Transforming theory into practice

Generative artificial intelligence (AI) is revolutionizing the nature of work across almost every industry, around the globe. We see it in creative work, where large language and image AI models generate automated content. In manufacturing, generative AI augments human-based product design efforts, optimizing key processes, and improving quality control. It’s changing how healthcare is delivered, improving patient outcomes through better diagnosis and information sharing.

To better understand AI’s role in transforming the procurement landscape, Arroyo Strategy and our AI partner, dSilo, present our co-authored Insights series, “The Future of AI in Procurement.” In our previous blog in this series, we explored how companies are using “passive” AI technology with large language models (LLMs) to deliver new insights from their data and act upon these insights. Rather than using the power of generative AI to take automated action on these insights, it’s left to human intervention to capture the value–when teams are already too busy. In the third installment of this series, ‘Transforming theory into practice,’ we discuss how game-changing AI can connect strategic and tactical procurement processes while reducing manual activities significantly. We delve into the transformational potential of AI in procurement, and the unique traits making it a prime use case for AI adoption in our industry, to an extent that previously hasn’t been possible.

As you might expect, we’re seeing real imagination on the part of the early adopters as they seek to tap into the generative capabilities of AI. As we noted previously, generative AI isn’t fundamentally changing the procurement process, but it is closing gaps and improving efficiency and effectiveness in new ways. These improvements aren’t limited to procurement; we have clients that are applying these capabilities into other areas. One of these is in commercial contracting, where clients are looking to accelerate the sales cycle while also providing insights that can enhance revenue.

In this blog, we’re also going to share use cases for generative AI that might spur similar efforts at your organization. Here are some of the most exciting examples we have seen (and supported) thus far, along with innovative ideas for the near future:

When the company is transacting on the supplier’s paper (or, from the sales side, on the client’s paper), the first step is for legal to review and to redline the document to the point that the language is within the bounds of acceptability. To date, it’s been a manual exercise, consuming the time of an expensive resource, and it delays the execution of the agreement.

Now, LLMs are taking aim at such inefficiency. They are trained to compare the proposed language with the company’s own language and to propose edits that will meet their own requirements. For now, this still needs review before returning to the supplier/client, but it’s a much more efficient way of managing this task. As the LLMs continue to learn, the reviews will be less time consuming and will become unnecessary.

A further idea proposed by one customer is that for repetitive issues like limitation of liability or indemnification, the tool can also capture how differences have been resolved in prior contracting negotiations for the reviewer to select from, rather than relying on memory (especially if there are multiple lawyers involved in the review process, or a change in personnel).

One of the clients that Arroyo Strategy worked with used this LLM capability to completely restructure the clause library for optimizing fallback positions and negotiations strategy. They started with simple P (Price) *Q (Quantity) contracts where market conditions, previous contract negotiations, and risk parameters were limited. Over the period extended, this use case was leveraged in complex, labor-based contracts (with different risk profile negotiations) and, after twelve months of training the platform, they are using this capability for even the most complex and variable clause management. This process was an organic and human-at-center model to keep making the engine outcome stronger for a year and now it’s delivering results on providing the recommendations much more autonomously.

Stakeholders collaborating with suppliers need to understand aspects of the contract for their day-to-day responsibilities. But they don’t want to have to plow through tens or hundreds of pages to find the information that’s pertinent to their job. In some cases, you also don’t want to share access to confidential information such as pricing.

Generative AI can create contract summaries that pull out the specified operational information required. It can also generate different summaries depending on the role of the user. As a result, A/P only sees what it needs to, while maintenance’s view is similarly customized.

Consider the example of a railway operator that maintains a fleet of engines and cars and has uptime metrics currently focused on repair and return to service. If their maintenance crews gain easy access to information on warranties in supplier contracts, they can quickly decide whether to repair an item, or replace it and send the defective item back for repair under warranty.

One of Arroyo Strategy’s clients is using this capability to manage internal firewalls between various groups and managing conflict of interests between stakeholder, legal, risk, and procurement teams. This capability made the information sharing significantly more streamlined and avoided any version control issues and possible gaming issues.

The increasing number of regulations that require terms in contracts is a challenge both for legal and risk management, especially for evergreen contracts that have been rolling forward for years or newer contracts that didn’t anticipate these new rules. These required clauses are not a part of the meta data typically captured in a CLM system, so risk management is essentially blind as to whether clauses for GDPR, DORA, Prop 65, Dodd-Frank, LKSG, CSRD and more are in the contracts or not.

In the past, knowing whether these risk management clauses were in place was an impossible task to perform manually. Now, LLMs can review all contracts and identify where these clauses should be present but are absent. Generative AI can then produce a draft amendment that adds the necessary language and either send it to the appropriate contract administrator for review, or even send it directly to the supplier.

At Arroyo Strategy, we are seeing a lot of point solution approaches with our customers to solve this problem, considering CLM product roadmaps many times don’t allow a custom solution. We have worked with some of our clients to build the unique, one-time utility to take care of historical information using technology and using Gen AI to keep it learning and maintained.

In our last blog we talked about the ability to use LLMs to reconcile contracts to invoices and flag discrepancies in pricing or terms, or with the application of discounts and rebates. With passive AI, these discrepancies are flagged, but it’s still left to the A/P clerk to take action to resolve these discrepancies. One customer is now developing the capability for the generative AI tool to draft an email to the supplier explaining the discrepancy and requesting a resolution, for example, by re-issuing the invoice or updating their systems to reflect correct pricing terms on the invoice.

One of Arroyo Strategy’s clients is using LLM to review around 500,000 documents to read, identify unique fields that are important for decision making, and then using that information to flag any future issues—while receiving invoices with super complex deliverables and working with the P2P systems to optimize PO structures to reduce reconciliation. The system is designed to provide recommendations at a transaction level or at a function level (change management).

This is a use case driven by a sales leader, but it has relevance for procurement as well. In this instance, the company wants to understand which of its contracts allows for a customer to reduce costs by limiting the number of users during the contract period or at the renewal point. It also wants to know which contracts allow them to raise pricing during the contract and with what limitations or timing requirements. Sales’ goal is to maximize revenue, but procurement could also use this information to understand when they could reduce costs within the contract terms.

The list of use cases is by no means exhaustive. It continues to expand by the day as organizations become more familiar with generative AI and consider how to deploy it to address their most pressing needs or issues. As awareness grows, leaders are seeing how their organizations will reap the competitive benefits of its enhanced efficiency and effectiveness.

If your organization is already working with passive AI, it’s time to think about how the generative tools can take on rote, repetitive tasks and free up your resources for the higher value-add work that’s being done far too infrequently today.

In the next installment of our “Future of AI in Procurement” Insights series, we’ll look at the human implications of generative AI and how procurement professionals can adapt. We’ll continue to explore the transformative role of AI in elevating procurement capabilities, with a focus on how Arroyo Strategy and dSilo are currently implementing these practices to take their clients’ use of game-changing generative AI to the next level.