The four layers of workforce trust: a way of looking at gig economy background checks and identity verification

Conceptually, let’s consider a framework for workforce trust. In this way we can think about workforce trust across four distinct layers: identity, records, signals, and orchestration. Each operates along two axes: proactive versus reactive, and point-in-time versus continuous.

Background checks cover one of those four layers. A complete approach to gig workforce trust covers all four, against both axes.

The implication for gig platforms: thinking about background screening as a list of products to buy makes it harder to keep up with regulations and threats that don’t respect product boundaries. Thinking about it as layers makes the architecture decisions easier.

Here’s why. A background check confirms against multiple records. That’s extremely important to do, but it isn’t the whole of what an employer needs to know about a worker. Employers also need to confirm that the person showing up for work is the person who passed the background check. Identity verifications help to detect a synthetic identity built from stolen documents or changes after an employee is onboarded. Background checks are one input into a larger decision.

For most of the past twenty years, gig economy background checks and identity verification have been sold as separate products and bought as separate workflows. That was workable when the threat surface was smaller and the workforce was less mobile. It’s becoming less workable now. What seems to be happening across the operators we observe is a quiet shift towards thinking about background screening and identity verification as layers. This article is one way of describing those layers.

The Trust Stack. What are the four layers of workforce trust?

The pattern we see across organizations rebuilding their background screening and identity verification approach is simple. Let’s break them down into four layers, from most foundational to most contextual.

Layer one — Identity. Who is this person? Establishing that a real, unique, living human is presenting themselves — not a synthetic identity, not a duplicate account, not a stolen credential being reused. Document verification, biometric matching, liveness detection, anti-spoofing, and uniqueness checks. Everything else depends on this layer being sound.

Layer two — Records. What does the world know about them? Criminal records, employment history, education, credentials, qualifications, regulatory standing, and sanctions. This is what most people mean when they say, “background check.”

Layer three — Signals. What are they doing now? Device behavior, geolocation, behavioral anomalies, social and adverse media, and continuous monitoring. The ongoing read on a worker after onboarding, rather than the one-time read at hire.

Layer four — Orchestration. What does it all mean for this role? The decisioning layer that translates evidence into action — calibrating the level of trust required to the work, maintaining the audit trail, applying the adjudication matrix.

Each layer answers a different question. None alone is enough. A program that invests heavily in records while neglecting identity may still fail against synthetic identities. A program that monitors signals without anchoring them to a verified identity is monitoring noise.

Takeaway. Background checks (layer two) are an essential layer of workforce trust. They are not, on their own, a complete approach to it.

Why does this layered view matter in 2026?

Three things have changed in the past two years.

Fraud has moved between layers. Pre-hire fraud used to simply take the form of falsified records such as a misrepresented degree, or a covered-up conviction. Layer two was where it lived. The most consequential fraud today sits at layer one: synthetic identities, deepfake interviews, stolen documents matched to fabricated personas. According to HYPR’s 2025 research, 95 percent of organizations experienced a deepfake incident in the past year (HYPR report). Pindrop Security tracked a 1,300 percent surge in deepfake fraud attempts across 2024, with attacks in contact centers jumping from one per month to seven per day. (Pindrop report). In those cases, the records often check out. The identity doesn’t.

Work has become continuous. A worker who passed every check on Monday can become a different risk by Friday. Gig platforms have learned this directly. For example, a driver who passed onboarding can have a license suspended a week later, a delivery courier can sublet an account to someone who never passed any check at all. Platforms operating in front of this problem are running daily selfies, substitution flows, and continuous monitoring. Platforms operating behind it are running annual re-checks.

Regulation operates across layers. The EU Platform Work Directive, the UK’s Right to Work extension, eIDAS 2.0, the EU AI Act’s human-oversight requirements for high-risk employment AI — none of these sit cleanly on a single layer. Each has implications for what evidence is required, how it is gathered, how it is decided, and how the audit trail is maintained.

Takeaway. If you think about background screening and identity verification as a list of products to buy, you’ll struggle to keep up with regulations and threats that don’t respect product boundaries. If you think about it as layers, the decisions get easier.

How do proactive vs reactive and point-in-time vs continuous change each layer?

The four layers describe what the approach contains. Two axes describe how it operates.

Proactive vs. reactive. A criminal records check is reactive- it tells you what’s already happened. A device-signal anomaly that flags an account-sharing event in progress is proactive — it tells you what’s happening now, in time to do something about it. Most screening programs are weighted toward reactive evidence. Most of today’s threats require proactive detection.

Point-in-time vs. continuous. Traditional background checks are point-in-time. Continuous monitoring runs in the background, on a cadence chosen by the operator. Most programs sit somewhere in between, with different layers operating at different cadences that weren’t necessarily designed together.

Plotting your background screening and identity verification approach against these two axes is one of the more useful self-diagnostics available. It tells you what you’re actually doing, as distinct from what you think you’re doing.

How are organizations and teams using this view in practice?

Three ways the layered view is being put to work, in our experience.

1. As a diagnostic. Mapping the current verification approach against the four layers and the two axes. Which layers are well covered? Which get least attention? Where is the program reactive when it could be proactive? Where is it point-in-time when continuous would catch more?
2. As a procurement frame. A vendor pitching a “complete solution” is usually pitching deep capability in one or two layers with thinner coverage of the others. Operators using the layered view organize vendor conversations around which layers matter most for their context, rather than around feature lists.
3. As a risk & safety architecture sequence. Organizations rebuilding their background screening and identity verification program tend to sequence the work: layer one first; layer two against the layer-one anchor; layer three against both; layer four as the orchestration that holds everything else together. The most common reported error is skipping layer one, investing in records and signals while leaving identity under-verified.

A few reflections for you and your team

• Which of the four layers does your current background screening and identity verification approach cover well? Which gets the least attention?
• At layer one, are you confident the person working today is the person who passed your background checks?
• At layer three, are you running anything continuously, or only checking in periodically?
• Who in your organization owns layer four, the decisioning logic itself? If you had to explain it to a regulator tomorrow, could you?

Frequently Asked Questions