---
title: "AI and technology due diligence for UK manufacturers"
description: "AI and technology due diligence for UK manufacturers: ERP, MES and OT questions for deals, data room metrics, and AI that pays back for SME manufacturers."
url: https://anywayventures.com/sectors/manufacturing
updated: 2026-10-25
---

[← ALL SECTORS](https://anywayventures.com/sectors) SECTOR · MANUFACTURING AND ENGINEERING · UK

# AI and technology due diligence for UK manufacturers.

Manufacturing was the UK's third busiest sector for deals in 2025 and the busiest in Yorkshire and the Humber. Most of the businesses being bought still run on an ERP that predates the iPhone, a planner's spreadsheet and machines nobody has mapped. That is where the risk sits in a deal, and where AI pays back first for an operator.
[Book 30 minutes](https://calendly.com/brannan-anywayventures/30min) How we help

FOR · PE deal teams, operating partners, and CEOs, COOs and FDs of UK manufacturers and food producers
UPDATED · October 2026
BY[Brannan Coady](https://anywayventures.com/brannan-coady)
FREE TOOLS · [Technical due diligence checklist](https://anywayventures.com/resources/technical-due-diligence-checklist) , [AI due diligence question bank](https://anywayventures.com/resources/ai-due-diligence-questions) , [AI readiness scorecard](https://anywayventures.com/resources/ai-readiness-scorecard)

ON THIS PAGE
1. Where does AI change the economics of manufacturing? 2. Manufacturing in numbers: deals, adoption and Yorkshire 3. AI use cases that pay back for manufacturers 4. How should SME manufacturers in the UK start with AI? 5. Where AI threatens the manufacturing business model 6. Technology and AI due diligence questions for manufacturing deals 7. Metrics to test in the data room 8. Regulation and standards that shape AI and technology here 9. A 100-day plan after acquiring a manufacturer 10. Red flags and green flags 11. Case example: an owner-managed distributor made AI-native in a quarter 12. How we work with investors and operators in manufacturing → How we help

## KEY TAKEAWAYS

For UK manufacturers, AI pays back fastest in office and planning work such as quoting, order entry, supplier documents and quality paperwork, because shop-floor gains depend on machine data most SMEs do not collect.

Only 2% of UK manufacturers say AI is widely embedded across their operations, and 83% use it in HR, finance and admin against 6% in quality control (Make UK, 2026).

Manufacturing recorded 986 completed UK deals in 2025, the third most active sector, and 129 deals in Yorkshire and the Humber, where it was the region's most active sector (Experian MarketIQ, 2026).

In manufacturing deals, the three questions that reveal most of what the capex line is missing are the ERP support date, what runs outside the ERP, and whether OT is separated and backed up.

After acquiring a manufacturer, spend the first 30 days separating OT from the office network, taking offline backups and testing a restore, then prove one quoting and one overhead AI use case on real data.

AI and technology change the economics of UK manufacturing mostly away from the machine: in quoting, order entry, planning, purchasing, quality paperwork and the data that ties them to the ERP. On the shop floor the gains are real but slower, because they depend on machine data most SME manufacturers do not yet collect. Investors should test in diligence whether the ERP, the production data and the operational technology (OT) can support the value creation plan, and what it costs to fix them if not. Operators should start with the office and planning work that sits between an enquiry and a shipped order, because that is where AI pays back fastest and with the least risk to production.

I've been on both sides of technical due diligence, and at Netsells, the York product consultancy I co-founded and ran for thirteen years, we built digital products for manufacturers including Jaguar Land Rover and Warburtons. The pattern in manufacturing deals is consistent. The information memorandum talks about Industry 4.0. The site visit shows a Sage 200 install with three bolt-on Access databases, a production planner who rebuilds the schedule in Excel every morning, and CNC machines that have never been networked. None of that is unusual. Pricing it as if it weren't there is the mistake.

## Where does AI change the economics of manufacturing?

A UK manufacturer with 50 to 1,000 staff makes money on a few things: winning the right work at the right price, getting it through the factory on time with little scrap, and keeping the overhead around production lean. AI touches all three, but not evenly.

Winning work is the clearest case. A subcontract machinist or fabricator might receive dozens of RFQs a week as PDFs, STEP files, drawings and emails, and turn them into quotes with one or two estimators who carry decades of pricing judgement in their heads. Quote turnaround is slow, hit rates are rarely measured, and when the senior estimator retires the business loses its memory of what it charged and why. AI can read the enquiry, find the closest past jobs, pull material, machine and labour rates from the ERP and draft a quote for the estimator to correct. That raises capacity and shortens turnaround without taking the pricing decision away from the person accountable for margin.

Getting work through the factory is where the hype lives. Predictive maintenance, AI vision inspection and AI scheduling all work when machine data is collected and trusted, someone actually recorded the maintenance and quality history, and there is enough repeat volume for a model to learn from. Many owner-managed manufacturers have none of that. Fix data capture on the constraint machines first, then apply AI to a narrow problem with a known cost.

The overhead around production is the fastest win. Customer orders arrive as PDFs and get retyped into the ERP. Supplier certificates of conformity, mill certificates and delivery notes get checked by hand. Non-conformance reports, 8D reports and customer complaint responses are written from scratch. Month-end, credit control and purchase ledger work look the same as in any other business. Language models are good at this work today, and it carries little risk to production.

