## Most AI Vendors Give You a Slide. We Give You a Timeline. Before you sign anything with an AI vendor, ask them one question: **"What does month 1 through month 6 actually look like?"** You'll get a lot of slides. Venn diagrams. Future-state architecture charts. Case studies from companies nothing like yours. What you won't get — unless you ask, and unless you know what to do with the answer — is a realistic week-by-week picture of what the implementation looks like, when you'll start seeing results, and what the payback looks like at 6 months. This post gives you that picture. Specifically for ceramics manufacturers. --- ## What AI Actually Costs for a Ceramics Firm Let's start with the number every business owner asks about first. Based on implementations we've scoped across small and medium ceramics firms (typically 20–100 employees, 2–8 kilns), typical costs for a first AI application in 2026: | Application | Typical Monthly Cost | Typical Setup Fee | |---|---|---| | Kiln scheduling optimisation | £400–£900/month | £6,000–£15,000 | | Quality control (vision-based) | £250–£700/month | £10,000–£25,000 | | Glaze and materials intelligence | £150–£400/month | £3,000–£8,000 | | Production admin automation | £100–£300/month | £2,000–£5,000 | **The relevant range for a first AI application in a ceramics firm:** the monthly subscription after setup is typically **£300–£1,200/month**. For context, that's less than the cost of one day's kiln runtime at many medium-sized ceramics operations. Most ceramics firms starting out choose either kiln scheduling optimisation or quality control as their first application. Both have clear, measurable ROI. --- ## Case Study: Kiln Scheduling at a North Staffordshire Ceramics Firm This is the example we use because it's real, specific, and representative of what we've seen across the sector. **The situation:** A medium ceramics firm running two kilns. They were spending 2–3 hours a day manually constructing firing schedules — working out load configurations, temperature ramp rates, and changeovers. Kiln downtime between firings was running at roughly 22% of available production hours. The scheduling team was experienced but reactive — always behind, always firefighting. **The AI application:** Kiln scheduling optimisation. The AI reads historical cycle times, defect patterns by load type, and kiln configuration data to produce firing schedules that maximise utilisation. **The cost:** Monthly subscription of £500/month. Setup fee of £8,000. **The results at 6 months:** | Metric | Before | After 6 Months | |---|---|---| | Kiln utilisation | ~78% | ~92% | | Changeover time | ~3 hours/day | ~90 minutes/day | | Defect rate (differential firing) | ~6.5% | ~4.2% | | Energy cost per firing | £185 | £153 | **The payback maths — month by month:** - **Month 1–2:** Baseline measurement. AI tool reads historical data, identifies scheduling patterns. No operational change yet. - **Month 3:** AI starts producing recommended schedules. Operations team reviews and approves manually before running. - **Month 4:** Schedules running autonomously. Kiln utilisation moving from ~78% toward ~88%. - **Month 5:** Full autonomous operation. Changeover time down significantly. Energy savings showing in utility bills. - **Month 6:** Full measurable results. The picture is clear. **At 6 months, the cumulative return:** - Energy savings: approximately £1,400/month (on two kilns, ~12 hrs/day, 5 days/week) - Defect reduction: approximately £800/month (on 500 units/week, average value £20/unit, 2.3% defect improvement) - Changeover time recovered: approximately £500/month (in labour cost equivalents) **Total monthly return by month 6:** approximately £2,700/month **Monthly AI cost:** £500/month **Payback period:** Under 4 months from the start of autonomous operation. After month 4, the returns compounded. The AI continued learning, schedules continued improving, and the cost of the subscription became negligible against the operational savings it was generating. --- ## The Honest 6-Month Implementation Timeline Here's what the ceramics firm's implementation actually looked like — no marketing gloss, just the reality: | Phase | Weeks | What Happens | |---|---|---| | Scoping + diagnostic | 1–2 | 30-minute conversation. We confirm the problem is AI-suitable and scope the specific application. | | Data baseline | 2–4 | Kiln runtime data, defect logs, changeover durations — existing records used to build the baseline. Runs alongside normal operations. | | Vendor + contract | 1–2 | We issue RFQ, compare vendors, negotiate terms. You don't need to do this yourself. | | Implementation | 4–8 | Tool configured, integrated with existing systems, tested on historical data before going live. | | Shadow mode | 2–4 | AI produces schedules; operations team reviews before running. This is the validation period. | | Full operation | Week 12+ | Schedules run autonomously. Team monitors, adjusts tolerance, flags exceptions. | | **Measurable results** | **Weeks 12–24** | The first 6 months of real data. This is what the payback calculation is based on. | **The honest answer:** you won't see dramatic results in month 1. The first two months are setup, data gathering, and testing. Month 3 is when things start changing operationally. By month 6, you have enough data to know exactly what the AI is doing for your business — and the payback picture is real and measurable. --- ## What Ceramics Firms Get Wrong (And How to Avoid It) **1. "We want to use AI for our production"** This is not a project brief. "We want to reduce unplanned kiln downtime by 20%" is a project brief. The difference matters enormously. When the problem is specific and measurable — "our changeover time is running at 3 hours and it's killing our Thursday shift capacity" — you can track whether the AI is solving it. When the problem is vague, you can't tell whether it's working, and you lose confidence before the implementation has had time to produce results. **We always start with a 30-minute diagnostic.** We'll tell you honestly whether your specific problem is right for AI at this point. If it's not, we'll tell you what would need to change first. **2. Skipping the baseline measurement** AI tools learn from data. If your kiln cycle times have never been logged consistently, if your defect data is "what the QC person remembers," there's a foundation missing. A 4-week baseline measurement before any implementation costs almost nothing and takes minimal effort — but it means the AI starts with real information, not guesswork. We build this into every scoping engagement as standard. **3. Expecting month 1 results** AI implementations follow a pattern: slow at first, then accelerating. Month 1 is data gathering. Month 2 is configuration and testing. Month 3 is the first operational changes. Month 4–5 is when the operational rhythm normalises. Month 6 is when the results are measurable and real. If you're evaluating an AI investment at month 2, you're evaluating it at the wrong time. Set a 6-month milestone, not a 30-day one. --- ## Is the Timing Right for Your Ceramics Business? The specific payback at 6 months depends on your current operation — what your biggest cost is, how efficiently your kilns are running, whether you have decent baseline data. For some firms, the payback at 6 months is strong enough to proceed immediately. For others, there's preparatory work that makes the difference between a smooth implementation and a frustrating one. The entry point is a free 30-minute AI diagnostic. Bring your production data — or even just the numbers you're guessing at. We'll scope the problem, give you a realistic cost estimate, and tell you honestly whether the payback case is strong enough to proceed now. We turn down projects where we don't believe the outcome will justify the investment. We've done it before. We'll do it again if the numbers don't work. Book a free audit →