Happy Webs
№ 01 / Use caseAI use case

Spot the Defect Before It Leaves the Bay.

You don't need a million-pound machine-vision rig to catch quality issues. A modern AI model running on a phone or a £400 fixed camera can spot missing welds, wrong components, surface defects, and assembly errors — instantly, consistently, on every part. We build the pipeline that turns your existing photos into a quality safety net.

№ 02The promise

The promise

< 2 sec inspection time per part
£0 extra hardware in most cases
24/7 consistent — no Monday-morning fatigue
№ 03Where it hurts

Where it hurts

The real friction we hear about.

01

QC is the most boring job in the building

Eyeballing 300 identical parts for the one with a missing weld is exactly the work humans are worst at. Attention drifts. Defects slip through. Costs end up on your warranty bill, not your QC line.

02

The good QC person leaves

Quality control depends on experience — knowing what to look for, what "normal" looks like for this product. When that person retires, quality dips for six months while you train the next one.

03

Customers want photo evidence anyway

Modern B2B customers ask for pre-dispatch photos. You're already taking them. You're not yet using them to actually inspect — they're just sitting on someone's phone.

04

Off-the-shelf machine vision is overkill

Industrial vision systems are £20k+ per station and assume a production line that runs the same product all day. For SMBs with varied work, that maths never works.

№ 04How it works

How it works

What we actually build.

01

Define what "right" looks like

A handful of reference photos and a written checklist per product type. We translate that into a structured inspection schema the AI can apply consistently.

02

Capture however you already capture

Phone, fixed camera, drone shot of a roof, scanner output. The same model handles any image source. No specialist hardware in 90% of cases.

03

Inspect against the schema

For each photo: pass/fail per criterion, with reasoning. Missing weld? Flagged. Wrong colour finish? Flagged. Damaged corner? Flagged. Clean? Through to dispatch with a confidence score.

04

Build an audit trail

Every inspection, every decision, every reason — stored against the job. When a customer raises a warranty claim six months later, you can prove the part left your facility right.

№ 05Proof

Proof

Real outcomes, not slideware.

Kingsland Fabrications

AI-driven QC inspection on fabricated parts pre-dispatch. Catches defects (missing welds, surface issues, wrong configurations) before they ship — including the cases human reviewers were previously missing on busy days.

№ 07Pricing

Pricing

Fixed fee, phased delivery.

£3,200–£10,000 for build, ~£150–£400/month for hosting + accuracy monitoring

From £3,200

Typical qc inspection build

  • Workflow audit + spec phase
  • Build, integrations, and tuning to your data
  • Live deployment in your environment
  • Monthly support and accuracy monitoring
№ 08Questions

Questions

Things people ask us about this.

01 Is this actually accurate enough to trust?

Yes — provided you set the confidence thresholds sensibly and keep humans in the loop on edge cases. We typically run with the AI handling 80-90% of inspections fully autonomously, with the rest reviewed by a human. The compound effect is that humans only look at hard cases, where they're much more reliable.

02 What about ISO 9001 and traceability?

Every decision is logged with the input image, the schema applied, the confidence scores, and the outcome. That's a cleaner audit trail than a paper QC log signed by a person who looked at 200 parts that day.

03 Does this replace our QC inspector?

It changes their job, not eliminates it. They become the person who reviews flagged cases, tunes the system, and handles complex inspections — much less boring, much more skilled. Most clients keep the same headcount and just take on more work.

04 What about parts that need physical measurement?

AI vision can't replace a vernier gauge. But it can flag "this looks visually wrong, take a measurement" — which is often the trigger that's missing today.

05 Where does the data go?

Photos and decisions stay in your environment. Models we use don't train on your data. We can run fully on-premise if that's a requirement.

Let's see if we can help.

A 15-minute chat with Chris & Kay. No slides. No pitch deck. You tell us what's on your plate; we follow up by email with real thinking.

Studio81 Penny Meadow, Ashton-under-Lyne, OL6 6EL
HoursMon–Fri · 09:00–17:00 GMT