Microsoft Announces Copilot in Viva, Introduces “Viva Glint”, and Shows Why it Matters
April 20th, 2023
This morning, at its Viva Summit, Microsoft announced the integration of Copilot into Microsoft Viva and also the formal addition of Glint to the Viva family in July 2023. Microsoft also announced a new “Performance Equation,” via its Work Trend Index Special Report (from the People Science team at Glint), which shows some compelling correlations between engagement and financial performance. (If you need the backstory on all of this, see our previous RedThread articles on Microsoft and Glint here, here, here, and here.)
As I’ve mentioned on social media, I had a chance to participate in today’s keynote panel, so got to peek under the hood before today’s announcement.
Here’s the TL;DR of it all:
- Copilot in Microsoft Viva = LLMs + Microsoft Graph + Viva Apps. Put another way, Copilot in Microsoft Viva applies the power of large language models (LLMs) to data in the Microsoft Graph and Viva apps. Practically speaking, this means that users can now query, in natural language, any Microsoft Viva app, and get a result generated by the LLM from across all the apps.
- “Viva Glint” = Copilot + Glint data + Microsoft behavioral and collaboration data. We’ve known for a while that Glint was moving to Viva, and that by doing so, it would be in the same “family” as Microsoft’s behavioral and collaboration data (provided via Viva Insights). However, it’s one thing to be in the same family – it’s another thing to share data effectively. The addition of Copilot changes the game, as it allows all those data to be accessed easily in an interface that is already well-known by Glint customers.
- Productivity x Engagement = Performance. Microsoft also released a new study (conducted by the People Science team at Glint) that shows why business leaders should care about engagement: because there is a strong relationship between engagement levels in organizations and financial returns. Based on this, they have thus come up with what they are calling the new performance equation: Productivity x Engagement = Performance.
Let’s dive into the details more.
Copilot in Microsoft Viva = LLMs + Microsoft Graph + Viva Apps
As mentioned above, the addition of Copilot to Viva means that users can now ask questions in natural language and, assuming the information is in Microsoft's systems or, specifically, the Viva ecosystem (the Microsoft Graph or other Viva apps), the user will get the answer back in natural language. In a nutshell, this is like having a private #ChatGPT instance applied to all your Microsoft data.
We all know that every #WorkTech vendor out there is currently trying to figure out how to apply #ChatGPT to their technology. A key difference here is that the Microsoft Graph has a bucket full of data on users that most Work Tech vendors do not, given that Microsoft is also many organizations’ productivity suite.
As a result, Copilot has the potential to:
- Increase productivity by combining data from productivity and people systems, as in the example of Viva Goals suggesting draft goals based on a business strategy document in Word (see Figure 1). For other examples, see the Microsoft blog here.
- Connect relevant talent-related activities to where the work is happening by providing workforce and talent data, insights, and nudges into productivity tools where people are already working (and not just in Teams).
- Provide better starting points for feedback and communication via Copilot, within the systems in which they will be used (e.g., Viva Engage).
- Find learning content, topics, and experts more easily using Copilot in Viva Learning and Viva Topics.
- Provide far more personalized recommendations and nudges, as it has much more granular information on individual employees.
Further, given Microsoft’s strong relationship with IT, it can also more easily handle security, compliance, and privacy needs.
Figure 1: Example of Copilot and Viva Goals | Microsoft, 2023.
“Viva Glint” = Copilot + Glint data + Microsoft’s behavioral and collaboration data
Today’s announcement confirms that Glint will be a part of Microsoft Viva as of July 2023 and will heretofore be known as “Viva Glint.”
While you might be tempted to think this is just a marketing / naming convention shift, with Glint moving from being a product "owned" by LinkedIn to one "owned" by Microsoft, it is actually a bigger deal than that. By bringing Glint officially into the Viva family, Microsoft is setting up Glint to integrate cleanly with its other Viva apps, like Goals, Engage, Learning, and Insights. But, as I have said before, it is one thing to bring things together. It is another to make them work together.