On the augmentation, transformation and disruption scale, most UK manufacturing sits at augmentation, with transformation in quoting, planning and quality for businesses that already have good data. Disruption is narrower than in software or services: the physical product still has to be made. The exposure sits in pricing (customers expecting a share of the efficiency), in engineering services sold by the hour, and in businesses whose real edge was a quoting or design skill AI now makes common.

## Manufacturing in numbers: deals, adoption and Yorkshire

Manufacturing is one of the most active sectors in UK M&A, even in a slower market. According to [Experian MarketIQ's full-year 2025 UK M&A review](https://www.experian.co.uk/content/dam/marketing/uki/uk/en/pdf/MA-Full-year-report-2025-V3.pdf) (Experian, 2026), manufacturing recorded 986 completed deals in 2025, down from 1,094 in 2024, behind only technology, media and telecoms (1,334) and professional services (1,007).

986UK manufacturing deals completed in 2025, the third most active sector ([Experian MarketIQ, 2026](https://www.experian.co.uk/content/dam/marketing/uki/uk/en/pdf/MA-Full-year-report-2025-V3.pdf) )
129manufacturing deals in Yorkshire and the Humber in 2025, the region's most active sector ([Experian MarketIQ, 2026](https://www.experian.co.uk/content/dam/marketing/uki/uk/en/pdf/MA-Full-year-report-2025-V3.pdf) )
2%of UK manufacturers say AI is widely embedded across their operations ([Make UK, 2026](https://www.makeuk.org/insights/reports/ai-skills-and-future-uk-manufacturing-sector) )
6%of manufacturers use AI in quality control, against 83% in HR, finance and admin ([Make UK, 2026](https://www.makeuk.org/insights/reports/ai-skills-and-future-uk-manufacturing-sector) )

### Why Yorkshire and the North matter for manufacturing deals

The same Experian review shows how concentrated manufacturing deal flow is in the North. In Yorkshire and the Humber, manufacturing was again the most active sector, with 129 transactions in 2025 against 130 in 2024, about a fifth of the region's 628 deals. In value terms it delivered close to £2bn, most of it from Qualcomm's £1.8bn acquisition of Leeds-based Alphawave IP Group. The North West added another 130 manufacturing deals (down from 154). Taken together, the two regions saw more manufacturing deals than the Midlands, the traditional heartland, which recorded 218.

Food and drink is a big part of that northern picture. The [Food and Drink Federation's 2026 regional report](https://www.fdf.org.uk/globalassets/resources/publications/reports/trade-reports/fdf-regional-report.pdf) (FDF, 2026) puts food and drink manufacturing in Yorkshire and the Humber at £4.1bn of GVA, 61,500 jobs and 985 businesses, which is 25.8% of all manufacturing employment in the region. Meat and bakery are the biggest employers. Across the UK the FDF calls food and drink the largest manufacturing sector.

The investors are local too. Leeds-based [Endless](https://endlessllp.com/) lists food and beverage, industrials and manufacturing among its sectors and invests from a £400m fund in businesses with £50m to £1bn of revenue, with a smaller £100m fund for £10m to £50m revenue businesses. Its recent activity includes buying Wolverhampton coatings specialist Anochrome, a sale [announced by adviser BDO in June 2026](https://www.bdo.co.uk/en-gb/deals/sale-of-the-anochrome-group-to-endless-llp) , and selling Hovis to Associated British Foods for £75m, a deal [completed in July 2026](https://www.ajbell.co.uk/news/articles/brief-ab-foods-completes-purchase-hovis-form-hovis-bakeries) (AJ Bell, 2026). BGF, which had backed 646 companies with over £4.9bn by the end of 2025 according to its [UK Enterprise Fund 3 annual report](https://www.coutts.com/content/dam/coutts/documents/wealth-management/ukef/ukef-3/annual-reports/BGF_UKEF3_annual_report.pdf) (BGF, 2026), is a regular minority investor in northern engineering businesses. [Inflexion](https://inflexion.com/our-firm/sectors/industrials) reports 17 industrials investments and €3.1bn invested in the sector, with industrial tech one of its stated focus areas.

### Why investors are leaning towards industrials

KPMG UK's mid-year review of private equity, published in July 2026, reported that UK PE deal volumes fell 10% year on year in the second quarter. Alex Hartley, KPMG UK's head of corporate finance, said investors would look for "assets that are less exposed to AI, with greater focus on blue collar services and industrials that offer secure, recurring revenues" ([KPMG UK, 2026](https://kpmg.com/uk/en/media/press-releases/2026/07/ai-uncertainty-and-conflict.html) ).

Reasonable, with a catch. "Less exposed to AI" means the product is hard for AI to replace, not that the business has nothing to gain from AI or is safe from a competitor who adopts it first. Good diligence prices both.

### Where AI adoption actually stands

Adoption in manufacturing is wide and shallow. Make UK's June 2026 report, [AI, Skills and the Future of the UK Manufacturing Sector](https://www.makeuk.org/insights/reports/ai-skills-and-future-uk-manufacturing-sector) (Make UK, 2026), found that only 2% of manufacturers say AI is widely embedded across their operations, fewer than 40% use it in some areas and nearly one in five have not adopted it at all. Use is concentrated in the back office: 83% use AI in HR, finance and admin, against 11% in production, 7% in supply chain and logistics and 6% in quality control. More than half cite skills shortages as the main barrier.

