How to Choose an AI Consultant for Your UK Manufacturing Business (Without Getting Burned)
The AI Consultant Market is a Minefield
Every consultancy claims AI expertise. Most are repackaging generic software implementations dressed up in machine learning terminology. You're sitting in a meeting room with someone in a sharp suit talking about "enterprise AI platforms" and "operational transformation strategies," and you're thinking: is this actually going to move the needle on my quoting bottleneck, or am I about to spend six figures on a PowerPoint deck?
You're right to be sceptical. The truth is that 73% of AI projects never reach production. They're abandoned. They sit in the cloud gathering dust while your team is back to the old process, and your budget is gone.
The difference between a consultant who delivers and one who consumes your budget without results comes down to a few critical factors: do they understand your commercial reality? Can they show you a working prototype in weeks, not months? Do they have a guaranteed ROI model, or are they selling vaporware?
This guide cuts through the noise. It's what you need to know before you sign a contract with anyone claiming to be an AI expert.
The Three Types of AI Consultant
Not all AI consultants are created equal. Understanding the category you're talking to is half the battle.
1. Big Consultancies: Accenture, Deloitte, PwC
The brand names. The blue-chip firms. The ones with offices in every major city and glossy case studies.
What they do well: they're credible. They have enterprise experience. They can handle massive implementations with hundreds of staff and complex governance structures.
What they're wrong for: you. If you're a £50M manufacturing business with 80 staff, you are not their core market. Their minimum engagement is typically £250k to £500k. They staff projects with expensive senior consultants who hand off to junior analysts once the scope is defined. They move slowly because their delivery model is designed for FTSE 100 clients with 18-month programmes and formal governance committees. Your six-week prototype timeline? They'll tell you it's not feasible. Their discovery phase alone can cost £50k and tell you nothing you don't already know.
They're also generalists. They do AI, process overhaul, supply chain, HR. They're not manufacturing specialists. They'll spend the first six weeks learning your business.
2. Tech-First Boutiques
These are the sharp technical people. Software engineers who love building cool things. They speak in frameworks, architectures, and model performance metrics. They're usually ex-FAANG or ex-startups. They've read the latest papers on neural networks. They can probably implement a transformer-based language model before breakfast.
What they do well: they build. They ship. They understand software architecture. They love hard problems.
What they miss: commercial reality. They'll build you a technically beautiful solution that nobody uses because your sales reps won't change their workflow. They optimise for AI performance metrics, not for business outcomes. They often don't have manufacturing experience, which means they're solving problems as generic AI problems rather than as commercial manufacturing problems. Their timeline estimates are optimistic. Their budgets overrun.
3. Operations-First Partners
These are consultants who understand your business first and build second. They've worked in manufacturing or B2B sales. They speak in your language: margin, throughput, revenue reclaimed, capacity. They understand that the best AI system in the world is worthless if your sales team won't use it. They build working prototypes because they know that seeing something work beats reading about it. They measure success in money, not in model accuracy.
They're typically smaller, more specialized firms. But they're the ones who deliver.
Ell Advisory sits in this category. We started as operators. We understand quoting bottlenecks, sales capacity constraints, and the commercial drag of manual processes. We build working prototypes in weeks. We guarantee ROI. We understand that your sales reps won't change their behaviour for a beautiful UI, so we build invisible UI, where the system works around their existing workflow.
Red Flags: What to Avoid
Before you meet with anyone, know what to watch for. These are the warning signs that someone is going to waste your time and money.
"We'll start with a discovery phase." Translation: we're going to charge you £30k to £50k to do research that you could do yourself. You need discovery, yes. But it shouldn't cost more than £2k to £5k, and it should result in a clear, actionable roadmap. If the consultant is asking for a five-figure discovery fee, they're either gold-plating the work or they're planning to use it as a wedge for a massive follow-on engagement.
