The decision made before anyone reads your CV
Picture a machinist at one of Grantham's engineering firms — fifteen years on the shop floor, a solid record, redundancy notice in hand. He finds a role advertised by a similar manufacturer, spends an evening writing up his experience, and hits send. Then nothing. No acknowledgement beyond an automated receipt. No call. No rejection letter explaining what fell short.
What he almost certainly does not know is that a human recruiter may never have seen his application at all. Between the send button and anyone's inbox sits a layer of software — an applicant tracking system, or ATS — designed to score, rank, and in many cases discard CVs before a recruitment team is involved. The filter is silent. It leaves no trace visible to the applicant, and in most cases triggers no obligation to explain.
This is no longer a quirk of large technology companies. It has become ordinary practice across UK hiring, including in the kinds of manufacturing, logistics, and food-processing businesses that form the backbone of employment around Grantham and the wider South Kesteven area. The algorithm decides first. The question worth asking here is a local one: who was that algorithm built to recognise — and is a Grantham machinist one of them?
What automated screening actually does
The mechanism itself is straightforward. When an employer posts a vacancy, the ATS is loaded with keywords drawn from the job description — specific job titles, qualifications, software names, sector terms. Every CV that arrives is parsed for those terms. Applications that match enough of them are ranked higher; those that do not are ranked low or removed entirely. In most cases, no human sets eyes on the discarded ones.
The scale of this is now considerable. By 2024, 90% of large private-sector businesses in the UK had adopted AI in their recruitment processes broadly, and 70% were running CVs through ATS software as a matter of routine. Use of AI recruitment tools tripled in a single year between 2022 and 2023 — a pace that suggests many job seekers are encountering these systems for the first time without knowing it.
More sophisticated versions go further than keyword matching. AI-augmented tools can score a candidate's overall profile, flag employment gaps as potential negatives, and in the case of video interview platforms, analyse tone of voice, facial expression, and speech patterns. Up to 75% of applications may be eliminated before any recruiter is involved, which means the filter is, in effect, the hiring decision.
The underlying calibration matters here. Most of these tools were trained on historical hiring data drawn predominantly from white-collar, office-based recruitment. The keyword libraries, the CV formats they expect, and the career patterns they reward were built around a particular kind of applicant — one whose working life looks quite different from that of a machinist or a food-processing line supervisor.
Why trade CVs fail the algorithm
Take a single job title. A worker applies for a machining role and writes 'CNC Machinist' on his CV — an accurate description of what he does, the term his colleagues use, the phrase that appears on his payslips. The ATS, however, has been loaded with the keyword 'Computer Numerical Control Operator'. The two mean precisely the same thing. The system does not know that. His application is ranked low or removed before anyone notices the discrepancy.
This is the first and most common failure mode: terminology mismatch. Manufacturing job titles evolved on the shop floor and in trade directories, not in corporate HR systems. The keyword libraries inside most ATS platforms were assembled from white-collar job descriptions and existing databases weighted toward office-based hiring. The vocabulary gap is not the worker's error — it is a calibration problem baked into the tool.
The second failure mode is formatting. Trade workers often use simple, single-column CV layouts — sometimes a single printed page, sometimes a scanned form from a previous employer. Many ATS parsers struggle with anything outside a standard multi-section template: tables, text boxes, and scanned images are frequently misread or stripped entirely, leaving a blank or garbled record that scores near zero regardless of what it contained.
The third concerns credentials. Shorthand that is universally understood on a building site or factory floor — CSCS card, NVQ Level 3, CPCS ticket — may not match the spelled-out string the system is looking for. A worker who has accumulated years of accredited training can appear, to the algorithm, to have none.
Underneath all three failures sits the same structural fact: a career built through on-the-job progression, brief application forms, and practical assessment produces exactly the kind of CV that keyword logic was not designed to read. The algorithm does not see incompetence. It sees an absence of the right words — and discards accordingly.
Grantham's labour market caught between two transitions
Greater Lincolnshire's manufacturing sector contributes £3.4 billion and roughly a fifth of the region's total economic output. Grantham sits at the centre of that base: advanced engineering, automotive components, metal processing, and food processing are the pillars of local employment, and the workforce that fills those roles has, for generations, been recruited informally — through agency referrals, a word at the gate, a brief form at the HR office.