Size matters. In the earlier [Make UK and Autodesk report, Future Factories Powered by AI](https://makeuk.org/insights/reports/future-factories-powered-ai) (Make UK and Autodesk, 2024), large companies were more than twice as likely as SMEs to be applying AI, [71% against 28%](https://www.themanufacturer.com/articles/industry-set-to-increase-investment-in-ai-make-uk-autodesk-report/) , and only 16% of companies regarded themselves as knowledgeable about AI. Across the whole economy, the [Office for National Statistics](https://www.ons.gov.uk/businessindustryandtrade/business/businessservices/articles/artificialintelligenceinukbusinesses/2023to2026) (ONS, 2026) estimates that around 35% of UK businesses with 10 or more employees use AI, and reports that manufacturers most often adopt it through free-to-use software. Usually that means a free chatbot with no policy behind it.

For a buyer, that gap is the opportunity: most targets have done little with AI in operations. It also means any target claiming advanced AI deserves a hard look.

## AI use cases that pay back for manufacturers

These are the workflows where I would expect AI to return time or margin in a UK manufacturer with 50 to 1,000 staff, roughly in the order I would try them. The readiness signals column matters most. Without those signals, the use case is a project to fix data first.

Workflow
What AI does
Evidence or typical effect
Readiness signals

Quoting and estimating (RFQ to quote)
Reads the enquiry, drawings and specs, finds similar past jobs, pulls material, machine and labour rates from the ERP, drafts a costed quote for the estimator
Faster turnaround and more quotes per estimator; strongest on repeat and family parts. Measure hit rate and achieved margin on AI-assisted quotes separately
Two or more years of quotes and job costs in the ERP, linked to part numbers; estimators willing to log corrections

Sales order entry
Extracts lines, part numbers, dates and prices from customer PDFs and emails, checks them against price lists and contracts, posts to the ERP for approval
Removes most retyping; catches price and revision mismatches before they reach the shop floor
Order volume high enough to matter; a clean customer part cross-reference; ERP with an API or import route

Supplier documents and goods-in
Reads certificates of conformity, mill certificates and delivery notes, checks them against the PO, specification and material grade, flags exceptions
Fewer missed non-conformances at goods-in and faster release of material
Specifications held digitally; PO data accurate; a quality lead who owns the exception queue

Quality paperwork (NCRs, 8Ds, complaints)
Drafts non-conformance reports, 8D and CAPA responses and customer complaint replies from the notes and history, suggests root causes from past cases
Faster closure and more consistent responses for automotive and aerospace customers
NCR and complaint history in one system rather than scattered Word files; a quality manager who reviews every output

Engineering and shop-floor knowledge search
Answers questions across drawings, work instructions, setup sheets, FAIRs and past NCRs ("how did we hold tolerance on this part last time?")
Less time hunting for documents; some protection against key-person loss
Documents in one place with sensible revision control; permissions that match who should see what

Production planning support
Suggests schedule changes when orders, materials or machines change, explains the knock-on, drafts customer delivery updates
Planner time back and fewer late surprises. Rarely a replacement for the planner
Routings and cycle times in the ERP that match reality; a planner who will test it alongside the Excel schedule

Predictive maintenance on constraint assets
Learns normal vibration, temperature or current patterns and flags drift before failure
Pays where downtime on a bottleneck is expensive and failures give warning. Made Smarter's 2026 toolkit includes machine performance monitoring among its SME case studies
Downtime logged by cause; sensors or controller data on the bottleneck machine; a known cost per hour of downtime

AI vision quality inspection
Checks surface defects, labels, seals, fill levels or assembly presence on the line
Consistent inspection at line speed; payback depends on volume and the cost of escapes. Only 6% of manufacturers use AI in quality control (Make UK, 2026)
Stable lighting and presentation; hundreds of labelled good and bad examples; a clear reject process

Food: specification, allergen and label checks
Compares supplier specs with raw material records, checks label artwork against recipe and allergen matrix, drafts specification updates
Fewer label errors reaching print; faster response to retailer and customer technical questionnaires
Recipes and allergen matrix in a system, not a spreadsheet; specs held centrally; a technical manager who signs off

Food: traceability and recall drills
Pulls lot, batch and supplier records into a one-up, one-down trace in minutes, drafts the incident pack
Faster mock recalls and audit responses
Lot data captured at goods-in and dispatch; production records linked to batches

Two points from doing this work. Quoting is usually the best opening use case in subcontract and engineer-to-order businesses, because it ties straight to revenue. And a use case only pays back when the ERP is the source of truth. If the real costing lives in an estimator's spreadsheet, AI will faithfully reproduce whatever errors that spreadsheet contains. The [AI readiness scorecard](https://anywayventures.com/resources/ai-readiness-scorecard) has a data readiness test that will tell you which state you are in.

