Updated
Industrial AI that runs on the factory floor.
Most UK manufacturers do not need an AI transformation programme. They need one painful, repetitive workflow handled by AI properly - RFQ triage, sales engineering knowledge retrieval, technical content production, supplier comms. I help engineering and operations teams pick that first workflow, build it, and hand it over so the team can keep using it without me.
SME vs enterprise note: "one narrow first workflow" applies to mid-market UK manufacturers (typically up to about £500M revenue). Enterprise manufacturers - Emirates Global Aluminium, Ecolab, Pfizer Global Supply, SOCAR - run portfolio AI programmes in quarterly waves and deliver $100M-class outcomes by design. That work needs a different governance model and a dedicated internal digital leader. If you are at that scale, McKinsey/QuantumBlack case studies and the AI for Manufacturing case-study library are better starting points than a solo consultant.
Five workflows where AI earns its keep.
Not "AI for everything". Not "let us run a six-month transformation". Five concrete jobs that come up in almost every UK manufacturer I talk to.
| Workflow | Manual today | After | Time back |
|---|---|---|---|
| RFQ triageSales + estimating | Inbox of PDFs, drawings, emails. Estimator reads each one, categorises, drafts a response, chases missing info. | AI extracts requirements, flags missing info, drafts the response. Estimator reviews and sends. | ~2 hrs/RFQ→15 min |
| Technical content productionMarketing + engineering | Engineer writes a draft, marketing rewrites for the web, both go back and forth for weeks. | AI drafts from engineering's source notes, in the brand voice, ready for engineer fact-check. | ~3 wks/page→3 days |
| Sales engineering knowledge retrievalPre-sales support | Sales rep asks engineering team the same questions repeatedly. Engineering loses hours to internal support. | AI answers from your real engineering docs, datasheets, past quotes. Engineering only fields the genuinely new questions. | Endless→Instant |
| Supplier communicationsProcurement + ops | Standard order acknowledgements, delivery confirmations, certificate-of-conformance chases all done by hand. | AI handles the standard exchanges. Procurement handles the exceptions and the relationships. | ~5 hrs/wk→30 min |
| Spec sheet generationEngineering documentation | New product variant means a new datasheet. Engineering writes from scratch, formats it, gets it signed off. | AI generates the first draft from the parent product datasheet plus the variant deltas. Engineer reviews technical accuracy. | ~1 day→1 hour |
Not on the list? Tell me what your team repeats and I will tell you whether AI is the right fix.
Why most manufacturing AI projects never make it past the pilot.
Blockers that kill more manufacturing AI pilots than the technology does. None are unsolvable. All need addressing before any workflow goes live.
Legacy system integration
Your ERP is from 2009 and the MES is bespoke. AI without a data feed is a clever demo. The first work is usually a clean integration layer - API or file-based - that the AI can read reliably.
Data fragmentation
Spec sheets in PDFs, drawings in DWG, pricing in Excel, customer history in CRM. AI needs a coherent view. The fix is usually less ambitious than a data lake - just a working knowledge base.
IP and compliance friction
You cannot just paste customer drawings into ChatGPT. The deployment decision - public API, private cloud, local model - is made up front based on what data the workflow touches.
Workforce adoption
If the engineer who uses the tool was not in the room when it was designed, it will not get used. The build process includes the people who do the work, not just the people who buy the software.
How one workflow gets built and handed over.
Not a six-month transformation. One workflow, properly built, properly handed over.
1. Process mapping
I sit with the people doing the work. Map every step, every handoff, every decision point. The map shows where AI fits and, just as important, where it does not.
2. Prompt and workflow design
Prompts, examples, system rules, exception handling. The workflow gets built around the decisions a human still needs to make, not around the AI's capabilities.
3. Human review checkpoints
Every workflow has explicit review points. The team knows what AI can ship without approval and what always needs a human eye. No automation creep, no surprises.
4. Hand-over docs for engineering
Operating guide, prompt library, model selection notes, change-management approach. The team can run, maintain and extend the workflow without me.
The kind of work the portfolio backs up.
My background is operations and search consulting at scale - 3M technical SEO audit, BlackRock enterprise content, Unilever brand work, E.ON regulated B2B and B2C. AI workflow consulting is the same discipline applied to a newer toolset.
Other UK manufacturing AI consulting options - and when each fits.
