When Computer Weekly met Intel CIO Kim Stevenson last year, she extolled the benefits of business intelligence (BI). This year, BI is being taken to a new level.
As global CIO at the world's largest chip manufacturer, Stevenson (pictured) manages a team of 6,000 IT professionals.
Despite the scale, her role is much the same as any other CIO – "I am still plugging away at delivering more value," she says – but this year, one of her main areas of focus has been Intel's PUE 1.06 datacentre.
The datacentre runs at 5MW and holds 20,000 blade servers in tall racks that go all the way to the ceiling. The hot aisle runs at 120°F and the cold aisle at 68°F, and it uses free air cooling.
Stevenson says Intel is using the datacentre's capacity for high-performance computing (HPC) to solve a particular business problem – one that should be familiar to large manufacturing firms that need to integrate building blocks and components developed across multiple teams.
The problem, she says, it that it takes too long to get system on a chip (SoC) products to market.
Using BI to speed up product development
An SoC design integrates motherboard components, processor, memory controller, graphics and sound on a single wafer, which simplifies computer designs. This means the final product can be produced at a lower cost because fewer external components are needed.
It is an area the engineering team has been working on, and IT has aligned with engineering to tackle the problem.
More CIO interviews from Computer Weekly
"Everything Intel has always done is complicated, but the big difference in an SoC is that you have to integrate different blocks, each of which is developed to its own schedule. And these parts have to be integrated and tested, just as a typical large manufacturer would need to do," she says.
By analysing the engineering process for SoC designs, Stevenson says, the IT team identified around 25 projects which could help the business by reducing the cycle time.
One of these projects involved optimising the compilation of engineering changes. This process would typically take eight hours to run, with an engineer only seeing one job a day.
"Thanks to the new datacentre, engineers can now see their output four times a day," she says. "I can now take my longest lead time item and make it faster. The integration work is also four times faster."
Stevenson says the principles Intel used to speed up its integration work could be applied in any company with high-performance computing (HPC) requirements, such as high-frequency trading or geothermal analysis.
"You'd get tonnes of performance improvement at a lower cost," she says.
There is also a BI element, she adds: "The more you get into data analysis and predictive analytics, the more you'll see this kind of HPC requirement due to volumes of data. And if you can do it cheaper, why wouldn't you?"
A way to engage business units in a dialogue
“If you go to a business unit and say we have data scientists who can solve a problem, they'll say they are too busy," she says. "But how about if you say we can create a $10m return on investment in six months?”
This is the approach she took with Intel's outbound call centre, which is tasked with selling Intel designs to manufacturers. Using predictive modelling, the business developed a set of criteria which defined the likelihood of a customer choosing an Intel product.
I hate the term shadow IT. People across the company are becoming more tech literate. Everyone is using services IT is not aware of
Kim Stevenson, Intel
The criteria combined weblogs from Intel's websites with information on availability of a budget and whether the company in question had Intel-trained engineers.
This information was used to model how the outbound call centre could be optimised, and included information on what call centre agents should talk to the companies about based on how they accessed and used Intel's website.
“The call centre was previously organised by territory and region," she says. Her team suggested organising calls based on customers Intel was most likely to win.
"We started in the Asian call centre. In a single financial quarter, the agents in Asia were five times more productive than elsewhere, all based on predictive analytics, based on the data we already had such as weblogs and who had taken training,” says Stevenson.
The approach is now being rolled out across Intel.
The changing role of IT
Stevenson agrees that it has never been easier for people to acquire technology. But she does not see this trend as shadow IT, where the IT department loses control.
"I hate the term shadow IT. People across the company are becoming more tech literate. Everyone is using services that IT is not aware of," she says.
Stevenson urges CIOs who may have worked in a command and control style of IT service to relinquish control in situations where IT cannot add any value.
More on BI and data analytics
- The Path to Payoff on big data analytics
- Taking advantage of data analytics
- Business information university: Game plan for big data analytics playback
- Business intelligence and the cloud
- Forward-looking business intelligence
- Computer Weekly buyer's guide to business intelligence
"The worst place to be as a CIO is to convince yourself you have control, when in fact you don't," she says. "You don't want to stop the flow of innovation with IT. You have to learn to operate under chaos theory."
In fact, Stevenson believes it does not really matter who the people using their own IT report to. Horizontal enterprise business processes, such as order-to-cash or procure-to-pay, are best run in a centralised fashion, reporting to someone who knows something about IT.
But she says specialised workloads, such as manufacturing design and engineering, are the way IT creates value at Intel. "I see a divide between these different types of workloads." As such, IT may not necessarily be best placed to run them.
That said, enterprise resource planning (ERP), customer relationship management (CRM) and other enterprise workloads are rapidly moving to cloud.
But, in Stevenson's experience, core value workloads are not moving to the cloud. "They are becoming deeply analytical because analytics gives better operational performance metrics such as lower cost and faster time to market," she says.
IT departments have to become more customer-centric, according to Stevenson. "We need to prioritise our work to help the business get work done. If we can't do it as an internal IT organisation, these people cannot wait for IT – they will buy the technology elsewhere,” she says.
Stevenson believes such user power is a good thing, but it requires a change in IT's role, particularly in terms of how security and privacy are managed. CIOs have to change their approach and align with the business, while still maintaining compliance.
Is security and privacy there to help the business move faster or are they there to say 'no'? Most IT departments act like they are the 'no patrol'
Kim Stevenson, Intel
"There are governance concerns, particularly around security and data privacy," she says. "Is security and privacy there to help the business move faster or are they there to say 'no'? Most IT departments act like they are the 'no patrol'."
Stevenson urges IT departments to change their approach to security to directly align with business objectives.
"Rather than say 'no', IT can show the business how to achieve its objectives in the most secure way," she says. For instance, IT can show the business how to develop secure code.
Of course, this is easier said than done. IT's role in governance and compliance has often been seen as a barrier to innovation. Users circumvent IT, using consumer tech to access corporate systems, and marketing and other departments are routinely buying more IT than the IT department.
But, as Stevenson has found, IT's true value is in turning ideas into proven concepts that can scale the business.