## How should SME manufacturers in the UK start with AI?

The [Made Smarter AI toolkit](https://www.madesmarter.uk/resources/ai-adoption-in-manufacturing/) , published in April 2026 with a foreword by Chris Dungey, the government's AI Champion for Advanced Manufacturing and CTO of the High Value Manufacturing Catapult, makes the same argument I would: start with tasks, problems and outcomes, begin with business and operational support systems, and only then move into production and engineering systems. Its scan, pilot and scale model is sound. Where I'd add detail is the pilot. Set the success measure before you start (hours per quote, errors per hundred orders), use a month of real examples, keep a person checking every output, and write down what the tool got wrong.

Made Smarter's adoption programme is the obvious first stop for SME manufacturers in England. [Made Smarter Yorkshire and the Humber](https://www.madesmarter.uk/adoption/in-my-region/yorkshire-the-humber/) offers a free digital readiness assessment and roadmap, technical support with up to 50% match funding, funded digital internships, and technology grants of up to £20,000 on a 50% match. Its fully funded Digital Leaders and Digital Champions courses, delivered by Sheffield Hallam University, started in York and Leeds in February 2026, and Made Smarter says 88 manufacturers in the region had been through them since 2021 ([Made Smarter, 2026](https://www.madesmarter.uk/resources/news-made-smarter-skills-programmes-to-help-yorkshire-manufacturers-build-ai-ready-leaders/) ). Check the current grant terms with the regional team, as they change each funding round.

For harder technical problems, the University of Sheffield's [Advanced Manufacturing Research Centre](https://www.amrc.co.uk/) on the Advanced Manufacturing Park in Rotherham is part of the High Value Manufacturing Catapult, which runs centres across the UK. AMRC is the right call for machining, vision, robotics and process research. It is less suited to the office-side AI work that pays back first, and that is where most SME manufacturers should begin. The [how to implement AI in a mid-market business](https://anywayventures.com/insights/how-to-implement-ai-mid-market-business) guide covers the order of work in more detail, and [why AI pilots fail](https://anywayventures.com/insights/why-ai-pilots-fail) covers the traps.

## Where AI threatens the manufacturing business model

Nobody prompts a gearbox housing into existence, so the direct threat is low. The indirect threats show up in margin before volume.

Customers will want the saving. Large OEMs and retailers already run aggressive cost-down programmes. Once they know AI cuts quoting, admin and inspection costs, they will expect a share of it in the next price negotiation. Suppliers who capture the gain first and reprice on lead time, quality and responsiveness will hold margin. Suppliers who wait will be told the number.

Quoting speed becomes the price of entry. In subcontract manufacturing, the supplier who responds first with a credible price often wins. Online manufacturing platforms already offer instant quotes for CNC machining, sheet metal and 3D printing on standard geometries. As AI-assisted quoting becomes common, a three-day turnaround on a simple RFQ starts to lose work it used to win.

Engineering hours get cheaper. Businesses that sell design, CAM programming or technical documentation by the hour are more exposed than those selling finished parts. AI-assisted CAD and CAM compress the hours a customer will pay for.

The moat might have been a person. In many owner-managed manufacturers the edge is a founder or senior estimator who knows what a job will cost and which customers to avoid. AI narrows that advantage over competitors who never had such a person. In a deal it cuts both ways: the key-person risk is real, and so is the chance to capture that knowledge before it walks out.

Who captures the efficiency gain? In my experience, it splits three ways: the customer through price, the workforce through fewer hours of overtime and agency labour rather than redundancies, and the owner through margin and capacity to take on more work at the same overhead. A value creation plan should state which share it assumes. Our [AI disruption risk](https://anywayventures.com/ai-disruption-risk) work scores this per business, and the [AI disruption risk scorecard](https://anywayventures.com/resources/ai-disruption-risk-scorecard) is the free version.

## Technology and AI due diligence questions for manufacturing deals

Operational and technology due diligence on a manufacturer has to cover three estates: the business systems (ERP, CRM, quality, finance), the shop-floor systems (MES, scheduling, machine data, PLCs and HMIs), and the people who keep them working. Most technology diligence reports cover the first well and the other two thinly. These are the questions I ask. The general [technology due diligence checklist](https://anywayventures.com/insights/technology-due-diligence-checklist) and the [AI due diligence questions](https://anywayventures.com/resources/ai-due-diligence-questions) cover the rest.

Question
Why it matters
Red flag

Which ERP runs the business, which version, and when does vendor support end?
Unsupported ERP means no security patches and an unplanned migration. Microsoft's extended support for [Dynamics NAV 2018](https://learn.microsoft.com/en-us/lifecycle/products/dynamics-nav-2018) ends in January 2028, and plenty of mid-market manufacturers still run older releases
ERP out of support, or a reseller who has gone quiet; migration not in the capex plan

How heavily customised is the ERP, and who wrote the customisations?
Custom code is what makes upgrades slow and expensive, and it often encodes the business's real costing and planning logic
Hundreds of custom objects written by a contractor who has left, no source code or documentation

What runs outside the ERP?
The spreadsheets and Access databases around the ERP are where planning, costing and quality really happen
A planner's schedule, the estimator's rate card and the quality log all in Excel on one person's desktop

How accurate are BOMs, routings and standard costs?
Margin by product, quoting accuracy and any AI use case depend on them. Bad routings mean the IM's product margins are guesses
Standard costs last updated years ago; variance written off at month-end without analysis

