The frontline is HR’s biggest AI blind spot

If you ask most HR leaders how their workforce is performing, they can pull up a dashboard in seconds. They will see engagement scores, flight-risk predictions, productivity trends, and all of it increasingly powered by AI. It feels like we finally have a real-time read on our people, yet a quiet problem hides within those confident dashboards. For a worrying share of the workforce, the AI is just guessing.

The people it is guessing about are our frontline employees, and they are actually the majority of the working world. According to the World Economic Forum, frontline employees make up an estimated 80% of the global workforce. Deskless employees stock shelves, drive routes, staff wards, run production lines and serve customers. Yet many of the digital signals organisations increasingly use to understand engagement, retention, and productivity come most readily from knowledge employees; the same is true of industry benchmark figures used for comparison. Frontline teams, by contrast, have almost no digital footprint in the organisation. That mismatch is HR’s biggest AI blind spot, and it is getting bigger as adoption accelerates.

The problem: Our models can only see the desk

Modern people analytics runs on data. When employees produce digital breadcrumbs, the model can see them clearly. Knowledge employees leave many breadcrumbs because they produce many emails, calendar invites, collaboration tool logins, document edits and instant messages. All of these provide signals that AI uses to draw inferences about engagement or burnout.

Frontline employees leave almost none of that behind. Many do not have a corporate email address or a company laptop, and their work does not generate calendar density or message volume. When they clock in, serve customers, and clock out, that shift is real and valuable, but it is largely invisible to a system built to count digital activity.

So when we point AI at “the workforce,” it only points at the part it can see. The frontline gets underweighted, averaged away, or dropped from the sample entirely. The dashboard still returns a clean number. It just is not a number about everyone.

Why it matters: We are optimising for the visible

In my experience, this disconnect is easy to miss because our dashboards often look complete. However, the consequences go well beyond data quality. The issue is that the employees AI sees the least are often the ones the business most needs to understand. Frontline roles typically carry the highest turnover, the largest share of labour cost and the most direct impact on customer experience and safety. They are exactly the people whose engagement and retention we most need to understand and the ones our tools understand least.

When AI-driven insights skew toward desk employees, three things occur

1. Our retention models miss the very population with the worst retention, so we are least accurate where the stakes are the highest.

2. There is little investment in the frontline. If frontline activity barely registers in the data, it does not come up in planning conversations either. The resources flow to the employees who were already easiest to retain.

3. And we start to mistake a partial picture for the whole one. A confident engagement score that quietly excludes most of your headcount is not a neutral gap. It is a systematic bias that compounds every time we act on it.

Another knowledge dimension keeps me up at night. Frontline teams hold vast amounts of institutional knowledge, and sadly, much of it walks out the door when they leave. If we cannot see their engagement clearly, we certainly are not capturing what they know before they go. As AI reshapes HR, the risk is not only that we make decisions on incomplete data. It is that we automate that incompleteness and scale it.

What HR leaders should do differently

The good news is that once you see a blind spot, it is fairly easy to fix. Here’s how:

• Start by auditing where your people data actually comes from. Ask a blunt question about every signal feeding your analytics: Does producing it require a corporate email, a company laptop or a desk? If so, your frontline is likely missing from that signal. This audit is often a wake-up call, as leaders discover that most of their “workforce” data is really knowledge-worker data wearing a company-wide label.
• Next, give the frontline a real digital front door. People cannot show up in your data if they cannot easily show up in your systems. That means meeting them where they already are—on a phone, in the flow of work—without forcing them through corporate credentials built for office employees. Simple, secure, mobile-first access is the precondition for everything else.
• Then, design the listening for their reality. For example, a 30-minute annual survey built for a laptop is wrong for someone with a locker and a 10-minute break. Instead, try short, in-the-moment pulse check-ins, recognition and shift-level feedback. Capture this frontline sentiment in a form that fits their day. Aim for a steady stream of frontline voice, not one heroic survey a year.
• Finally, hold your AI vendors and your own dashboards to a coverage standard. Before you act on any insight or benchmark, ask what population the model was trained on and what percentage of your frontline is represented. Then pair those signals with the human sensor layer you already have: frontline managers who can tell you what the model cannot see. Make sure your data actually reflects your workforce before you act on it.

Close the frontline data gap

None of this means AI has no place in frontline HR. If we use it well, it can finally give this population the attention it has always deserved. But that only happens if we feed it a complete picture.

READ MORE: Tourism offers jobs, not careers. Mandai Wildlife Group is trying to prove otherwise

AI is only as good as the data behind it. If the data is skewed, the results will be too. Every cycle of AI makes the model more confident in the same skewed inputs, and a blind spot we tolerate today becomes one we have hard-coded tomorrow. The organisations that get frontline engagement and retention right in the AI era will be the ones deciding now to make their frontline visible, before the technology quietly decides for them. That means giving frontline employees access to the right tools and ensuring their experience is represented in the data.

Closing that gap is not a data project. It is a choice about whose experience of work actually counts.

About the Author:

Trecilla Lobo is Senior Vice-President of People and Culture at Flip, where she leads the organisation’s global people strategy. This article was first published on HR Executive.