Your Questions, Answered

A context layer is a foundational AI infrastructure that translates raw data into meaningful, governed business context for AI agents. It enriches models with definitions, rules, and relationships so they can accurately reason over company information without hallucinating.

A context layer sits as middleware between your foundational data infrastructure (bottom) and your AI applications/agents (top).

• The Bottom Layer (Systems of Record): Data warehouses, SaaS applications (like HRMs), vector databases, and semantic layers (Gloat has this built in).
• The Middle Layer (Context Layer): Unifies your data into an AI-ready knowledge graph, catalogs business terms, and manages governance policies. Gloat is an example of this kind of unification.
• The Top Layer (AI Applications): Domain copilots, chatbots, and agentic workflows that query the context layer to generate accurate, localized responses. This is what Gloat leverages to display answers to queries based on that unified data.

To secure a context layer, you can implement zero-trust data governance at the middleware level. This ensures AI applications only access authorized data. This can be done by enforcing Role-Based Access Controls (RBAC), implementing dynamic data masking, establishing real-time guardrails, securing the data pipeline, and auditing compliance and lineage.

We just did a podcast today with Rita Azevedo from Sana on that, that you guys might want to listen to. L&D is a really fun, easy-to-understand transformation. The traditional L&D function is a publishing department. Collect data, identify problems, author a course, publish the course, produce the course in the LMS, launch the course. It's very much like writing a book. And these books sit on the shelf, and then the employee is supposed to figure out what book to read, because they've got 500 courses in front of them. That's nice if you have time to read books, but people don't have time to read books. They don't have time to take courses. They don't know what course to take.

What happens with AI is the content development process is radically simplified because you can literally take a document or a recording or a video or a PowerPoint or whatever artifact you have, put it into the system, and it'll generate a course. And you can tell it, "I want a 5-minute course" or "I want a podcast" or "I want a 10-hour course with certification", and it'll generate that, and it'll be 75% perfect. Then you'll be able to tweak it, so you can generate content way, way, way faster.

And then the user has a personal AI. In our version of this, which is Galileo Learn, there's a Super Tutor and you can just start asking it questions, and it will either answer the question or teach you things, depending on how much information you want. And it will read the whole corpus of all the courses ever created — and you don't have to be within a course to do that. That's a very high ROI use case, because you can eliminate staff, you can eliminate external vendors, you can eliminate contractors, you can eliminate graphics tools, fancy video stuff. So, that's a high ROI area of agentic HR that has fairly mature solutions right now.

When you think about agents in this space, you can create a workflow where an agent tells you that there is a major skills gap around leveraging data that many people in Customer Success roles have, then creates new learning assets around these skills gaps, then pings the people in those roles with information about what they need to learn and shares it with them, then after a pre-set period of time of learning (say 4 weeks), deploys an assessment to them to see what level of competency they've achieved. People leverage their wisdom and feedback on the process, create the strategy and manage the overall execution, but the agents do the heavy lifting.

Every state and jurisdiction has different rules for AI testing in talent acquisition, and there are two large lawsuits in the US (Workday and Eightfold) that have yet to be settled, each claiming bias based on race and gender. Despite these rules, most employers don't use AI for hiring, they use it for screening, and the "human, manual" screening process is just as "imperfect." At this point, the volume of AI interviewing, sourcing, and screening is so high most employers use their own judgement to test it, and I personally don't think legislation will do much except force vendors to "prove" that their AI is not gender, race, or age biased.

This is something that's been big on my list for a long time. So let's suppose you went to an AI agent and you said "We're trying to roll out a new goal setting approach in the company. We have a bunch of practices and methods, how you set your goals every year, we want to explain to people why we're doing this, how to do it, what is the role of the manager, etc." You could build a 15-hour course on this, and nobody will ever take it. But you do it anyway, you roll it out, and then you look in your HR system, and you realize only 10% of the people in the company have even changed their goals. Nobody's even done it.

