ATS, test workflow and ownership when recruiting teams build their own AI tools

On 28 October 2026, HR Dive reported from the Gartner HR Symposium/Xpo, where Jamie Kohn, senior director analyst in Gartner's HR practice, set out four trends shaping talent acquisition right now. One of them is about who owns the technology. According to Kohn, recruiters can now pick up new features from the vendors they already use, from the company's general enterprise AI tools, and from no-code and low-code options that let them build what they are missing themselves. HR technology leaders said they are building AI agents inside their HCM platforms. The downside is duplicate tools, inconsistent processes and technical debt. Kohn's summary was that the problem is not a shortage of innovation but the difficulty of keeping up with it.

The part that matters most for assessment is her distinction between what recruiting teams can reasonably build on their own and what should be managed centrally. Labour market analysis tools are fair game for a team to build. Candidate matching and rediscovery need more oversight. And assessments, as a general rule, should be enterprise-managed, because the risk reaches well beyond the individual team.

The article gives no sample sizes or methodology behind the Gartner figures, so they should be read as an analyst's view rather than a measured effect.

It can sound like an organisational footnote, but this is exactly where many test workflows break. When a recruiter can attach an AI function to the ATS without going through procurement or IT, it becomes tempting to solve a local problem locally. One team drowning in applications builds its own filter. Another sets up its own AI interview. Each step can be perfectly reasonable on its own. The trouble starts when the test, validated against specific criteria and designed for a specific place in the process, suddenly sits behind a filter nobody at the centre knows about.

Kohn also said that, by her account, application volume has more than doubled in two years, and that the barrier against AI in screening and assessment is weakening. Recruiters will use AI whether it is approved or not. Her advice was to weigh high-volume, business-critical hiring against the situations where AI cannot reliably or fairly judge qualifications. Risk rises when skills are hard to observe and when it is unclear what success in the role looks like. Stable, repeatable roles with clear criteria are, by contrast, the lower-risk places to expand AI use, AI interviews included.

That is a useful rule of thumb, but it assumes someone has an overview of where AI already sits in the process. That overview is precisely what disappears when tools are built team by team.

A day later, on 29 October, HR Dive described how Gap Inc. has launched an AI-enabled career guide called Bridgette for job seekers across Old Navy, Athleta, Banana Republic and Gap. The tool is built on technology from Paradox and Workday Recruiting. It helps candidates find and apply for roles, answers frequently asked questions, schedules interviews and sends reminders. It has been piloted since June at more than 100 stores and six regional distribution centres, where nearly 45,000 people have used it, and it has scheduled more than 7,000 interviews. Gap expects around 200,000 applications for seasonal roles.

The article mentions neither tests nor screening, and that is a data point in itself. At that volume the question is not whether a tool sits in front of the interview, but which one, who owns it, and whether it lets any test do its job or quietly makes it redundant.

Who is allowed to attach something to the ATS? If the answer is anyone with access to a no-code platform, you do not have one test workflow, you have several.

Where does the test sit relative to the newest AI steps? If a locally built filter sorts candidates before the test goes out, the test's documented validity no longer covers the population that actually reaches it.

Is assessment defined as enterprise-managed? Gartner's recommendation is clear on this point, and it belongs in your governance, not only in a conference slide.

Are you measuring the right thing? Kohn suggested slate quality, time to productivity and candidate confidence as alternatives to the old speed metrics. A test workflow measured only on time to hire will be pushed to skip steps.

The news from Gartner is not that AI is entering recruitment. It is that it is coming in through several doors at once, and someone needs to hold the key to the door the test stands behind.

Kilde: https://www.hrdive.com/news/4-trends-explaining-why-recruiters-are-struggling-right-now/873036/ and https://www.hrdive.com/news/gap-ai-hiring-tool-bridgette/873179/