Is there an MES or shop-floor data collection, and what does it capture?
Real OEE, scrap and labour data only exist where something collects them. Without it, operational claims are estimates
OEE quoted in the IM but no system that records stops, speed or rejects

Which machines are networked, and how old are their controllers and the PCs that run them?
Old HMI and line PCs often run unsupported Windows. [Windows 10 reached end of support](https://learn.microsoft.com/en-us/lifecycle/products/windows-10-enterprise-and-education) in October 2025
Line PCs on Windows XP or 7, connected to the office network, with no compensating controls

Is the OT network separated from the business network, and who can reach it remotely?
Flat networks let ransomware cross from email to the production line. Machine builders' remote access is a common route in
Permanent remote access tools on PLC programming laptops; shared vendor passwords; no record of who connected when

Is there a current record of OT assets and connections?
The NCSC says you cannot secure OT without a definitive view of it, and its August 2026 alert put that first
Nobody can produce a list of PLCs, HMIs and their connections; "the integrator has it"

Are OT systems, PLC logic and ERP backed up offline, and has a restore been tested?
Recovery time after an attack decides whether the business loses days or weeks
Backups on the same network; never tested; PLC programs held only on the controllers

Who are the key people for the ERP, planning, maintenance and estimating?
In SME manufacturers these are often one person each, sometimes the same person
One ERP administrator with no deputy; the senior estimator retiring within two years with no succession

What AI or Industry 4.0 claims does the IM make, and what evidence sits behind them?
AI and "smart factory" language inflates value. Ask for the system, the data, the users and the measured result
A dashboard demo with no production users; "predictive maintenance" that is a calendar-based PM schedule

How are customer portals and EDI integrations run?
Automotive, aerospace and retail customers often require EDI, supplier portals and specific label formats. Failures mean penalties or lost lines
EDI run by one person through a legacy translator; scorecard penalties not disclosed

What do customer and certification audits say about systems and data?
IATF 16949, AS9100, ISO 9001 and BRCGS audits record system weaknesses an IM will not mention
Repeat findings on document control, traceability or calibration records

For food: how are traceability, allergen and specification data held?
Recall speed, retailer confidence and legal compliance depend on them
Allergen matrix in a spreadsheet maintained by one technical manager; trace exercises that take a day

How is staff use of AI tools governed?
Free AI tools with customer drawings or recipes pasted in can breach NDAs and expose IP
No policy; engineers uploading customer drawings to free tools; no list of approved tools

What is the digital plan in the investment case, and is it costed?
ERP replacement, MES and cyber work often appear as "upside" without cost or time
ERP migration assumed to take six months for a multi-site manufacturer with heavy customisation

If I could ask only three, they would be the ERP support date, what runs outside the ERP, and whether OT is separated and backed up. Those tell you most of what the capex line is missing. For the people side, [key-person risk in due diligence](https://anywayventures.com/insights/key-person-risk-due-diligence) covers how to price it.

### What the Norsk Hydro attack means for diligence

On 19 March 2019 Norsk Hydro, the Norwegian aluminium group, was hit by LockerGoga ransomware. The attack started in IT, but it reached production across the whole group. Hydro isolated its plants and switched to manual operations wherever it could. Its extrusion business, which makes the profiles and components most like a UK subcontract manufacturer's work, was the worst hit: [Hydro reported](https://www.hydro.com/en/global/media/news/2019/update-on-cyber-attack-march-22/) it was running at about 50% of normal capacity three days in, and its building systems units were still at around 20% a week after the attack. Encrypted PCs and servers had to be rebuilt from backups. [Hydro estimates the total cost at around NOK 800m](https://www.hydro.com/en/global/media/on-the-agenda/cyber-attack/) (Hydro, 2024), roughly £70m at 2019 exchange rates. A business of that size, with a large security team and cyber insurance, still lost weeks of output. An SME manufacturer with one IT manager and a flat network has far less room.

The supplier side runs both ways. In March 2022 Toyota [suspended all 28 lines at its 14 plants in Japan](https://global.toyota/en/newsroom/corporate/36964956.html) for a day because of a system failure at one plastic parts supplier, Kojima Industries, which had been hit by a virus. The [Associated Press reported](https://baynews9.com/fl/tampa/ap-top-news/2022/03/01/toyota-to-resume-japan-production-after-virus-hits-supplier) (AP, 2022) that the physical mechanics of production were not affected. The systems that carried parts orders were, and in a just-in-time supply chain that was enough. A supplier's incident can stop its customer, and a customer's stoppage can stop its suppliers, who still have staff and loans to pay.

For diligence, test the target's own controls first: OT separated from IT, offline backups of the ERP and PLC programs, a tested restore and a plan for running the plant manually. Then test its dependence on a large customer. Ask how long it could pay its staff if its biggest customer stopped calling off for five weeks, and whether its customers' supplier contracts require cyber controls it cannot yet evidence.

## Metrics to test in the data room

Manufacturing KPIs are easy to quote and hard to evidence. If the source is a spreadsheet someone fills in on Friday, treat the number as an estimate.

