We think of procurement software as a system to support certain kinds of workflows — the user browses suppliers, creates purchase requests, gets requests approved, generates purchase orders, and so on.
But an AI assistant could theoretically take on some of the activities performed by a person purchasing from another entity, creating a radically different paradigm.
So, would procurement software disappear in five years? Probably not.
Procurement Software Does More Than Create Purchase Orders
Enterprise-level procurement software manages a good amount of structure, and provides information that is valuable throughout the organization:
Supplier management
Approval workflows
Purchase orders
Contract and invoice management
Compliance
Audit trails
ERP integration
Financial controls
In many ways, an agent could help fulfill the same objectives, but most agents need existing information to be of value — in other words, you might not want to replace your procurement software, even if you could use agents in procurement management.
Therefore, procurement would be an interesting area in which a traditional procurement solution serves as the system of record, and AI agents serve as the system of action.
The Power of Having the Procurement Agent Focus On Actions vs. Workflows
Traditional procurement software makes sense — you define various kinds of objects and their values, approvals, and the like.
An agent could likely go more directly after a goal.
Here's a simple example where the goal is an outcome:
Assume we get the input "We need enough packaging material for the next six weeks while minimizing costs".
The agent could then potentially follow a set of steps, each of which represents some kind of action:
Determine inventory level
Calculate utilization based on past performance
Get production forecasts to estimate future requirements
Anticipate requirements based on known variables (demand forecasts)
Determine relevant suppliers based on criteria such as minimum order size, lead time, and price
Estimate cost implications
Determine risk based on available capacity at various suppliers
Evaluate options and make recommendation
Create purchase request for approval
Compare this to the alternative where a human being determines all this.
Yes, that agent has a lot of value.
And, crucially, it took data, performed analysis, executed actions and — ultimately — recommended a purchase to leadership.
In effect, you off-load a bunch of the busywork that traditionally falls to procurement departments while empowering your human purchasers to make higher-level decisions.
But, there's another key advantage.
Integrating the procurement agent into the wider organization
An intelligent agent has a real opportunity to make better decisions if it can integrate and analyze data from related systems within the organization.
Those might be systems related to the procurement process such as ERP systems, inventory management systems, supplier databases or contract and supplier management repositories and systems supporting core production processes.
IoT data from physical production facilities might also provide valuable inputs: An AI analyzing that kind of data could determine that production throughput levels are higher than expected, and therefore recommend increasing materials orders.
Intelligence from an AI-assisted, IoT-connected enterprise system could create a feedback loop with the procurement agent, thereby enabling it make smarter and smarter recommendations.
For example:
IoT sensors flags accelerated rate of material usage -> AI detects increased risk of inventory shortfall -> procurement agent identifies alternative material suppliers -> human approver approves increased orders to supplement existing inventory.
Why procurement software won't disappear
There are several possible reasons why a procurement agent would never completely replace existing procurement software.
Those include:
Governance: Large organizations benefit from clear approval processes.
Accountability: Someone needs to be responsible for large purchases.
Compliance reasons: Certain industries have well-established processes around procurement, and may not want to change.
Data quality: The quality of an AI's results depends on the training data, which may include existing procurement software.
Security: An independent agent may be able to make purchases and recommendations, but you would still need someone to manage it, negotiate with suppliers, sign contracts and the like.
As a result, human purchase managers probably have a long future in front of them — particularly when it comes to strategic sourcing.
What procurement departments might look like in five years
Procurement software for the next five years will likely enable the transformation from a dashboard of views to an enterprise intelligent environment.
Rather than browsing purchase orders, requests, approvals and the like, procurement managers and purchasers will be able to rely on AI agents to perform many mundane or time-consuming tasks such as risk analysis, and to support humans in making complex purchase recommendations based on factors that span the entire enterprise.
There will be less of a need to constantly review, analyze and consume data regarding purchase orders and the like, since the AI can do those activities in real-time.
Meanwhile, the humans can focus on the larger picture.
Would AI agents replace procurement software? A closing thought
We think the answer to the question is "no" — not really.
An AI procurement assistant might help transform and optimize the work and tools currently supporting procurement management, but the underlying system of record is likely to remain the procurement software.
But maybe an interesting combination of enterprise data + traditional procurement software + operational data + AI agents + human management is what's really going to win in procurement over the next decade.
Those exploring how AI connects with physical systems and enterprise data in general are already looking ahead to this grand AI+IoT future — you can learn more about how we're approaching that at Aperture Venture Studio.
Top comments (0)