They talk about "AI strategy" before they ask about your sales process. If they're selling you an AI strategy before they understand how you sell, they're selling strategy, not solutions. A good consultant will ask: how many quotes do your reps write per week? How long does each quote take? Where are the manual steps? What's your win rate? If they skip this and jump to "let's build a unified data lake," walk away.
No manufacturing experience. Manufacturing has unique constraints: technical specifications, complex quoting, regulatory compliance, long sales cycles. A consultant who's only worked in SaaS or insurance will miss these. They'll build something elegant that doesn't fit your world.
They can't explain ROI in your language. If they're talking about "model accuracy" and "training data" instead of "revenue reclaimed" and "hours saved," they're not thinking about your bottom line. A good consultant will tell you upfront: if we implement this, you'll reclaim this many hours per rep per week, which at your burdened cost is worth this much per year. Guarantees might be hard, but the framing should be commercial.
They want to replace your CRM. Your CRM is fine. It does what it does. Rip-and-replace projects are death spirals. Any consultant pushing a CRM migration as part of an AI project is solving for their own complexity, not yours. Stay away.
They're quoting a 12-month timeline for a quoting assistant. Twelve months? For an AI system that helps your reps quote faster? That's not a project; that's a consulting career. A solid team should be able to show you a working prototype in six weeks. Full production rollout in 16 weeks. If they're asking for a year, they either don't know what they're doing or they're stuffing scope.
There's no commercial guarantee. "We'll build you a system and hope it works" isn't a partnership; it's a services engagement with one-way risk. You want someone who says: "If we deploy this and you don't see these specific commercial results, we'll keep working until you do."
What Good Looks Like
So what should you be looking for instead?
They speak your language. Not jargon. Commercial language. They talk about sales capacity, margin per order, revenue reclaimed, quoting throughput. They understand that a minute saved per quote multiplies across 50 reps and 200 quotes per week. They do the maths in your presence, not in a PowerPoint deck.
They offer a low-risk entry point. A good consultant will start with a workshop (£1.5k to £3k) or an audit (£5k to £10k). Not to nickel-and-dime you. To prove value quickly and build trust. They want you to see something work before you commit to a larger engagement. This is the signal of someone confident in their work.
They show you working prototypes, not PowerPoints. They'll sit you down and show you a system that actually works on your data, with your spec, handling a real quote scenario. Not a mockup. Not a slide deck. Not a "what it could look like." A thing that works. This is non-negotiable.
They have manufacturing case studies with numbers. Not anonymized success stories. Real numbers. "We reduced quoting time from 45 minutes to 12 minutes per quote for a £40M precision engineering firm, reclaiming 351,000 hours per year across their sales team." Or: "We built a specs-to-quote system for a mid-market fabricator that increased their quote success rate from 31% to 67%, adding £10.1M in value." Numbers. Specificity. Manufacturing context.
They understand invisible UI. They know that you can't ask your reps to change their workflow. So they don't. The system integrates into the tools your reps already use. Email, CRM, spreadsheets, whatever. Your reps keep doing what they're doing. The AI system does the heavy lifting invisibly. No behaviour change required.
They're honest about what they don't know. If you ask about integration with some obscure legacy system and they say "I don't know, but we'll figure it out," that's better than "absolutely, no problem" when they haven't worked with it before. Honesty scales.
They offer protection against failure. At minimum, a time guarantee: "If we're not in production by this date, the engagement converts to time-and-materials at a reduced rate." Better yet, a commercial guarantee: "If we deploy this and you don't see X hours reclaimed, we keep working."
The 10 Questions to Ask Before You Sign
Here's your checklist. Ask every consultant these questions. Their answers will tell you everything.
1. What's your manufacturing experience? Are they coming to manufacturing fresh, or have they lived it? How many manufacturing clients do they have? In what subsectors? Can they name a specific customer and talk about what they solved? Vague answers are a bad sign.
2. Can you show me a working prototype? Not a mockup. Not a "here's what it could look like." Something live. Running on data that looks like yours. Ideally running on your actual data. A consultant who can't show a prototype is selling theory, not results.
