Skip to main content

Where Should a Small Manufacturer Start With AI? (Not on the Shop Floor)

Operations professional reviewing an automated workflow at a computer

Most small and mid-size manufacturers should start with AI between sales, operations and administration, not on the shop floor. The fastest wins are quote preparation, SOPs and work instructions, reporting out of the ERP, and routine supplier communication. Robotics, predictive maintenance and computer vision can pay off later, once the data and processes underneath them are ready.

Manufacturing gets some of the most aggressive AI pitches of any industry. Robotics vendors, industrial IoT platforms, quality control software, predictive maintenance systems: there's no shortage of people selling transformation to operations managers who are already stretched thin.

Most of it is aimed at the wrong part of a small or mid-size operation. The rest, while potentially useful someday, needs a level of capital, data infrastructure and integration work that puts it out of reach for manufacturers under $100M in revenue who are trying to run lean. The near-term opportunity lives in the connective tissue of the business, where manual work compounds quietly and nobody is measuring the cost.


What should a small manufacturer not start with?

Shop floor automation and robotics. This isn't a software project. Automating physical production takes capital expenditure, facility changes, safety validation and ongoing maintenance that's out of scope for most smaller operations right now. The vendors aren't wrong that it can work. They're wrong that it's the right starting point.

Predictive maintenance platforms. Predictive maintenance is genuinely valuable when you have consistent sensor data, a mature data infrastructure, and engineering staff who can act on what the system surfaces. Most smaller manufacturers don't have that yet. Buying the platform first means paying for a system that underperforms and eventually gets blamed for not delivering.

Computer vision quality control. Same story. Vision QC works with clean, consistent training data, a controlled production environment, and integration with your existing quality workflow. Without those, you're not deploying AI. You're funding a pilot that will need significant engineering support to get off the ground.

TechnologyWhat it needs firstFit for a small plant today
Robotics and shop floor automationCapital, facility changes, safety validationUsually later
Predictive maintenanceConsistent sensor data and staff to act on itLater, after the data exists
Computer vision QCClean training data, a controlled line, quality integrationLater, as a scoped pilot
Quoting and order supportAccess to customer specs and past quotesNow
SOPs and work instructionsPeople who know the processNow
ERP reportingA connection to the ERPNow, if the data can be reached

These technologies aren't bad. But starting with them before the operational foundation exists is how AI investments fail quietly: not with a visible disaster, but with a slow fade as the system underperforms and adoption stalls. We've written more about that pattern in why small business AI projects fail before they start.


Where does AI help a small or mid-size manufacturer?

Order and quoting support. The back-and-forth between sales and operations on quotes is where many small manufacturers lose time they can't bill. Pulling spec requirements from customer documents, summarizing lead requirements for production planning, assembling quote packages: this is structured, repetitive work AI handles well. A sales coordinator who spends hours on a complex quote package can cut that significantly with the right tools in place.

Production documentation and SOPs. Writing and updating work instructions, quality procedures and standard operating documents is always behind. AI drafts and revises faster than anyone does by hand. The expertise still has to come from your people, because the model doesn't know your process, but a long documentation project becomes a much shorter review and edit. For manufacturers building ISO-ready documentation or onboarding new operators, this is one of the highest-return uses available.

ERP data and reporting. Most manufacturers have a lot of operational data in their ERP and not enough analyst time to get regular, useful insight out of it. AI-assisted reporting (production performance, patterns in material usage, summary reports for plant management) can fill that gap without a dedicated hire. The key is connecting AI to where the data actually lives.

Supplier and supply chain communication. Purchase orders, change requests, supplier follow-ups, delivery confirmations: manufacturers send a high volume of structured supplier communication. AI drafting cuts that time the same way it does anywhere there's repetitive outbound correspondence. The operations team still reviews before anything goes out, but the draft is already there.


Why does the ERP decide what AI can do?

This is the part most AI vendors skip, and it decides whether an AI project works in a manufacturing environment.

Your ERP holds production schedules, inventory levels, customer orders and supplier lead times. An AI tool that isn't connected to that data only knows what someone typed into a prompt or uploaded. That's useful for some things, but it puts a ceiling on what AI can do for your operation.

Any serious AI strategy in manufacturing has to answer how the tools will work with the ERP: direct integration, scheduled exports, or middleware. That's doable, and easier on some platforms than others, but it takes someone who has done this kind of integration and knows the practical options, not just the vendor's sales diagram. Our guide to system integration for small business compares those options.

