savings. Avoiding unnecessary cost or risk can be just as valuable.

Agentic AI can help you maintain greater visibility over:

• upcoming contract renewals
• procurement policy compliance
• contract terms and obligations
• supplier risks and performance
• incomplete or stalled procurement processes

Instead of relying on someone to remember what needs attention, relevant information can be surfaced when action is needed. This is particularly valuable as your supplier base, contract portfolio and purchasing volume grow.

4. Strategic capacity

Perhaps the biggest long-term impact is what your procurement team can do with the time it gets back.

When technology takes care of more information gathering, coordination and routine analysis, procurement professionals can spend more time on negotiation, supplier relationships, category strategy, risk management and decisions that create value for the wider business.

That makes the ROI of agentic AI broader than simply reducing the cost of a procurement task. It can increase the amount of strategic work your procurement team is able to take on without increasing its workload at the same rate.

Real results with Najar

Agentic AI is still an emerging part of procurement technology, so Najar Agent is en route.

However, Najar is already using AI across several parts of the procurement process, focusing on reducing manual work and giving you better information at the point of decision-making.

The AI Copilot for Purchase Briefs uses the company’s existing context to guide the conversation, ask for the information that is missing and build a more complete purchase brief. Procurement and other stakeholders therefore receive the context they need earlier in the process, while employees spend less time working out which fields they need to complete.

AI can also support procurement decisions once the information is in the system. Najar can assess whether contract pricing is above, below or in line with market benchmarks.

AI Contract Diagnosis adds another layer during contract review.

It analyses contracts to surface important issues such as unfavourable terms, missing clauses and potential compliance gaps, giving teams a faster way to identify what needs attention before a decision is made.

Najar also uses OCR technology to extract information from documents automatically.

Najar’s existing customer stories show what can happen when procurement processes become more structured, connected and easier to manage.

Vusion is a useful example.

As the company grew, its procurement team was dealing with manual sourcing, purchases made outside the procurement process and limited visibility into pricing benchmarks. Internal teams, particularly IT, were also handling supplier discussions alongside their other responsibilities.

Najar helped Vusion bring purchases and renewals into a more structured process, improve visibility across its SaaS tools and provide access to pricing insights and negotiation support, plus:

• 10x ROI in one year
• 25% average savings per negotiation
• 25% reduction in internal procurement time
• €70,000 saved on a security software renewal worth around €200,000
• 25% cost savings on overall IT spend

Vusion’s experience also illustrates why agentic AI is the next logical development.

Once procurement data, processes and decision points are connected, AI has more context to work with. The opportunity is then to move beyond supporting individual tasks towards coordinating more of the work required to reach a procurement outcome.

How to implement agentic AI into procurement

Introducing agentic AI into procurement is not as simple as connecting an AI model to your purchasing workflow.

An agent needs reliable data, access to the right systems, a clear understanding of your business rules and logic that works across different purchasing scenarios.

This is particularly important when an agent is expected to make recommendations rather than simply retrieve information. A useful negotiation recommendation, for example, depends on having enough pricing and contract data to establish what a reasonable deal looks like.

That becomes more difficult when you are negotiating with a new supplier and have no internal history to use as a benchmark. The same applies to purchase approvals: an agent needs to understand different budgets, approval thresholds, departments and business rules rather than applying one generic workflow.

Building that context and logic in-house can therefore become a substantial undertaking.

You need to connect data sources, maintain business rules, account for exceptions and continually test whether the agent is making appropriate decisions.

Before implementing agentic AI, you need to make sure you have the data, infrastructure and procurement expertise required to make its decisions reliable.

Start with the right procurement processes

The strongest starting points are processes that happen frequently, involve multiple steps and generate enough structured information for an AI system to work with.

These might include purchase intake, approval routing, supplier research, contract renewals or spend analysis.

However, even apparently straightforward processes can contain significant business logic.

An approval agent may need to account for different budget holders, departments, spending thresholds and procurement policies. A renewal agent may need to consider contract value, supplier performance, usage, previous negotiations and notice periods before determining what should happen next.

Starting with a defined process makes it easier to establish the data and rules an agent needs, measure its performance and identify where human intervention is still required.

Connect the data and systems an agent needs

Agentic AI is only as useful as the context available to it.

Procurement data can sit across ERP and finance systems, procurement platforms, contracts, supplier records, purchase histories and other business systems. Connecting these sources is therefore a fundamental part of making an agent useful rather than simply giving an AI model access to a collection of documents.

Pre-built connectors can reduce some of this complexity by allowing relevant financial and procurement data to feed directly into the systems using it.

Najar, for example, has a range of pre-built connectors designed to bring the relevant business data into its procurement workflows.

The quality and breadth of that data also matter. Pricing recommendations are more useful when they can be informed by real procurement and negotiation history rather than generic information.

Contract analysis becomes more valuable when the system can understand the wider commercial context. The aim is to give an agent the right information, with the right permissions, at the point where it needs to make or support a decision.

Build procurement expertise into the logic

Data alone is not enough. An effective procurement agent also needs business logic that reflects how your organisation actually buys.

That can include approval thresholds, preferred suppliers, purchasing policies, budget ownership, category-specific requirements, contract rules and exceptions. It also needs to account for situations where the usual process does not apply.

This is one reason procurement-specific AI can be difficult to replicate with a general-purpose AI tool. The value comes from the combination of procurement data, workflows, rules and domain expertise that determines how the model should operate.

Najar’s AI capabilities are supported by its procurement experience and a large body of procurement data, including more than €6 billion in spend data and over 3,500 negotiations.

This provides a broader foundation for pricing and negotiation insights than an organisation starting from its own limited purchasing history.

Keep humans in control where judgement matters

Procurement involves financial, commercial and operational consequences, so autonomy needs clear boundaries.

You might allow an agent to gather supplier information, analyse a contract or prepare a renewal, while requiring human approval for supplier selection, high-value purchases, contract approval, negotiations and exceptions.

The appropriate level of autonomy will also vary between organisations and processes.

A routine, low-value purchase may require little intervention, while a strategic supplier decision can depend on commercial context that is difficult to encode into a fixed rule.

Move gradually from assistance to autonomy

You do not need to move directly from manual procurement to fully autonomous agents. A more realistic progression is:

Assist → Recommend → Execute with approval → Execute within defined boundaries → Continuously optimise

At the first stage, AI helps your team gather and interpret information. It can then begin making recommendations, carrying out approved actions and eventually managing defined processes within carefully established parameters.

The important point is that greater autonomy should follow evidence that the underlying data, logic and controls work reliably.

Agentic procurement gives AI models enough trustworthy information, procurement expertise and business context to act appropriately and defines where human judgement remains essential.

From procurement automation to action

Procurement has evolved from manual coordination to workflow automation, AI-assisted decision-making and now increasingly agentic processes.

The next step isn’t simply automating more individual tasks. It’s connecting those tasks so procurement can move from reacting to requests to continuously managing the decisions, suppliers and contracts behind them.

Agentic AI can take on more of the work between a business need and its outcome: gathering information, analysing options, coordinating processes, monitoring progress, and taking authorised actions.

For procurement teams, that creates an opportunity to spend less time managing the process itself and more time on the decisions that require judgement, negotiation and strategic thinking.

The shift won’t happen by handing procurement over to AI overnight. The organisations that get the most value are likely to be those that connect their data and processes, define clear boundaries for AI, and gradually increase its autonomy where it can deliver measurable value.