Metric
How AI moves it
How to evidence it

Quote turnaround time and hit rate
AI-assisted quoting shortens turnaround and lets estimators respond to more RFQs
CRM or ERP quote records with dates and outcomes, by customer and part family, for 24 months

Achieved margin against quoted margin
Better recall of past job costs narrows the gap
Job costing from the ERP against the original quote; variance by product family

On-time in-full (OTIF)
Planning support and supplier chasing reduce late orders
Dispatch records against promised dates from the ERP, not a manually kept scorecard

OEE on constraint machines
Condition monitoring and faster changeover planning reduce stops
MES or machine data. Check how availability, performance and quality are each measured

Unplanned downtime, MTBF and MTTR
Predictive maintenance targets these directly
Maintenance system or CMMS records by asset and cause. Paper logs are a finding

Scrap, rework and cost of poor quality
Vision inspection and faster root-cause analysis reduce escapes
NCR system data, customer returns and scorecards, credit notes for quality

Order entry error rate
AI extraction and validation removes most retyping errors
Credit notes and amended orders, customer complaints coded as admin error

Overhead hours per £1m of revenue
Office-side AI returns hours in sales admin, purchasing, quality and finance
Headcount by function and agency spend, against revenue, for three years

Inventory accuracy and stock days
Better purchasing signals and demand planning
Stock-take variances and cycle count results; slow-moving and obsolete stock provisions

Customer complaints and scorecard ratings
Faster, better 8D and complaint responses
Customer portal scorecards from automotive or aerospace customers, complaint logs

Food: giveaway, label errors and trace time
Line vision and label checks, faster traceability
Checkweigher data, label-related withdrawals, timed mock recall results, BRCGS audit grade

Two cautions. IM OEE figures are often calculated on planned production time rather than all available time, which flatters them, and labour productivity often excludes agency and overtime hours. Rebuild both from source data.

## Regulation and standards that shape AI and technology here

Manufacturing is lightly regulated on AI itself, but heavily regulated on the things AI touches: product safety, food safety, worker data and cyber resilience.

Operational technology security. The NCSC's guidance on [creating and maintaining a definitive view of your OT architecture](https://www.ncsc.gov.uk/collection/operational-technology/definitive-architecture-view) , published with CISA and other partners in 2025, sets five principles for knowing what OT you have and how it connects. In August 2026 the NCSC warned of [increased targeting of OT across multiple sectors, including in the UK](https://www.ncsc.gov.uk/news/disruptive-cyber-activity-highlights-risk-from-internet-exposed-systems-and-edge-devices) (NCSC, 2026), and told organisations not to assume their OT is unreachable from the internet without checking. The [ISA/IEC 62443 series](https://www.isa.org/standards-and-publications/isa-standards/isa-iec-62443-series-of-standards) is the international framework for industrial control system security, built on zones and conduits. Few SMEs will certify against it, but a buyer should expect its basic ideas: segmentation, controlled remote access and known assets.

Cyber legislation. The UK's Cyber Security and Resilience (Network and Information Systems) Bill had not received Royal Assent by late October 2026. It passed the Commons in June 2026 and finished Lords committee stage in September, with Lords report stage [scheduled for 26 October 2026](https://bills.parliament.uk/bills/4035) (UK Parliament, 2026), so most of its duties will arrive later through secondary legislation. It extends the NIS regime to data centres, managed service providers and designated critical suppliers. As [Brodies noted](https://brodies.com/insights/ip-technology-and-data/uk-government-publishes-cyber-resilience-bill) , it does not bring manufacturing or food and drink into scope directly, unlike the EU's NIS2. That leaves a manufacturer's own controls and its customers' contract requirements doing most of the work. Its managed service provider may now be regulated even if it is not.

Worker monitoring. Cameras for AI vision inspection, productivity tracking on the line and biometric clocking all process workers' personal data. The ICO's [guidance on monitoring workers](https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/employment/monitoring-workers/) expects a data protection impact assessment for intrusive monitoring and for biometrics, clear information to staff, and an alternative to biometric access for those who do not want it. Point inspection cameras at the product, not the operator, wherever you can.

Machinery sold into the EU. UK manufacturers who build machines for EU customers face the new EU Machinery Regulation 2023/1230 from 20 January 2027. Safety components with self-evolving behaviour using machine learning move onto the high-risk list and need third-party conformity assessment rather than self-certification ([Law Society Gazette, 2026](https://www.lawsociety.ie/gazette/in-depth/2026/may/the-eus-machinery-directive-is-being-updated/) ). If a target's roadmap includes AI-driven safety functions, diligence should check that this has been priced.

Food safety and traceability. Food businesses must be able to trace food one step back and one step forward under general food law, and the FSA's [general food law guidance](https://www.food.gov.uk/business-guidance/general-food-law) sets out traceability, withdrawal and recall duties. The FSA's [allergen labelling technical guidance](https://www.food.gov.uk/business-guidance/food-allergen-labelling-and-information-requirements-technical-guidance-summary) covers the 14 major allergens and precautionary labelling, and full ingredient labelling for [prepacked for direct sale food](https://www.food.gov.uk/business-guidance/introduction-to-allergen-labelling-changes-ppds) has applied since October 2021. The [BRCGS Food Safety standard](https://www.brcgs.com/certification-listing/food-safety/) , which BRCGS says supports over 24,000 businesses in more than 130 countries, is the certification most UK retailers expect. None of these rules forbid AI. All of them hold a named person accountable, so an AI tool that checks labels or specifications must leave an audit trail and a human decision.