That is changing simultaneously on two fronts. Local firms are adopting automation and robotics to address skilled labour shortages — a rational response to a tight market, but one that reshapes what the remaining human roles look like and narrows how many of them exist. At the same time, those firms are moving to online application portals and ATS-filtered shortlisting as standard recruitment practice. A worker made redundant from a production line who applies for a new role at an engineering firm down the A1 may now be filtered not by a recruiter who knows the sector but by a keyword system that does not.
The timing creates a specific compounding problem. A machinist or line operative who has spent fifteen years being hired by word of mouth and a phone call has had no reason to learn CV optimisation — and, as the next section addresses, no clear institutional pathway to do so. The workers most likely to be displaced as automation reshapes the factory floor are, structurally, the same workers whose application documents are least likely to pass an algorithm calibrated on white-collar hiring data (no named local employer has been publicly linked to a specific platform, but the sector-wide failure modes are well-documented). The exposure is structural, not incidental. It follows directly from how this workforce was recruited, trained, and retained.
The digital skills gap that closes the exit
The obvious response to all of this is simple enough: learn to write an ATS-optimised CV. Guides exist. Tutorials are free. The information is out there.
What is harder to find, in South Kesteven, is a clear institutional route to that knowledge for the workers who need it most.
Lincs Digital is Lincolnshire's principal community digital inclusion provider — the organisation designed to bridge exactly this kind of gap. South Kesteven District Council does not appear among its partner network. That absence is not a technicality: it means the region's main community digital skills infrastructure does not formally extend to Grantham's local authority area.
Funding does exist in other forms. South Kesteven's Economic Development Strategy 2024–2028 lists digital innovation as a UK Shared Prosperity Fund priority, and interventions have been commissioned through Grantham College and Steadfast's Connect2Grow programme. But coverage is partial and self-directed uptake is low — conditions that consistently favour those already confident online, not those encountering digital job applications for the first time.
The people the TEDx Grantham investigation identified as most exposed — adult workers in low-wage roles, sole traders, micro-employers — fall structurally outside these pipelines. A machine operative or plumber who has never written a digital CV, let alone calibrated one for keyword scanning, has no signposted route to acquire that skill. Connect2Grow and Grantham College's provision together do not close that gap. That is a policy and infrastructure problem, and the workers left outside it are not a marginal case — they are the backbone of the local economy.
Bias, regulation, and what workers can realistically do
The filtering problem goes beyond keyword mismatches. Documented research shows algorithmic systems tend to penalise employment gaps, non-linear career paths, and class-adjacent signals such as postcode — patterns that map closely onto the workforce profile already described. University of Melbourne researchers found AI recruitment tools discriminated against applicants with names perceived as Black, those requesting disability adjustments, and non-native English speakers. Amazon's résumé screening tool — now a standard reference point in the literature — downgraded CVs containing the word 'women's' and penalised graduates of all-women's colleges, despite having been built to improve efficiency, not encode prejudice.
UK equality law offers partial cover. Discrimination on grounds of race, disability, and sex can, in principle, be pursued — but socio-economic status is not a protected characteristic under the Equality Act 2010. Class-adjacent signals encoded in algorithms (postcode, employment history, non-elite education) are, for now, legally harder to contest.
On paper, regulation exists to address this. The GOV.UK March 2024 Responsible AI in Recruitment guide, developed with the ICO and EHRC, obliges employers to build in meaningful human review for automated decisions and to audit for bias; it even identifies digital exclusion as a novel risk created by these tools. But the guidance works through employer compliance, not external inspection — a meaningful gap when the workers most affected are the least likely to know what questions to ask.
Practical steps can reduce the risk of being screened out: spell credentials in full rather than abbreviating (National Vocational Qualification, not NVQ), mirror the exact wording used in the job advertisement, and use a plain single-column format with no tables or text boxes. The difficulty is not that the advice is hidden. Knowing what to do and having the time, confidence, and support to do it are rarely the same thing — and for Grantham's workforce, that gap is as structural as any the algorithm creates.
- [1] Fair by Design? The Legal and Ethical Challenges of Algorithmic Hiring. (2025). https://doi.org/10.56397/slj.2025.08.01 https://doi.org/10.56397/slj.2025.08.01
- [2] A Comprehensive Review of AI Techniques for Addressing Algorithmic Bias in Job Hiring. (2024). https://doi.org/10.3390/ai5010019 https://doi.org/10.3390/ai5010019