This is where the addition of Copilot, which relies on the Microsoft Graph for collaboration and behavioral data, likely changes the game. By adding Copilot to the mix, Glint customers should be able to ask questions, right within Glint, that span all of these different data sources (the Microsoft Graph, the other Viva applications, and Glint).
This breadth of data access, analysis, and natural language reporting capability will make Glint one of the most powerful employee engagement and experience tools on the market.
That isn’t to say you couldn’t get at these same insights by creating a data lake or use a multi-source analytics platform with all these data. You could. But it wouldn’t be prebuilt into your employee engagement/experience platform and pre-designed to roll out to all your employees. And, it may not be very cost-effective, either.
To be clear, what Microsoft is announcing today is, of course, the vision. And I haven’t seen the reality (just demos) and customers aren’t using it yet, so we don’t *really* know how it works. All that said, the vision is pretty compelling.
Let’s return to the real world, though and talk about what Viva Glint will be able to do come July. According to Microsoft, Viva Glint customers can use Copilot to:
- Summarize and analyze thousands of employee comments (see Figure 2)
- Provide leaders a fresh way to explore feedback by asking questions via natural language
- Help leaders accelerate and deepen organizational understanding by bringing together aggregated employee engagement data coupled with behavioral and collaboration data from Viva Insights and the Microsoft Graph
Figure 2: Viva Glint Comment Report | Microsoft, 2023
But we obviously know this is just the beginning. Some other potential ways I could see Copilot (or any LLM, really, in an employee experience platform) being used in the future include:
- Summarizing survey results quickly and in real-time in natural language, which could be especially useful for employees who may be less data literate.
- Generating customized reports on specific aspects of employee experience data via integration with Office365 products.
- Creating first drafts of responses to employees about questions that arose as a result of employee feedback.
There is clearly a lot of potential with this addition of Viva Glint to the Viva + Copilot family. It will be interesting to see what real customers think of it once they get their hands on it.
Productivity x Engagement = Performance
Finally, let’s turn to the last part of the announcement, the results of the Work Trends Index Special Report, which provides the raison d'être for this entire product set.
According to this study, there is a clear correlation between employee engagement and financial returns. Thus, the logic follows, even in an uncertain market, folks should still be investing in employee engagement. And, as you might expect, much of what Viva offers aligns with what drives engagement. To be fair, our RedThread research on performance management essentially says the same thing – that clear goals, regular feedback, and strong communication are all keys to engagement and performance.
Let’s dive into the details of this study a bit before we parse it out. You can read the study’s full methodology at the bottom of the page here, but my understanding is that the People Science team basically looked at what would happen if you invested $10,000 in a stock portfolio of the top 10% of engaged companies and the bottom 10% of engaged companies (based on Glint’s survey data). They then tracked those companies’ stock over the course of a year-long period and, surprise surprise, the most engaged companies had stronger stock performance.
(Ok, maybe that last sentence has a bit of irony in it. Maybe sarcasm? One of those. (Alanis Morrisette probably knows which one.))
Perhaps that is because I am a bit like (as I imagine my fellow people data nerds are also thinking), “Did we really need ANOTHER study that shows a relationship between engagement and performance?” And a part of my heart says no. We don’t. We know this works. Can’t we just do the right thing for people, people?!
And yet, I still included it in this blog, even though it wasn’t a “headline” in Microsoft’s announcement. And the reason is this: this study has several unique aspects worth your attention, especially if (God forbid) you still are trying to convince leaders in your org to care about engagement. So here are the reasons this study matters:
- Financial returns are based on stock price over a year. Stock market performance is a metric that business leaders can sink their teeth into, not some other softer finance-related metric that they’ve never heard of or one that is based on employee perceptions. Also, the difference is significant: there is a +$46,511 difference in market cap per employee per additional point of engagement (see Figure 3).