The UK manufacturing AI market splits into four model types. Four options worth knowing before commissioning anyone, including this page. Honest framing because Consultancy.uk's "Diamond / Platinum / Gold" tiering is essentially a paid-placement ranking.
| Option | How it is structured | Best for |
|---|---|---|
| UK industrial AI specialistGravitonic (industrial AI strategy), B13 (Birmingham, "AI for Manufacturers" guide) | Small to mid-sized UK teams focused on operational AI - predictive maintenance, computer vision for quality control, supply chain optimisation. Sit between strategy consultancy and data science. | Manufacturers ready to deploy operational AI use cases (shop floor, supply chain, QC) with senior strategy input. Less ideal if your need is broader AI literacy and workflow design rather than specific industrial models. |
| SMB-focused regional AI agencyZoravar (Birmingham SMB - "60% cost reduction, 90-day ROI"), Geeks Ltd | Smaller regional agencies positioning on affordability, fast ROI claims, chatbots + basic automation + analytics. Aggressive marketing positioning. | Manufacturers wanting to start with chatbot and basic automation work at sub-£3k/mo. Verify the ROI claims with case studies before committing - "60% cost reduction in 90 days" is rarely sustained at that level. |
| Enterprise consultancy with AI practiceProtiviti, Bell Integration, Big Four (Accenture, Deloitte, KPMG, EY) | Multi-hundred-person teams, AI alongside many other practices. Heavy partner-juniors model, six-figure minimum engagements, multi-month discovery before any build. | Large manufacturers with £100k+ engagement budgets, regulatory complexity, multi-site rollouts, and internal procurement teams that prefer named brand consultancies. Overkill and slow for mid-market. |
| Senior independent consultantThis page - 3M, BlackRock, Unilever, E.ON enterprise background applied to AI workflow design | One senior IC working directly with operations or marketing leadership. AI workflow design, prompt engineering at scale, agentic tooling integration, AI search positioning - the side of AI most enterprise consultancies under-staff. | Manufacturers wanting senior judgement on which AI investments are real versus hype, and how to embed AI into existing workflows rather than buy another tool. Best when the value is senior input, not headcount. |
If you want help reading an AI consulting proposal before signing, send the scope and I will tell you what is real work and what is consulting theatre.
Find it. Build it. Embed it. Hand it over.
Map the workflow
Sit with the team, map the current process, pick the candidate with the clearest inputs, clearest outputs and lowest blast radius if AI gets it wrong.
Workflow plus review points
Prompts, model selection, integration, exception handling. Every step has a defined human checkpoint where it matters.
Run it with the team
The first month is supervised. The people who will use the tool are in the room. Drift gets caught early, the prompts get tightened, the operating guide gets written from real use.
Leave a working system
Operating guide, prompt library, model notes, change log. The team runs it without me - and knows how to fix it when something shifts.
When AI is the right answer. When it isn't.
Good fit
- Mid-size UK manufacturer with at least one painful, repeating workflow
- Internal team that wants to learn, not outsource the thinking
- Existing data sources - ERP, CRM, file shares - that the AI can read
- Leadership willing to start small and scale only if version one delivers
- A first project where mistakes cost time, not money or safety
Bad fit
- An enterprise looking for a company-wide AI transformation programme
- A first project touching production control, safety systems or financial controls
- An organisation expecting AI to replace headcount within months
- A team with no internal champion who will use and own the tool
- An unwillingness to discuss data residency, IP and compliance up front
The questions manufacturers ask first.
Will this work with our ERP?
Yes. The integration approach is API-first where the ERP exposes one (most modern systems do), or file-based where it does not. Either way the AI workflow sits alongside the ERP, not inside it.
What about IP and data leakage?
The deployment decision gets made up front based on what data the workflow actually touches. Public model API, private cloud deployment, or local model - each has different IP and compliance trade-offs. For sensitive drawings, specs or customer data, local Llama or Mistral models running on your own hardware is the usual answer.
Can shop-floor staff actually use this?
They have to. If the tool requires a degree in prompt engineering to operate, it does not get used. The handover includes operator-level training and a single-screen interface for the people doing the work.
Do you understand Industry 4.0 and the wider digital transformation picture?
Yes - but this work is narrower than that. Industry 4.0 is a multi-year programme. This is one painful workflow, automated properly, that pays for itself in months. Both are valid - this is the more practical starting point.
How does this fit with our existing MES or SCADA systems?
The AI layer typically reads from these systems rather than writing to them. Predictive flagging, anomaly detection, automated report generation. The system of record stays the system of record.
What is the first AI project a UK manufacturer should run?
For most, it is RFQ triage or technical content production. Both are high-volume, repetitive, and have clear inputs and outputs. The wrong first project is anything involving production control or safety - those need a different governance model entirely.
Tell me what your team repeats.
Half an hour to map the workflow that hurts most. By the end of the call you know whether AI is the right fix, the wrong fix, or the right fix for a different problem you have not raised yet.
Book a call