3. What happens if we don't see ROI? Do they have a guarantee? What's their recourse? Are they willing to keep working until you see the benefits? Or do they take the cheque and move on? This question separates operators from salespeople.
4. How long until we see results? If they say 12 months, walk away. If they say 6 weeks for a prototype, 16 weeks for production rollout, that's realistic. A long timeline isn't a sign of rigour; it's usually a sign of bloated process.
5. Do my reps need to change their workflow? If they say yes, that's a problem. Behaviour change is the graveyard of AI projects. Your reps won't do it. Good systems integrate into what they already do. No training required. No new tools to learn.
6. What's your smallest engagement? Do they have a workshop option? An audit? A low-risk entry point? Or do you have to commit six figures before you see anything? The former signals confidence. The latter signals desperation.
7. Who will actually deliver the work? Will you meet the person building it? Or will a sales director hand you off to someone junior after the contract is signed? You want the person in the room to be the person doing the work.
8. Can you estimate this in fixed price? Time-and-materials is fine for discovery. For production work, there should be a cap. A consultant who won't fix a price either doesn't know what they're doing or plans to overrun on scope.
9. What's your measurement framework? How will you know if this is working? Not vague KPIs like "efficiency gains." Specific, measurable outcomes. Hours reclaimed per rep per week. Revenue impact. Order processing time. Timeline to measurement.
10. Can you reference other clients like us? Not anonymized case study abstracts from their website. Real introductions to manufacturing companies similar to yours. People you can call. You're buying experience; check it.
The Cost Reality
What does AI consulting actually cost? Here's the breakdown for mid-market manufacturers.
Workshop / Audit: £1,500 to £5,000. This is the low-risk entry. A day or two on-site. You walk away with a clear problem statement, a proposed solution, and an understanding of what's involved. This should be low-cost because the objective is mutual discovery, not to generate billable hours.
Roadmap / Scoping: £5,000 to £15,000. A detailed specification of what you're building, how long it will take, what it will cost, and what ROI you'll see. This is the document you use to decide whether to proceed. A good roadmap is worth the cost because it de-risks the next phase.
Prototype / MVP: Custom fixed price, typically £15,000 to £50,000. This is the working system. A real, production-quality first version that solves your primary problem. Your reps can use it. You can measure the impact. Most consultants should be able to give you a fixed price here based on the roadmap.
Full Production & Rollout: Custom, typically £30,000 to £150,000 depending on complexity. Adding additional features, hardening the system, training your team, ongoing support. This should also be fixed price or time-capped.
Why cheap is expensive. If someone quotes you £5,000 for a prototype that should take eight weeks, they're either going to cut corners or the timeline will stretch to six months. And six months at £5,000 is a bad deal for you. Pay for quality. Get it in weeks, not months.
Why expensive isn't always good. A big consultancy quoting £250k for a 12-month engagement isn't premium; it's bloated. They're charging for their overhead and their sales team. You're paying for the name and the process, not for results. Mid-market manufacturers should expect to pay £30k to £80k for a full AI system that delivers measurable ROI. Anything above that, you're buying brand or bureaucracy.
Why Most AI Projects Fail
73% of AI projects never reach production. Your odds are worse than a coin flip. Why?
No commercial anchor. The consultant falls in love with the technology and loses sight of the business problem. They build a system that's technically elegant and commercially irrelevant. It's beautiful. It doesn't do anything your business needs.
Solution looking for a problem. This is the opposite. A consultant comes in wanting to build a specific type of AI system, and they squeeze your business into that shape. You didn't need a natural language model; they came with a natural language model, so that's what you get.
No adoption because behaviour change is hard. The system is deployed. Your reps are supposed to use it. They don't. Why? Because it means changing how they work. It requires new training. It requires new tools. It's a friction point in their day. They go back to the old way. The project dies quietly.