- The analysis covers an especially volatile year – and engagement proves more important as uncertainty increases. As you can see in Figure 3, as the realities of inflation and a potential recession set in (around mid-April 2022), there is a significant divergence in stock price between the most and least engaged companies. That difference only exacerbates with time, and by late summer the most engaged companies are also beating the S&P 500 index. Of course, there can be confounding factors driving these differences, but apparently, industry (17 are represented), org size (all large orgs), and geography (worldwide distribution of orgs) were not some of them.
- The data set is really large (over 3 million employees and more than 225 organizations), addressing some concerns in smaller studies with fewer data points. The size of the data set makes these results even more compelling.
Finally, and this is not really an explicit reason you should pay attention to this study, but I will mention it anyway: The Glint People Science Team is one of the nerdiest sets of people data nerds out there. I have a pretty high level of trust in what they produce — and you all know I hardly trust anyone outside RedThread when it comes to data and analysis. Therefore, I am giving this study more credence than I do most vendor-created studies.
I know you all don’t need more convincing that engagement matters. But some folks you know may, and I think this study is worth your (and their) attention.
Figure 3: 2022 financial portfolio returns comparison of the top 10% highest and lowest engagement scoring companies and the S&P 500 | Microsoft, 2023
My take on all of this
Ok, this got long. #sorrynotsorry. Let’s synthesize and then wrap.
I have been saying for years – maybe a decade! – that it drives me crazy how much rework we ask employees to do when it comes to things like inputting goals, OKRs, and feedback into HR Tech systems.
Further, I have long thought it was a bit nuts that we thought employees would use templates that exist within an HR system to guide them when planning for a performance-related meeting or to draft their feedback. I mean, honestly, how many times have you written up feedback in Word and then just cut and paste it into whatever HR system you had to put it in, without looking at the system before you did it? Yeah, I can’t count the number of times either.
Therefore, I love, love, LOVE the idea of marrying productivity and Work Tech systems and having them work together. The increased efficiency alone makes my heart happy. Then when you think about the power of AI across these systems to get people started on communications, providing feedback, and understanding insights quickly… It is incredibly powerful.
Further, the idea of being able to more easily combine perception, behavioral, and collaboration data is seductive. Think of all the things we can learn. Imagine all the changes we can make to practices, org designs, policies, norms, and cultures to make the world of work better. The possibilities seem endless.
In so many ways, this seems like a dream come true.
Yet, I know that not every dream, in reality, turns out as we thought it would. There are lots of places where this dream could potentially go awry, and the realist in me requires us to think through those, too. So, let’s have a look at them. The worries, as far as I can see them, are both practical and philosophical concerns.
Practical Concerns
There is a whole range of practical concerns – here are just a few that jump to mind (but I’d love to hear more from you, if you have others to add).
- Accuracy: We’ve seen with #ChatGPT that if it doesn’t have the information, it will just make stuff up. That’s not ok if we’re talking about HR policies, interpretations of data that may impact people’s lives, etc. Even though Copilot is only looking at Microsoft data for returning results, what happens if the data are not in the Microsoft ecosystem? Does it make up answers?
- Bias: Bias exists in all of our human interactions, and thus shows up in the data upon which these models are trained. Therefore, bias will show up in the LLMs, just as it does in other algorithms. I know, I know – the data and model are likely to be less biased than we are as humans. That’s great. But here’s the rub. I, as a human, can only spread my bias so far so fast. When the bias – even if it is just a little bit – is within the tech, it literally expands exponentially faster. Further, when humans think they can trust the tech (because we are lazy when it comes to critical thinking and making assumptions), we won’t actually do the critical thinking to stop the bias. I don’t see enough bias interrupters being built into the tech to believe that this won’t happen with the suggestions and recommendations made by LLMs. So tech-enabled bias is more dangerous, IMHO.
- Data privacy compliance: LLMs, by their very definition, use large amounts of data to run and refine the model. How does that square with things like GDPR, where, for example, someone has to be forgettable? Once the data go into the LLM, can it ever come out / can someone be forgotten?
