Intel Corp., whose components can be found in countless high-tech products and systems the world round, never had a simple supply chain. But exponential growth in markets and applications, coupled with intensifying customer demands, has brought the company’s procurement, manufacturing and fulfillment network to an unprecedented level of complexity.

With annual indirect-materials spend of $58 billion, Intel was finding it increasingly difficult to assess the terms and conditions that made up more than 15,000 “legacy” contracts with 650 suppliers. Manual processes were no longer sufficient to get the job done.

The solution lay in adoption of a concept that is foremost on the minds of every global supply-chain executive today: digitization. Intel would proceed to embrace multiple aspects of modern-day automation, including machine learning, natural language processing, and robotic process automation, with the eventual goal of evaluating risks for more than 100,000 contract documents for indirect spend.

Traditional manual techniques for auditing contracts were slowing efforts to a crawl. Merely finding the right supplier in the contract database would take five minutes. Then five to 20 minutes more to search for all relevant folders, files and documents. Then up to an hour to score all the provisions, by way of a 120-question template. Finally, another five minutes for adding and compiling the scores into the master dashboard. Rinse, wash and repeat. With each assessment taking upwards of 90 minutes, Intel’s Indirect Materials team only had the time and resources to audit its top 80 suppliers.

Intel’s first step toward a solution involved posing a basic question: Could it come from an outside solutions provider, or did it need to be fashioned in-house? The former option would seem preferable, given the relative ease of applying proven techniques that were already on the market. In fact, Intel approached more than 10 outside providers of contract analytics, only to find that none of them had anything close to a solution.

Four-Step Process

In-house development was clearly the way forward, through a concerted effort by Intel’s Supply Chain Intelligence and Analytics (SCIA) team that would require four distinct steps:

• Data extraction, downloading thousands of contracts from the central repository in minutes with the help of RPA, and digitizing them through optical character recognition (OCR) in order to standardize the information and make it searchable;
• Data integration, in which raw data from the OCR process would be merged with metadata (essentially, data about data, such as supplier names and identification) from the central repository;
• Use of machine learning and natural-language processing to suss out risks associated with labeling data, especially of the unstructured variety, based on input from commercial and legal experts, and
• Creation of a color-coded dashboard to display the results in a meaningful way.

That dashboard gave Intel near-instant visualization of the legal risk attending each contract clause, based on more than 15 critical provisions as evaluated by machine learning. A user can click on any color-coded cell, generating a popup that provides an explanation of the color designation, including text taken from the original document as a means of validating the results. The entire document can also be downloaded in searchable form. All of which means that what previously took up to 90 minutes per supplier now is executed in seconds.

In the past, Intel would have needed to expend more than 575 hours to grade all of its indirect-materials suppliers. The digitization initiative improved process velocity by 99%. The near real-time nature of the tool effectively means that Intel can perform a 100% audit of all contracts, every day. A commodity manager can upload a new document at night, and find that the scoring has been completed automatically the next morning.

Intel estimates that it improved the strength and quality of indirect-materials contracts by 7.5%. The ability of managers to spot problematic documents — those registering on the dashboard as yellow and red — means that issues can be much more quickly resolved.

Ripple Effect

Overall savings by the Indirect Materials team is projected at $19 million a year. The result is more than a matter of cold numbers; the clarity of the dashboard and instant access to data reduce human bias when assessing the risks and quality of contracts. And that, Intel points out, has a ripple effect when it comes to negotiating with suppliers. The whole exercise becomes markedly more consistent and predictable, Intel says. What’s more, the company is now able to strategically allocate business to those suppliers that represent higher contract strength, while deploying risk-mitigation efforts for pursuing new and expanded procurement opportunities.

The joint effort between Indirect Materials and SCIA teams ended up garnering recognition both internally and externally. The principals were invited to present at the company’s annual statistics summit and Supply Chain Technical Leadership Forum. In addition, SCIA found itself recommended by an outside consultant, when the latter was asked by Intel’s Sales and Marketing Group about who could perform analytics on customer contracts.

Intel says it found no outside provider that could match the level of analysis of contract clauses — coded green, yellow and red — that it has achieved for all contracts within its repository. For the company, that constitutes a competitive advantage. But Intel isn’t stopping there: it plans to expand the system to its entire global supply-chain organization, representing $35 billion in spend and 100,000 documents. It already has 20 use cases on its platform roadmap, and is currently coaching data scientists in the Sales and Marketing Group on how to apply the technology to customer contracts.

For those companies wishing to follow a similar path, Intel urges them to define the business problem first, before settling on a particular type of technology to solve it. Equally essential, the company says, is the initial “make-versus-buy” determination. While not possible in this particular case, an external solution might be available that would save money and lead to faster results.

In a broader sense, Intel views the indirect-materials project as key to the dream of an “autonomous digital supply chain.” It likens the effort to switching out paper maps for automated navigation systems.

“We asked ourselves, ‘Why can’t we do this within our supply chain?’” Intel says in its submission to the Supply Chain Innovation Awards competition. “If we can digitally connect all these ‘silo’ functions, we will enable the path and be on our way.”