Rebuilding trust in hiring: why AI isn’t the problem

AI isn’t the problem with trust in hiring; the problem is hiring processes designed so that nobody is visibly accountable for what happens to a candidate. AI can make that worse or better, and which one it does depends on design choices about who decides, what the candidate can see and whether the experience feels like it was made by someone who cared.

Trust went before the AI arrived

It is tempting to date the loss of trust to the arrival of generative AI, but most candidates could have told you about it years earlier. The application that vanished. The automated email that addressed them by the wrong name. The rejection with no reason, or no rejection at all. Hiring had already become a process that candidates experienced as a series of anonymous touchpoints, fast but flat, with no sense of a person on the other side.

AI arrived into that and got the blame, partly because it is new and partly because it is a convenient name for a feeling people already had: that nobody is really looking. The fear is not that a model will read your application. It is that nobody will, and that a decision will be made that nobody can explain or will own.

That distinction matters, because it tells you where to put the work. If the problem were AI, the answer would be less of it. Because the problem is accountability and care, the answer is design.

Decide who decides

The first design choice is the one that matters most: what the software is allowed to do. At inploi the answer is fixed. Agents do the admin, humans do the hiring. The Hiring Agent scores applications with reasoning a recruiter can read, and its role stops at a recommendation. When a decision is made, it is made by a person, in the ATS you already use.

That is not a limitation we tolerate; it is the point. A candidate can trust a process where a named human owns the outcome, even if software helped that human read two hundred applications before lunch. They cannot trust one where the decision belongs to nobody. Every setting in Studio is versioned and attributed to the person who made it, for the same reason: when something goes wrong, you can find out what happened and who can put it right.

Not all friction is bad

The second design choice is less obvious. The industry has spent a decade treating friction as the enemy, as if the ideal application were one click and the ideal decision instant. Some friction deserves to go: a forty-minute form for an hourly role, a mandatory account, a CV upload that fails on a phone. But removing every step can remove the meaning with it.

In most decisions that matter to us, choosing a flat, a holiday or a school, people want to explore. They want to understand what they are choosing and to feel that the choice was theirs. A job is no different. A candidate who has looked at the shift pattern, understood what the role involves and answered a couple of questions that are genuinely about the job is more committed than one who fired off an application in four seconds, and the employer learns more from them too.

So we build for deliberate friction. Conversational apply asks the questions a role needs, in the right order, and lets a candidate who is not eligible find out before they invest time. The Candidate Agent lets people ask what they actually want to know before they apply, and helps them decide whether to apply at all. Those are small steps, but they are steps that build understanding on both sides, which is what trust is made of.

Taste is a signal of care

The third choice is about how the experience feels. Hiring technology has converged on sameness. Most careers sites follow the same template, most applications ask the same questions in the same order, and the result is an experience that looks familiar and says nothing. Benchmarking against competitors produces exactly this: everyone ends up in the middle.

Candidates read sameness as indifference. The opposite of indifference is taste, meaning the evidence that someone made deliberate choices about this experience for these people. It shows in copy that sounds like the brand, in careers sites that load quickly on a phone and in the details a candidate notices without being able to name them. We approach it the way you would approach a consumer product rather than a procurement exercise, because for a frontline candidate the careers site is often the first real contact with the employer, and often the same brand they buy their coffee or book their holiday from.

Taste is not decoration. It is how an employer signals, before a single conversation, that the person on the other side matters.

Where AI helps rebuild trust

Designed this way, AI becomes part of the answer. It can tell a care worker between home visits whether the role needs a driving licence, on a Saturday when otherwise nobody would have replied until Monday. It can make sure every application is read against the same criteria, rather than the first forty getting attention and the rest getting none. It can write down its reasoning so a recruiter can challenge it. And it can take enough of the admin away that the recruiter has time for the conversations that really build trust.

None of that works if the AI is a black box making final decisions. All of it works if a human owns the outcome, the candidate understands what is happening and the experience shows care.

What to do next

Look at your own process and ask three questions. For every decision about a candidate, can you name the person who owns it? Does every step you ask candidates to take teach either side something useful? And would a candidate recognise your careers site as yours with the logo covered up? The answers will tell you more about trust in your hiring than any view on AI.

For more on what trustworthy AI in hiring looks like in practice, read our trust in hiring collection, or book a demo to see how the pieces fit on your ATS.