Is AI missing the best talent?

AI can speed up hiring. It can also narrow your talent pool in ways you might never see.

When screening tools rely too heavily on keywords, formatting or historic hiring criteria, strong candidates can disappear from the process before a hiring manager ever sees them.

For employers dealing with high application volumes, particularly across FMCG manufacturing, that creates a difficult question: is your technology helping you identify the best people, or simply the people whose CVs are easiest for it to understand?

The bias problem nobody budgets for

Automation is very good at managing volume. Understanding context is harder.

Early-stage filters such as CV parsers, ranking tools and automated screening can be influenced by the criteria they have been given and the data or rules behind them.

That matters because good candidates do not always describe equivalent experience in equivalent ways.

A hiring process can therefore become very efficient at applying criteria that are too narrow.

As Libby Noble, Business Manager for Supply Chain, Operations & Engineering at Denholm Associates, explains:

“If you want to test your AI, run a quick false-negative audit. Pull a random set of auto-rejected applications from last quarter, mask identity and have hiring managers re-score them against outcomes, not pedigree. If you find even a handful who should have advanced, it’s time to review.

AI should find more great people, not hide them. That means human-in-the-loop decisions, fresher training data and a skills-first mindset.”

What can get lost in FMCG manufacturing?

Application volume can be high across FMCG manufacturing, while finding people with the right practical experience remains difficult.

Automated screening can help teams deal with that volume, but the way criteria are configured matters.

Think about how easily these candidates could be missed:

• A factory candidate has 15 years of relevant experience but a poorly structured CV.
• A Supply Chain Analyst and a Forklift Driver are assessed using broadly the same screening logic.
• A migrant worker describes relevant experience using different terminology or imperfect written English.
• A creative or non-standard CV creates problems for a parsing tool.
• An experienced frontline manager writes “batch runs” where the system expects “production cycles”.
• A maintenance technician describes “fixing downtime issues” rather than “equipment optimisation”.

The underlying experience may be there. The language the system expects might not be.

Define what good actually looks like

Better screening starts with a better definition of success in the role.

Hiring managers can play an important part here because they understand what good performance looks like on the ground.

Instead of relying heavily on years of experience, qualifications or exact terminology, consider the skills and outcomes that demonstrate whether someone can actually do the work.

For example:

• Can they run a line changeover effectively?
• Can they model inventory scenarios under variable demand?
• Have they reduced scrap, downtime or waste?
• Have they improved equipment reliability?
• Can they demonstrate experience in a comparable production environment?

That gives screening tools a richer picture of what to look for and gives hiring managers a more useful basis for assessing candidates.

Teach your screening process to recognise context

Equivalent skills can be described in very different language.

If strong Line Leads in your business use “batch runs” while your system looks for “production cycles”, that should tell you something about the screening criteria.

The same applies to transferable skills. Someone may have developed highly relevant experience in an adjacent environment without using the terminology associated with your sector or organisation.

Reviewing a sample of near-miss and rejected applications can help identify where those gaps occur.

That insight can then be fed back into your screening process so that the system becomes better at recognising the experience that matters.

Keep human judgement in the process

Automated screening can support decision-making, particularly when application volumes are high. Employers should still understand where automated decisions are being made and what those decisions are based on.

A useful approach is to regularly review a sample of rejected or near-miss applications with hiring managers.

Ask:

• Would we genuinely have rejected this person?
• What experience did the system miss?
• Are we applying appropriate criteria for this particular role?
• Are different types of candidates being screened fairly?

Over time, that creates a feedback loop between the technology and the people who understand the work.

A practical 90-day reset

You do not necessarily need to rebuild your recruitment process to find out whether automated screening is helping or hindering you.

Weeks 1–2: Map the hiring flow

Identify where applications are automatically rejected or deprioritised. Pull a sample of recent near-miss applications across different roles and ask hiring managers which candidates they would reconsider and why.

Weeks 3–4: Redefine success

Review success profiles for priority roles. Capture the skills, outcomes and terminology hiring managers associate with strong performance, including equivalent language used by people already doing the work.

Weeks 5–6: Look for patterns

Review pass-through rates at different stages of the process. Look for criteria that may be unnecessarily narrowing the pool, including rigid keyword rules, qualification requirements or knockout questions.

Weeks 7–8: Ask your technology providers better questions

Understand how the tools you use are evaluated, what information is logged and what control you have over screening criteria and automated decisions.

Weeks 9–12: Pilot and learn

Test the revised approach within one site, role family or function. Monitor shortlist quality, interview-to-offer ratios, time-to-shortlist and candidate feedback before deciding what should be rolled out more widely.

Compliance and candidate trust

Employers also need to understand their responsibilities when automated tools are involved in recruitment decisions.

Transparency about where automation is used, appropriate oversight of decision-making and a clear route for human intervention can all help create a more accountable process.

The Information Commissioner’s Office provides guidance on AI and data protection, including resources for organisations assessing AI-related risks.

Make your technology work harder for your talent strategy

The opportunity with AI is to build a recruitment process that can handle scale while still recognising the skills, experience and potential that matter.

Denholm works with employers to understand the talent they need, reach people beyond the obvious applicant pool and design hiring approaches around the realities of the market.

Explore our Talent Strategy & Delivery Models, learn more about Recruitment, or explore our FMCG and Supply Chain, Operations & Engineering expertise.

If your recruitment process is generating plenty of applications but struggling to surface the people you actually need, call us on 03303 359 818 or email connect@denholmassociates.com.

 

You might also be wondering...

What should you ask an AI recruitment technology provider before choosing a system?

Ask how the system makes or supports screening decisions, what data informs those decisions, what you can configure, how outcomes are monitored and whether you can review why a candidate was rejected or ranked in a particular way. You should also understand how the provider tests its system and what support is available when your hiring criteria change.

How can you tell if your ATS is rejecting good candidates?

Look backwards. Take a sample of candidates who were rejected or ranked poorly before human review and ask experienced hiring managers to assess them independently against the actual requirements of the role. Patterns in what was missed can reveal where screening criteria are too narrow or poorly calibrated.

Could AI screening disadvantage candidates with non-traditional career paths?

It can if the screening criteria rely heavily on conventional job titles, qualifications, career progression or exact experience. Employers can reduce that risk by defining the skills and outcomes that matter and considering how relevant experience might have been gained in different roles, sectors or environments.

What should employers tell candidates about AI in their recruitment process?

Employers should be clear about how candidate information is being used and understand the relevant data protection requirements. Where automated tools influence recruitment decisions, transparency can also help candidates understand the process and what happens to the information they provide.

Who should be responsible for checking whether AI recruitment tools are working properly?

Responsibility should sit with the employer rather than being left entirely to the technology provider. Talent or HR teams, hiring managers and those responsible for data protection or governance may all have a role, depending on how the technology is being used. Hiring managers are particularly useful in testing whether automated criteria reflect what successful performance actually looks like.

Can AI help identify transferable skills?

Potentially, but the quality of the result depends on how the system has been designed and configured. If it can recognise related skills, equivalent terminology and relevant experience from adjacent roles or sectors, it may help broaden a search. If it relies heavily on exact keywords or historic hiring patterns, transferable experience can be easier to miss.