The ERP question gets even more pressing during a change of systems. This year we helped a manufacturer changing ERPs. They needed to keep years of history and help their people work well in the new system. We built an archive that mirrors the system they left, plus tools that make the new one easier to work in. If you're facing the same move, see what happens to your old data in an ERP migration and our software and ERP migration work.


What about the knowledge that isn't written down?

Manufacturing has an advantage here: SOPs often exist. Quality systems, work instructions and safety procedures get written down because a certification, an audit or a customer required it at some point.

But documented and current aren't the same thing, and formal SOPs never capture everything. In most small manufacturers, a lot of knowledge lives in people's heads: the floor supervisor who knows which machine runs hot on Tuesdays, the purchasing coordinator who knows which supplier quotes aggressively but ships late. AI can't use what isn't written down.

That's an opportunity, not just a problem. AI is excellent at helping experienced people document what they know. An operator who would never write a work instruction will often talk through a process while someone drafts it with AI in real time. The documentation can happen alongside AI adoption rather than before it, as long as someone is actively driving it. The manufacturers that pull ahead aren't necessarily the ones with the best documentation to start. They're the ones with someone responsible for capturing knowledge as the work gets done.


How should a small manufacturer start with AI?

A small manufacturer should start with AI at the points where sales, operations, administration, suppliers and customers hand work to each other, because that's where manual, unbillable work piles up and process gaps are easiest to see. Typical first projects are quote package assembly, SOP and work instruction drafting, ERP reporting and routine supplier communication. Before adding any tool, document how those processes actually run, including the exceptions and judgment calls, not just the official procedure. Then build AI help around what's documented. Raise the ERP question early: it doesn't have to be solved first, but it decides which uses are within reach. Finally, give one person the time, authority and accountability for whether the AI work delivers. Shop floor robotics, predictive maintenance and computer vision can come later, once the data, processes and ownership can support them. That order turns small wins into a foundation instead of a pile of pilots.

Having that owner matters more than any tool choice. If you don't have someone internally with the time, that's what our AI strategy and fractional AI leadership work provides.


The Cedar Rapids–Iowa City Corridor has a substantial manufacturing base, from precision manufacturing to food production to industrial equipment, and regional context shapes supplier relationships, workforce dynamics and integration choices. If you're sorting through pitches for a plant in Iowa, start with our Advise, Educate & Lead page.

Frequently Asked Questions

How can a small manufacturer start using AI?

Start where sales, operations and administration meet. Good first uses are assembling quote packages from customer specs, drafting and updating SOPs and work instructions, pulling reports out of your ERP, and drafting routine supplier communication. These are structured, repetitive tasks with clear owners. Leave robotics, predictive maintenance and computer vision until your data and processes can support them.

Is predictive maintenance worth it for a small manufacturer?

It can be, once the foundation exists. Predictive maintenance needs consistent sensor data, a mature data infrastructure and engineering staff who can act on what the system surfaces. Most smaller plants don't have all three yet. Buying the platform first usually means paying for a system that underperforms and gets blamed for it. Build the data habits first, then revisit.

Does AI need to connect to our ERP?

For the most valuable uses, yes. Your ERP holds production schedules, inventory, orders and supplier lead times, and an AI tool that can't reach that data only knows what someone typed or uploaded. Drafting, communication help and document search work without it. Reporting and anything that feeds operational decisions needs a connection, through direct integration, scheduled exports or middleware.

Can AI help write SOPs and work instructions?

Yes, and it is one of the best early uses in manufacturing. The expertise still has to come from your people, because the model doesn't know your process. But an operator who would never sit down to write a work instruction will often talk through a process while someone drafts it with AI. That turns a long writing task into a shorter review.

Who should own AI in a small manufacturing company?

One named person with time, authority and accountability for whether AI projects deliver, not a committee and not whoever is most enthusiastic. They should understand how the plant actually runs, keep the ERP question on the table, and make sure each process is documented before it is automated. Enthusiasm alone doesn't sustain implementation across a production environment.

Drew Bloom, founder of Mosaic Solutions

Drew Bloom

Founder, Mosaic Solutions — Business Systems, Software & Automation

Drew is the founder of Mosaic Solutions. He spent several years as an automation consultant and Innovation Lead for Affinity, where he worked with law firms and enterprise clients on AI strategy, change management, and custom software. He works with small and mid-sized businesses in Iowa and beyond to fix the systems that slow them down — with better processes, software, and AI that actually works.

More about Mosaic LinkedIn Work with Drew

Bring Us the Decision in Front of You.

AI strategy, policy, vendor choices and fractional technology leadership, scoped first and billed for the time the work takes.