So you want to know what happened, what part of this they didn't understand, or what was not communicated well? You could send a survey, and you're going to get answers like "I didn't understand it," which doesn't help much. But if you told the agent what your problem was, using the language I just used, and the agent asked you a few more questions, and then went out and reached out to each employee in their own unique way and said "I'd like to have a 5-minute conversation with you about the new goal-setting process," and did either audio or video or text-based interview and then consolidated that — you could get the change management information you need from the organization, maybe in a few days. It might take you months through surveys.

That opportunity is sitting there waiting to be captured by somebody, and there are vendors now starting to move down this path. And I know in our case, where we have a lot of technology in our company, we already have our digital twins. I just haven't used it for this, to go and find out what's going on in people's roles that they could help us with. So that's where I think this is going. I don't think the survey industry is going to disappear overnight, but I can see it coming.

Imagine a world where as an employee you receive a "micro-pulse" ping once every 6 weeks that asks 2 questions (part of your overall Pulse set) to get a temperature of how people are feeling in the moment. Once you answer, you can see everyone else's answers as well, in real time. As an org leader, you can get pinged with a collective view of your org in real time.

You could also create an agent to monitor capacity and let you know if people on your team are working more than 50 hours a week for more than 2 weeks — or if they're consistently working nights and weekends — to monitor for capacity overload. That same agent could check in with the employee to see how they're doing. And if their sentiment is negative, it could encourage/coach the manager on how to check in with the employee or redistribute the work to others on the team — or identify other people outside of their team who would be interested in doing that work as part of career development towards their next role.

These agents go well beyond just pinging people to answer a survey — they not only summarize but also synthesize multiple data points, bring context (like Sam is a new parent who may be sleep deprived, so that's why work is getting done at 3am), and take action by sharing out data and insights without anyone asking for them.

I started using the phrase "power skills" a long time ago. The idea is that these are the non-technical human or business-related skills that we bring to work, usually developed over a long period of time — how do you collaborate, listen, educate, inform, align, train? Empathy. We talked a lot about this during the pandemic. How do we create empathy in the organization? As humans, we have flexibility, a sense of joy, a sense of happiness.

At the moment, with people worried about tech and job displacement and "re-engineering" (which means layoffs), people think "If we had really good HR technology, maybe we wouldn't need any HR people." But human beings are very emotional, imperfect beings, and thus the "power skills." When you walk into a meeting to try to solve a problem, it isn't linear what happens. People speak up, people have examples, people object, people are quiet. How you deal with change, how you facilitate change, how you understand the organization?

I talked about Shark Ninja. The things they do there are all cultural. There's no technology at Shark Ninja that you don't have in your own company, but the way they use it and the way they apply it and the way they go after their market is very unique to them. That's the way Exxon decides to build a refinery and operate it — it's different than the way BP does. Those are all the power skills things, and if you're in recruiting, you know this very, very well. The way you assess a candidate is very personal to you, if you're a hiring person. And so I think those kinds of skills, however you define them, are really, really valuable, and they're going to become more valuable.

I wanted to share a little bit about this from Gloat's side: it's not just for Fortune 500 folks, this is absolutely for smaller organizations, small-to-medium sized businesses — we've also worked with governments. There's a lot of different organizations that we can do this with.

I have recently started our AI Innovation Program. Our first cohort closed out pretty quickly. We have a lot of folks interested, so we're opening up another cohort because there is limited space in these. It is absolutely complimentary — there is no cost for anyone to do this. In these workshops, we spend time looking at current process and starting to unpack where agents might be able to come in and help you rethink the way that you are doing work within HR. This isn't (as Josh said earlier) sticking an agent on top of an old process — this is really rethinking the process and having the agent as a persona or an actor within the whole process.

So this is something that we do together in the workshop. We can give you access to play with the agent. We can leverage your data so you can see what happens in real time, if that's something that would be helpful for your organization. If you are interested, you can apply to be part of this program. We're super excited about this program and our partnership with Josh — and having Galileo as part of this overall ecosystem.