- Data ethics: This one is a whole ball of wax for the entire people analytics tech field (which is why we are studying it right now!), but specifically with something like this – is it appropriate for companies to be integrating all these data without telling employees what is being done with their data? What do employees need to be told? When? And at what depth?
- Ecosystem interoperability: Everything I wrote about above sounds great if you are a 100% Microsoft shop. But what if you are a Glint customer and also a Google shop. Or you have tools other than Viva Goals or Viva Insights for your goals / OKR or analytics tools? What then? Will you get any of the benefits of these new releases? (My guess is no, but we haven’t really heard about this yet.)
FWIW, Microsoft’s statement on the topic of Responsible AI is the following:
“Our work is guided by decades of research on AI, grounding and privacy-preserving machine learning as well as our Responsible AI Standard and core set of AI principles: fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability.”
While this is a good start, we need all need to do more.
Philosophical
You thought those practical concerns were rough? Well, let’s talk about a couple of philosophical ones.
- Has too much control been ceded? As I’ve spoken with HR leaders over the last month or so about this technology, one theme has come out loud and clear: A fear of losing control. They fear they can’t turn off all these data integration capabilities (especially since IT may be driving things), that employees will have information before HR and senior leaders can prepare answers (a fear amongst especially conservative orgs), or that employees might take actions on certain issues without HR knowing. You all know our research shows the benefits of pushing information and decision-making down into the organization and not holding too much information up at the top. However, there are some legitimate concerns here, and they will take some significant cultural and policy adjustments to address them.
- Is Big Brother finally here? The integration of all these data may give employees pause, as they wonder what information their employers have about them and the potential negative impacts of that. All the workforce monitoring technology that hit the market over the last 3 years isn’t helping with this, either. We here at RedThread firmly believe in #peopledataforgood, but we also know this doesn’t happen on its own. Organizations need to create a people data ethics charter, develop checks and balances on how and when information is used and why, and ensure these are baked into technology and operational systems. Finally, organizations need to develop a clear strategy for ongoing communication and transparency around data, the analyses that are being done, and the resulting decisions and actions being taken by orgs.
- Does this tech solve the problems that matter? Everyone is running to implement #LLMs into their Work Tech. But, as with anything tech related, we have to ask ourselves if we are solving a real problem or just running after the latest shiny object. Generally speaking, I think Copilot can dramatically increase productivity, which is a real win. Further, the ability so summarize employee comments with more specificity and to summarize data more quickly, in general, are real opportunities. We just have to be careful not to get seduced by the fancy new tech and not do the hard work of improving or changing the processes and practices that we identify as needing to be adjusted. Shiny tech, in and of itself, doesn't generally solve problems.
There are obviously mitigating steps that can be taken to address all these concerns. The point is we need to keep them in mind as we are experimenting with and rolling out this tech. We don’t think the genie can go back in the bottle, but it can be influenced and guided toward good.
Wrapping up
Today’s announcement is significant in our Work Tech industry and will have far-reaching repercussions. I strongly suggest folks watch the Summit, if for no other reason than to ensure you are up to date on the latest and greatest in this space. I’d love to hear your thoughts below. (And thanks for indulging me with this excessively long post.)
I know I am unusually positive about this tech in this blog. You all know I defend our reputation for unbiased, no bull$h!t research staunchly at every turn. So let me be clear about our financial incentives, in case that is of concern. Whenever any of us speak at a vendor event, we are paid a speaker’s fee to compensate us for taking the time to share our opinion or research. We do not take money to write blogs on products or events and we do not get those blogs approved by anyone, though we do often give vendors the opportunity to review our blogs for factual accuracy (on product details, executives’ titles, etc.). Our opinions are our own.
That said, we still have to pay our team and our mortgages. If you want more of this type of content and want to support our ability to produce it, consider joining our research community membership (it costs less than a conference pass and you will get to use it all year long to learn and network!). If you’re a vendor, consider joining our Tech Consortium, which explicitly supports our creation of high-quality, unbiased tech research.
