Artificial Intelligence (AI)
Discuss current events in AI and technological innovations with Intel® employees
412 Discussions

Transforming Intel with AI

5 0 3,543

I remember when Intel’s artificial intelligence (AI) journey started—it was 2008, and I, along with a small group of other business intelligence (BI) and data engineers, became excited about the business-transformation potential of the then-new AI technology. We realized that no matter how valuable our BI capabilities are, there is so much more value in a technology that is able to predict what will happen, learn and adapt. Today, I’m VP and general manager of Intel’s AI Solutions Group in SATG/OCTO (Software and Advanced Technology Group/Office of the Chief Technology Officer), and the group has grown to 220 employees, contributing USD 1.3B in value to Intel through additional revenue and cost reduction. 

Over the ensuing years, the AI Solutions Group and the field of AI have grown side by side—to some extent, predicting the future, learning and adapting were exactly what we had to do to grow an idea shared by a few friends into the global organization we are today. As an avid reader of science fiction, technology and history, I often find my inspiration for new horizons for the group by reading about the past and the future. It’s all about identifying which of the latest innovations in the AI domain are the ones that will positively disrupt Intel’s critical working processes.

The AI Solutions Group has focused primarily on implementing AI to accelerate product development, optimize manufacturing processes and increase revenue. Here are some examples of how we have put the power of AI to use: 

  • Embedding AI into pre-silicon validation processes.
  • Personalizing unit testing during high-volume manufacturing.
  • Automating manufacturing yield analyses.
  • Increasing the scale and reach of sales teams. 

Staying competitive requires Intel to focus on continuously improving battery life and maximizing performance. To achieve this, we are building AI inside. In partnership with Intel’s Client Computing Group (CCG) and Data Platform Group (DPG), we are embedding AI algorithms into Intel products that dynamically adjust the power limits of our processors based on the application’s workload. This means you get added turbo burst when needed and extended time in turbo for sustained workloads. We’re also using novel algorithms to reduce chip voltage at lower temperatures to improve battery life and performance. The next generation of Intel® Core™ processors will include more than 20 AI solutions across all layers of the product. These AI solutions will do everything from defining how chips balance battery life and responsiveness to maximizing performance. The algorithms have contributed to performance gains of more than 40% for specific workloads and a 5% increase in battery life for some mobile devices. Over 25 AI-based features have been developed so far, in partnership with Intel’s CCG and Design Engineering Group (DEG). 


AI Is Critical for the Future of Intel 

Intel’s business is becoming exponentially complex, and the company faces unique challenges from the combination of our IDM 2.0 strategy, increased focus on product execution and expansion into new verticals. However, although Intel is growing, we need to increase the amount of highly qualified workers to perform these complex tasks, which require outstanding analytics skills.

The AI Solutions Group believes that AI can relieve the knowledge worker bottleneck and we will work diligently to create AI that can make human-quality decisions at scale. Therefore, we can create AI that meets one of two primary objectives:

  • Augment human decision-making.
  • Embed AI in the automated processes.

Augmenting Human Decision-Making

AI can augment human decision-making by equipping knowledge workers with AI-generated information, and humans can take actions based on machine insights derived from enormous amounts of data that no human could process. For example, an algorithm can give sales staff insights derived from monitoring and synthesizing commercial and social media trends across thousands of sources like social media, external websites, business information, historical transactions, activity on Intel’s website and more. On top of this data, a tailored Sales AI platform continually uses natural language processing (NLP), a deep recommender system and other AI algorithms to mine knowledge about 750,000 existing or potential Intel customers and create intelligent customer profiles. All of this information is fed to sales staff. In just the last year, Intel’s Sales AI and Autonomous Sales capabilities contributed to over USD 300M of additional sales for Intel.

The AI-generated customer profiles combine Intel business strategies with external triggers to create timely and actionable insights. For example, a time-series change detection algorithm might identify a change to the company business. If it determines that this creates a potential sales opportunity for Intel, the information is relayed to the appropriate sales representative.

The AI system can even perform a fully autonomous sale. It can identify an opportunity, trigger a direct email to the company with an offer for relevant products and complete the sale without human involvement. Of course, just because the system can operate without human intervention, it is important to understand that AI does not now and will not replace humans. Rather, it augments human intelligence, enables humans to perform more complex intellectual tasks that require creativity, and breaks down the tradeoff between productivity and quality. For example, because the AI system takes care of the “easy” sales by itself, the sales staff can focus on more complex but potentially higher-value sales opportunities.

