Generative AI has become ubiquitous to any conversation about human capital management. However, most organizations have not gone much beyond the talking stage to really leverage GenAI in specific and strategic ways.
Kyndryl, the world’s largest provider of IT infrastructure services, is leading the way. The company, in a phased approach, introduced Copilot for Microsoft 365 to its workforce. Their leaders believe the technology will fundamentally transform the way people work and the skills they need.
But the key to successful use of GenAI is not the technology itself, Kyndryl leaders said, but laying the proper foundation across the enterprise. I talked to Darryl Wilson, Global VP of Kyndryl Consult for the Digital Workplace, and Risham Sahni, Application Architecture Lead for the Kyndryl CIO team, about their experiences and insights. You can watch the complete interview on the Brandon Hall Group™ Excellence at Work podcast here.
Building the Foundation
Let’s start by understanding Copilot for Microsoft 365. “It has a very similar function to ChatGPT in that you can query the AI, prompting it to give you information. The major difference is Microsoft’s commitment to responsible AI that keeps your data secure within your tenant and prevents your prompts and data from training the public model,” Wilson said.
“That’s a huge difference from ChatGPT, which is just using information that’s learned from the internet. With Copilot, you ask it a question and it responds with a value. And it’s embedded in all the applications and tools most employees use each and every day at work, whether that is Microsoft Teams, PowerPoint or Outlook.”
Here is a high-level look at Kyndryl’s approach:
- They used a phased rollout starting with an early access program for 300 users before expanding to 10,000 users, about 13% of their workforce of more than 80,000. They focused on high-value use cases and did thorough risk assessments.
- Individual use cases, many of them solicited from individual employees, were closely examined to understand how Copilot could improve employee productivity. 94% of employees felt Copilot addressed challenges and provided efficiency gains.
- Key metrics showed employees saved approximately 20 minutes per day finding information and saved 10 hours per week overall, with 35,000+ hours saved across all users in a month.
That last number is not a typo — 35,000+ hours saved within a month. Spread that over a longer period across an entire global workforce and you are looking at a complete overhaul of how people work.
“I equate it to the ’90s when the internet was born. It fundamentally transformed business. We got into eCommerce and the expectations of employees changed fundamentally. Same thing in 2007 when the iPhone was released. I think those events are nothing compared to what we are going to expect from the workforce in the next 10 years,” Wilson said.
For that reason, Kyndryl chose an intensely inclusive and collaborative approach involving all functions of the company in launching the Copilot project. That is a critical step because Brandon Hall Group™ research shows that less than half of organizations effectively collaborate cross-functionally for digital workplace technology projects. And only 36% of HR leaders surveyed said they are actively involved in their organizations’ GenAI strategies.
Sahni said the Copilot project started by creating a list of all the possible stakeholders who needed to be involved in the early access process. “We aligned with HR, Labor Relations, Legal, Procurement, Consulting, Data Privacy, and Security and Compliance.”
Once stakeholders were aligned, the group conducted a risk assessment:
- What kind of risks might be opened up by the use of the technology?
- What is the current state environment?
- Does the organization have the information architecture or the data to support the rollout?
A Human-Centric Approach Toward Fundamental Change
“Our strategy toward the overall rollout was very value-focused,” Sahni said. “We were looking at exploring and shaping our workforce dynamics through collaboration and bringing Generative AI here. We wanted to increase the currency of our organizational information to increase the productivity of our employees.”
“We wanted to know how Copilot could act as a game changer when you have to find that information. It’s not new to anybody that when you start your day, you spend time looking for the information you need,” she said. “‘Which article am I supposed to refer to? What Word document was I last working on? What tasks did my manager assign me?’ We wanted to simplify those things. Copilot was exactly the tool which could bring all of these things together.”
Wilson stressed that these types of GenAI applications are “going to fundamentally change the skills that are required for individuals. It’s going to fundamentally change our education programs as people enter the workforce. And so, it is critically important that we take a human-centric approach and not just treat GenAI as a technology project. That’s why working with HR, working with Compliance and Legal, working with IT is absolutely required for success. If this is just an IT project, you will not get the benefits of what this game-changing technology can provide.”
Kyndryl’s approach was exhaustive and relentlessly collaborative. They:
- Conducted legal, security and privacy reviews — involving multiple functions.
- Created use-case submission scenarios that allowed employees (called Kyndryls) to not only submit ideas, but to explain their value.
- Developed an in-house center of excellence, a community of Copilot-using Kyndryls, who went through training and provided input and feedback during the initial rollout.
- Started the Kyndryl AI Governance Committee, led by the CIO team, which invited all stakeholders to get involved.
- Implemented an Organizational Change Management (OCM) and Adoption program to ensure users have the best employee experience with new technology.
“When we were in the initial stages of the early-access program,” Sahni said, “we used to meet every day for two hours because we wanted to go through each assessment, each use case in depth. And we also had the submitters of use cases. So, it wasn’t just the committee making decisions, but it was also the Kyndryls who had submitted use cases bringing their viewpoints. Because at times, you know, when you’re on the other side of the table, you do not completely understand what’s the end goal.”
The committee has now expanded its horizons to other projects and services. “We now have a very structured approach to these projects that includes considerations from HR and legal and other areas,” Sahni said. “We now have a blueprint ready for how you are supposed to prepare for AI. But everything started with the commencement of the early-access program of Copilot.”
Lessons Learned
- Ensure your information architecture is clean — with all the correct types of metadata, data tagging and classification.
“A critical thing to understand is Copilot will have access to all the data that users have access to. If your data is not correctly structured, if it’s outdated, then you might not get the correct information,” Sahni said.
The IT team did a thorough cleanup of SharePoint sites, Teams groups and data to ensure it was structured correctly with updated metadata and tagging. They collaborated closely with HR to update policies at the country level and remove outdated COVID-related information.
When rolling out Copilot more broadly, the team initially failed to get approvals from the workers’ councils in some countries across Kyndryl’s global workforce. This caused them to pause the rollout. To address this, they presented to the workers’ councils on their rollout strategy, support plans, training, who to contact with questions, etc. This allowed them to get approvals in most countries.
A key learning was to involve labor relations teams early in the process to avoid delays in the wider rollout due to missing workers’ council approvals.
- Build deep partnerships to launch GenAI projects.
Developing close internal partnerships between the CIO and HR teams and building a collaborative approach to the Copilot project from the ground up enabled Kyndryl to transfer their experiences to help their external customers with their digital infrastructure projects.
“I was at a CHRO conference in December in Toronto with about 160 leaders,” Wilson said. “We had a breakout session around Generative AI in HR and in wider contexts. And I brought up our experiences that we’ve talked about. And most of the people were actually flabbergasted, like, ‘Oh my goodness, I wouldn’t even think about that.’”
The bottom line, Wilson said, is that any GenAI project is not a technology project, but a project to benefit the entire workforce and must be run that way.
“The project may be run by IT, but it must be in close partnership with HR, with Legal, with Compliance, with all stakeholders. That has to happen from the very beginning. It can’t be at the end or in the middle. Start at the beginning.”