No measurement framework. You deploy something. You don't measure the impact. A month later, you can't tell if it's working. No data. No conversation. The project fades.
No commercial guarantee. The consultant takes the money. Deploys something. Walks away. Whether or not you see results, they got paid. They have no incentive to do the heavy lifting of measuring impact and iterating to improve it.
Timeline creep. What was supposed to be a 12-week project stretches to six months, then nine months. Scope creeps. Stakeholder sign-off slows. Momentum dies. By the time the system finally launches, your team is fatigued and has moved on to other priorities.
Working in isolation. The consultant builds something cool in a vacuum. They don't involve your sales team in the process. When it's time to deploy, your team sees it for the first time. It's not what they need. It requires them to learn new tools. It gets rejected.
The best way to avoid all of this is to work with someone who starts small, measures relentlessly, and guarantees outcomes. The workshop. The prototype. Measure the impact. Iterate based on what you learned. Then scale. That's how you avoid the 73%.
The Invisible UI Difference
Here's a concept most consultants don't understand, but it's critical: invisible UI.
Your sales team has a workflow. They open their CRM. They check emails. They pull specs from a PDF. They open a spreadsheet. They do some calculations. They write a quote. They send it. That's their day.
A consultant might come to you and say: "You need to change your workflow. You need to learn this new system. It's how you'll interact with the AI." Wrong. That's where adoption dies.
A smart consultant says: "Your workflow stays the same. You check your CRM, same as always. You read an email, same as always. Behind the scenes, the AI is listening. It's parsing your specs. It's pulling pricing data. It's doing calculations. By the time you write the quote, 80% of the thinking is done for you. You type a bit faster. You send it. Your day is marginally faster and easier." That's invisible UI. That's what adoption looks like.
This is why manufacturing experience matters. A generic AI consultant thinks you'll adopt new tools. A manufacturing consultant knows that you'll reject anything that changes your workflow. So they build around your workflow, not against it.
The Bottom Line
Choosing an AI consultant is about risk. You want someone who:
- Speaks your language (margin, throughput, revenue)
- Shows you working prototypes before you commit
- Has manufacturing experience, not just AI experience
- Offers a low-risk entry point
- Guarantees ROI or commits to keeping working until you see it
- Understands that invisible UI beats beautiful workflows
- Can produce results in weeks, not months
- References real clients you can call
Avoid anyone who:
- Talks about "transformation strategy" before understanding your sales process
- Offers a five-figure discovery phase
- Can't show a working prototype
- Is quoting a 12-month timeline for a quoting assistant
- Wants to replace your CRM
- Has no manufacturing experience
- Speaks in technical jargon instead of commercial language
The market is full of consultants. Most will take your money and deliver theory. A few understand how to build things that your team will actually use and that will actually move the needle on your commercial performance.
The questions above will separate them. Ask them. Trust your gut on the answers. Then make your decision.
What Next?
Ready to explore how this could work for your manufacturing business?
We run a 15-minute discovery call to understand your process, your bottleneck, and what good looks like. No pitch. No pressure. Just a conversation.
After that, we typically recommend a £3k to £5k workshop. Two days on-site. We audit your process, identify where the manual drag is, and show you what an automated solution could look like. You get a prototype. You get a clear roadmap. You get a number for what it's worth.
That's low-risk. You see something real. You decide if it's worth going further.
If you want to explore this, here's what to do:
Book a 15-minute discovery call
Secure your AI Deep Dive Workshop slot
Further reading:
How to Unlock Hidden Sales Capacity in UK Manufacturing
AI vs Hiring a Sales Rep: A Cost Comparison for UK Manufacturers
CPQ vs Custom AI Quoting Systems for Manufacturers
The Hidden Cost of Slow Quoting in Manufacturing
About the author: Ell Advisory helps UK mid-market manufacturers unlock revenue and sales capacity through AI. We've reclaimed 351,000 hours and delivered £10.1M in value for manufacturing clients. Our focus: commercial outcomes, not technology for its own sake.