Embedding AI in Automated Processes

By injecting human-like decision-making into automated processes, an algorithm can make instantaneous and complex decisions that imitate a human’s expert judgment. For example, test algorithms actuate a different test scheme for every unit manufactured, based on previous test outcomes. This reduces the total duration of tests and improves yield and quality. In the last few years, these algorithms helped enable the production of millions of additional units and reduced test time by 50%, compared to legacy methods, with no negative impact on product quality. This amounts to hundreds of millions of dollars in cost reductions and additional units sold.

To enable unit testing at scale, the AI Solutions Group partnered with DEG’s Manufacturing & Product Engineering (MPE) team to create the MAESTRO platform, an AI infrastructure embedded into the testing procedures of Intel’s high-volume manufacturing. Full data access enables unit-level actuation, and over 200 AI algorithms written in Python are used for optimizing various test tasks to improve the various product health indicators. The team uses MAESTRO and highly efficient problem exploration techniques they’ve pioneered to create and implement new algorithms.

AI also improves product quality, resulting in fewer outgoing defects and faster response times when quality incidents do pop up. In the past two years, 11 critical cases were mitigated, with a turnaround time of days instead of weeks. The impact of this time savings on Intel’s bottom line is profound, and is felt in many ways, from decreased cost of materials and labor to increased customer satisfaction.


Technology Aids Execution

To enable these AI wins, our AI Solutions Group focuses on two pillars: strong execution coupled with advanced technology.

Strong Execution

On the execution front, the team applies modern Machine Learning Operations (MLOps) practices and has over 500 AI algorithms embedded into core business processes, with millions of inferences each week. To enable this scale of AI operations, the quality and reliability of the algorithms and AI systems we create must be impeccable, with 99.9% availability and accuracy. (Read more on AI model quality.)

Advanced Technology

Solving the complex challenges that Intel faces requires state-of-the-art technology. The AI Solutions Group harnesses the most impactful and relevant AI research to solve real Intel problems.

For example, the team is using the latest developments in AI-based language models to transform AI-customer interactions. They’re also exploring machine programming and advanced online learning techniques to improve Intel’s chip design processes. Partnerships are key to this high level of execution. The AI Solutions Group works very closely with its counterparts in the various business domains at every step of the way, from AI initiation to productization. These partnerships enable AI to go deep into the core business processes and create real value.

But There’s More

It isn’t all just methodology and technology, though. Our real competitive advantage comes from our human assets. We never compromise on hiring only the best—people whose hearts are as large as their AI skills. The AI Solutions Group creates many growth opportunities for its members, offers a personalized working environment and gives back to the community through medical, governmental and educational projects using AI. The result is a retention rate well above the average for the AI industry. By putting talented people to work on advanced technology that makes a difference for Intel and the world, magic happens.


Enabling “AI Everywhere” at Intel

To address the huge potential for AI technologies across Intel, in August 2021 the AI Solutions Group launched a new program called “AI Everywhere,” headed by Nufar Gaspar, to help other groups at Intel in their AI journey. We created the program with a clear vision: to scale AI value and usage across Intel. When building this program, we used the many learnings we gained over the years in the AI Solutions Group and made them accessible to the rest of Intel through hundreds of offline and live training hours, self-service AI tools and professional services that include consultation and use cases. An employee AI Portal provides access to all the information needed to apply AI at Intel. 

Many of the best practices and operating principles used by the AI Solutions Group are described in a recent article, “7 Things Managers and Leaders Can Do to Facilitate a Successful Artificial Intelligence Adoption in Their Organization” and an IT@Intel white paper, “Push-Button Productization of AI Models.”

My colleague Nufar says, “The way we scale the impact is through the internal AI community we foster. The impact and feedback we get from these activities is exciting.” The AI Everywhere program brings together individuals, teams and local communities of AI practitioners and enthusiasts from across the company into one active and growing community. Currently, the program has over 4,400 members who facilitate Meetups, mentoring and trainings, and encourage knowledge exchange.


Where Do We Go from Here?

Although generating USD 1.3B in value and deploying 500 algorithms is an impressive achievement, the potential that AI holds for Intel is even greater. The AI Solutions Group has set some bold goals for the next three years, including doubling the group’s business value and the enhanced product quality Intel derives from its internal AI capabilities. The team also aims to deliver a product performance improvement that’s equivalent to an additional product generation, using embedded AI algorithms. 

Of course, I and rest of the AI Solutions Group realize that although AI is already transforming how Intel works, there is much more to learn about the optimal human-machine division of labor, and there are many more business domains that can benefit from AI. We’re shaping our long-term plans to make AI a core differentiator in Intel’s strategic battle for its future—whatever that future brings. It is sure to be an interesting and inspiring journey.

For more information on Intel IT best practices, visit