Quality management standards. IATF 16949, AS9100 and ISO 9001 all require change control, so putting AI into inspection or document workflows is a process change that automotive and aerospace customers will expect to see managed.

## A 100-day plan after acquiring a manufacturer

This assumes a lower mid-market manufacturer with one to four sites, an ageing ERP and little shop-floor data. Adjust for scale.

Days 1 to 30: secure, map and stop the bleeding.

Separate OT from the office network where it is flat, remove permanent vendor remote access, and replace it with controlled, logged access.

Take offline backups of the ERP, PLC programs and HMI configurations, and test one restore.

Map every system, spreadsheet and Access database that the business runs on, with an owner for each.

Put a deputy and documentation behind the single ERP administrator, the planner and the senior estimator.

Publish a short AI acceptable use policy so staff stop pasting customer drawings into free tools. The [AI acceptable use policy template](https://anywayventures.com/resources/ai-acceptable-use-policy-template) is a starting point.

Days 31 to 60: pick two AI use cases and prove them.

Choose one revenue use case (usually quoting) and one overhead use case (usually order entry or supplier documents).

Run each on a month of real data, with a person checking every output and a success measure agreed in advance.

Start logging downtime by cause on the constraint machines if nobody does yet. That is the data any future maintenance AI will need.

Decide the ERP path: upgrade, replace or contain. Cost it properly, including data migration and the customisations.

Days 61 to 100: embed and plan the next year.

Put the proven use cases into daily work with named owners, and report hours saved monthly to the board.

Train the people who will use the tools, starting with estimators, planners and quality staff. [AI training](https://anywayventures.com/ai-training) for operators covers this.

Write the 12-month technology plan: ERP, OT security, MES or data collection, and the next three AI use cases, each with a cost and a measure.

Agree how the value creation plan will evidence AI gains at exit, so the next buyer's diligence finds numbers rather than claims.

For operators who are not part of a deal, the same plan works as a first 90 days. Skip the deal mechanics, keep the order. Secure the basics, prove two use cases, then plan.

## Red flags and green flags

Red flags I see in manufacturing targets:

An ERP out of vendor support, with heavy custom code and a reseller relationship that has lapsed.

The real production schedule kept in Excel by one planner, with the ERP updated afterwards.

Standard costs years out of date, and margin by product that nobody can rebuild from source data.

A flat network where an office PC and a PLC programming laptop can reach each other, and machine builders with always-on remote access.

PLC programs that exist only on the controllers.

OEE and productivity figures in the IM with no system that records them.

"AI-powered" or "smart factory" claims that turn out to be a dashboard or a supplier's demo.

A senior estimator or technical manager due to retire with no succession plan and no captured knowledge.

In food, allergen and specification data held in spreadsheets maintained by one person.

Green flags:

An ERP within support, lightly customised, with someone in-house who can explain every customisation.

Downtime, scrap and labour recorded in a system, by cause, for at least a year.

An OT asset list, segmented networks and a tested restore.

Quote history linked to job costs, so quoting accuracy can be measured.

Staff already using AI in a governed way, with a policy and approved tools.

In food, traceability exercises timed and recorded, and BRCGS findings closed promptly.

## Case example: an owner-managed distributor made AI-native in a quarter

The closest case I can share is a distributor rather than a manufacturer, but the workflows are the ones most manufacturers have in their office. In an owner-managed distributor we built a pricing tool and automated document handling over a quarter. Six months on, the tools were returning about 320 hours a month and running unattended, and the number of staff using AI every day had gone from 2 to 14.

The pricing work is close to quoting in a manufacturer: judgement held by a few people, pricing data split between the ERP and spreadsheets, and margin leaking through inconsistency. The document handling looks like order entry and supplier paperwork. What made it work was two workflows with a clear cost, measured from the start, and training for the people using the tools.

## How we work with investors and operators in manufacturing

For investors, we run technology and AI due diligence on manufacturing targets, covering the ERP, shop-floor systems, OT security, data and key people, and turn the findings into costs for the model and items for the 100-day plan. After completion we can act as an [AI operating partner](https://anywayventures.com/ai-operating-partner) across industrial portfolio companies, or help prepare a business for sale through [exit readiness](https://anywayventures.com/exit-readiness) work.

For operators, we start with an [AI audit](https://anywayventures.com/ai-audit) that ranks the workflows where AI will save the most hours in your business, then build and run the tools with your team through [embedded AI delivery](https://anywayventures.com/embedded-ai-delivery) . We are based in York and work across Yorkshire, the North and the rest of the UK, alongside Made Smarter and AMRC support rather than instead of it.

