The Future Leaders Hiding in Your Recognition Data: A Conversation with Workhuman’s KeyAnna Schmiedl

In conversation with KeyAnna Schmiedl, Chief Human Experience Officer, Workhuman

Workhuman has spent 25 years telling companies that recognition matters. Its newest product argues recognition data can identify your future leaders years before your succession planning would. I sat down with KeyAnna Schmiedl, who runs the entire people function at Workhuman and lives with this data every day, to ask what that actually means, and where it could go wrong.

You’ve been arguing for 25 years that recognition is the real-time data source for the employee experience. What’s changed?

What’s changed is the models. We’ve always used some level of AI in our tooling, but as the models have gotten more intelligent, we’ve been able to dig into data sets and outputs we couldn’t reach before. That’s what led to Future Leaders, a new category we’re calling AI-driven leadership development. Instead of static frameworks and annual succession cycles, it’s a live, self-learning tool that predicts future promotions from behavioural signals inside recognition moments.

It’s called the Ascend model, built on tens of thousands of distinct leadership signals from real workforce data over the last 25 years. It can predict about four years out who the future leaders of a company will be, typically mid-career employees building momentum and influence. What excites me most internally is that it can also spot when the definition of leadership itself is shifting. The leaders of 10 or 15 years ago didn’t necessarily emphasise what we need now. We don’t want the model telling us about capabilities that are already outdated. We want it to catch the trends as they’re emerging.

Have you mapped that to retention? If the model flags someone as a future leader, can you also see who’s a flight risk?

That’s exactly where my head went too when the team first showed me this. It’s essentially a high-potential pool, but it feels more objective because it’s crowd-sourced from peers rather than coming from one manager who happens to be good at advocating for their people.

The same data source tells you where the hot spots are and, by proxy, the cold spots. If the Ascend model says someone will be ready for a leadership role within six months, and that same person has gone quiet, recognised constantly a year ago and now getting one moment a quarter, that’s a serious retention signal. You’ve got two data points pointing the same direction: recognition has dropped, and this is one of your most ready people.

Some of the strongest performers I know don’t generate much recognition traffic. Their personality is to get on with things quietly, or their role just doesn’t lend itself to visible praise. How does a model built on peer recognition avoid mistaking visibility for value?

That’s extroversion bias, or proximity bias. We find recognition varies most by seniority and by role type. The more senior you get, the less recognition you tend to receive. Take total rewards teams. You typically only hear from them around bonus time, merit cycles, benefits renewal. The rest of the year they’re not visibly interacting with the company, but their role is critical for retention and pay equity.

So the model adjusts for frequency in two ways: how often someone is recognised, and when that recognition tends to land. Someone might get recognition rarely, but at the most impactful point of their work, when everyone is watching. That’s different from, say, workplace experience teams who get recognised constantly for visible, in-the-moment help. The model has to weigh those differently rather than treating all recognition as equal value.

We saw this clearly with a client like Rio Tinto, where the workforce is largely in mines. You’re not getting people to type up a recognition moment sitting at a desk. We had to understand how they actually work, so recognition could happen through something as simple as a phone line at the mine site, someone calling in mid-shift to shout out a colleague.

Historical recognition carries bias. Have you built anything into the model to counter that rather than replicate it?

Yes. One of our most recognised tools externally is the Inclusion Advisor. After you write a recognition moment, it flags things for you before you submit. It might say: you mentioned service, and that language tends to align with women receiving lower award amounts than men, or people being thanked for helping and service rather than innovation and urgency. It’s not telling you to change anything. It’s making you pause and consider whether, alongside being helpful, this person also pushed something forward in a way worth naming.

It colour-codes what it flags, gender, race, ethnicity, generation, location, because we have a genuinely global workforce across Dublin and Framingham with real cultural differences within that. With the newer agentic AI work, we can surface that more seamlessly in the moment rather than as an afterthought.

You mentioned neurodivergence affecting how people read tone at work. Is that something Workhuman has built for internally?

We have a neurodivergent employee resource group who ran their own AI hackathon, and the ideas were strong enough that we’re rolling several out. The first is called My Second Brain. It came from someone in the neurodivergent community who found it hard to judge urgency and tone across Slack, email, and a calendar all firing at once. The tool scans everything, your calendar, project management, email, Slack, and gives you a briefing for the day that you can refresh as things change.

I use it every day. It’s caught things for me, like a senior leadership meeting appearing on my calendar and the tool flagging that I hadn’t prepped for an update I’d have otherwise forgotten. It came from solving a specific challenge for the neurodivergent community, but it turns out to be a challenge most of us have in some form.

That instinct, building for one group and finding the whole company needs it, is what makes Future Leaders worth watching once it’s live in more organisations rather than freshly unveiled. KeyAnna and I are both heading to UNLEASH World in Paris this October, where we’re planning to pick this up in person. We’ve also agreed to revisit the Future Leaders story once Workhuman has run the model internally for six months, to see what the data actually surfaces and whether anything has surprised her.