## How we help in manufacturing

[Technical due diligence ERP, MES, shop-floor OT, integrations and key people, tested before you sign, with the cost of fixing what we find. MORE →](https://anywayventures.com/technical-due-diligence) [AI due diligence Which AI and Industry 4.0 claims in the IM are real, and how many office and planning hours AI can actually return in this plant. MORE →](https://anywayventures.com/ai-due-diligence) [Exit readiness One ERP, a mapped OT estate, clean BOM and costing data and evidenced KPIs before the next buyer's diligence team arrives. MORE →](https://anywayventures.com/exit-readiness) [AI operating partner For funds with several industrial assets: one person who runs the AI plan across the portfolio and reports to the deal team. MORE →](https://anywayventures.com/ai-operating-partner) [AI audit For operators: a ranked list of the workflows where AI saves hours, from quoting to order entry to quality paperwork, costed on your data. MORE →](https://anywayventures.com/ai-audit) [Embedded AI delivery We build and run the tools with your team: quoting and estimating, order intake, supplier documents, maintenance and quality records. MORE →](https://anywayventures.com/embedded-ai-delivery)

[FREE RESOURCE · PDF + XLSX Technical due diligence checklist DOWNLOAD FREE →](https://anywayventures.com/resources/technical-due-diligence-checklist) [FREE RESOURCE · PDF + DOCX + XLSX AI due diligence question bank DOWNLOAD FREE →](https://anywayventures.com/resources/ai-due-diligence-questions) [FREE RESOURCE · PDF + XLSX AI readiness scorecard DOWNLOAD FREE →](https://anywayventures.com/resources/ai-readiness-scorecard)

## Questions

### What should technology due diligence on a manufacturer cover?

Start with the ERP: version, vendor support dates, customisations, and how much of the business runs in spreadsheets around it. Then test the shop floor: MES or job tracking, machine connectivity, and how operational technology is separated from the office network. Check BOM, routing and costing data quality, key people (often one ERP administrator and one planner), cyber controls and backups, and every AI or Industry 4.0 claim in the IM. For food producers add traceability, allergen and specification data. Each finding should come with a cost and a timeline.

### How can SME manufacturers in the UK start using AI?

Start with a task, not a technology. Pick one workflow with a clear cost: quoting, order entry from customer PDFs, supplier certificates or quality paperwork. Test AI on a month of real examples, with a person checking every output, and measure the hours saved. Made Smarter's 2026 AI toolkit sets out the same scan, pilot and scale order, and in England Made Smarter offers SME manufacturers free advice, digital roadmaps and part-funded technology projects. Move to the shop floor once the data foundations are sound.

### Can AI produce quotes and estimates for a manufacturer?

Yes, for much of the work. AI can read an RFQ, drawings and specifications, find similar past jobs, pull material and machine rates from the ERP and draft a costed quote for an estimator to check. It is strongest on repeat and family parts, and weakest on genuinely novel work, where it should flag uncertainty rather than guess. The usual gains are faster turnaround and more quotes per estimator. Keep a person approving every price, and track hit rate and margin on AI-assisted quotes separately.

### Is predictive maintenance worth it for a small manufacturer?

Sometimes. It pays where unplanned downtime on a few constraint machines is expensive and the failure modes give warning, such as vibration, temperature or current draw on spindles, pumps, compressors and gearboxes. It rarely pays across a whole site of older machines with no sensors and no maintenance history. Start by logging downtime properly and adding condition monitoring to the bottleneck asset. If you cannot say what an hour of downtime on that machine costs, you are not ready for predictive models.

### What are the biggest technology risks when buying a UK manufacturing business?

The common ones are an ERP out of vendor support with heavy customisation, one person who understands it, production planning that really happens in Excel, BOM and costing data nobody trusts, and an OT network that is flat with the office network and reachable remotely by machine builders. Norsk Hydro's 2019 ransomware attack, which Hydro estimates cost it around NOK 800m, showed how far an office IT incident can reach into production. Each risk needs a cost in the price or the 100-day plan.

### How is AI used in food and drink manufacturing?

The most useful early work is in specification, allergen and traceability data. AI can check supplier specifications against your raw material records, compare label artwork with the recipe and allergen matrix, draft BRCGS document updates and pull lot records together quickly when a customer or the FSA asks. On the line, vision systems can check labels, seals and fill levels. Allergen and traceability decisions must stay with a named, trained person, and every AI check should leave an audit trail.

Written by Brannan Coady
Founder of Anyway Ventures. Co-founder and CEO of Netsells for thirteen years, then Chief Product Officer, COO and CEO of YourParkingSpace and a member of the Flowbird Group executive committee through its sale to EasyPark (now Arrive). Investor and board member; now leads technology and AI due diligence and delivery for PE-backed and mid-market businesses. [More about Brannan →](https://anywayventures.com/brannan-coady)

## Related

[ARTICLE · 10 MIN Technology due diligence checklist for private equity deals](https://anywayventures.com/insights/technology-due-diligence-checklist) [ARTICLE · 9 MIN Key person risk in due diligence: would the product survive the CTO?](https://anywayventures.com/insights/key-person-risk-due-diligence) [ARTICLE · 9 MIN Why AI pilots fail, and what the survivors do differently](https://anywayventures.com/insights/why-ai-pilots-fail)
OTHER SECTORS[Parking](https://anywayventures.com/sectors/parking) [Distribution and logistics](https://anywayventures.com/sectors/distribution-and-logistics) [Professional services](https://anywayventures.com/sectors/professional-services) [Field services, FM and TIC](https://anywayventures.com/sectors/field-services-and-tic) [Software and SaaS](https://anywayventures.com/sectors/software-and-saas)
