There is a particular kind of frustration that settles over a leadership team somewhere around the second or third conversation about AI. Everyone agrees that it is relevant. Everyone has seen the case studies. Someone has been to a conference. And yet, six months on, nothing has moved.
This is not a failure of ambition. It is a failure of starting point. AI is broad, the options are overwhelming, and the stakes feel high. The result is a kind of productive paralysis: lots of discussion, very little action. If this sounds familiar, you are in good company. It is one of the most common situations we encounter when working with mid-size manufacturers.
The good news is that getting unstuck does not require a grand strategy or a significant upfront investment. It requires answering two honest questions, and then being willing to start small.
The two questions that cut through the noise
Before any conversation about technology, vendors or budgets, every manufacturer needs to answer the same two questions clearly.
The first is: what data do you actually have? Not what data you think you have, or what data your ERP theoretically contains, but what is genuinely available, reasonably clean, and accessible without a six-month data engineering project.
The second question is: where is the pain that is costing you the most, right now? Not the most exciting AI use case you have read about, but the actual bottleneck, inefficiency or recurring problem your team mentions in every operations review.
The answer to both questions, taken together, almost always points clearly toward a starting place. If you have reasonable operational data and a persistent process problem, you probably start with automation. If your leadership team cannot agree on what matters most, you probably start with strategy. If your data is a mess, you may need to start with data readiness before anything else.
Why administrative automation is often the right first step
For manufacturers who are genuinely new to AI, we have a strong bias toward starting with business process and administrative automation. The reason is straightforward. It does not require shopfloor access. It does not put production at risk. And it typically delivers visible, measurable results within a few months rather than a few years.
The administrative layer of a manufacturing business - purchase orders, invoice processing, quality reporting, shift handover documentation, management summaries - is almost always heavily manual, surprisingly time-consuming, and completely overlooked.
When you automate even one or two of these processes, you free up meaningful hours, reduce errors, and build confidence internally that AI can actually deliver something tangible.
When to start with strategy instead
There are situations where starting with a hands-on automation project is the wrong move, usually when the leadership team does not have a shared view of where AI should focus.
If your operations director sees one opportunity, your finance director sees another, and your CEO is being pulled in a third direction by a vendor, the first thing you need is alignment, not a prototype. In this case, an AI strategy engagement is the right starting point: a structured process that maps your operation, surfaces the highest-value opportunities, and creates a shared roadmap.
Strategy-first is also the right approach if you are considering a significant investment and want an independent view before committing. Getting the prioritisation right at this level has a disproportionate impact on outcomes.
What to avoid
The most common mistake we see is starting with the technology rather than the problem. A vendor demonstrates a compelling platform, a board member reads about a competitor's AI initiative, and suddenly the question becomes "how do we use this tool?" rather than "what problem are we trying to solve?"
This almost always leads to expensive, underused deployments. The technology was chosen before the use case was clear, and the use case was chosen before data readiness was understood. Start with the problem. Find the data that speaks to it. Then choose the technology.
A simple framework for choosing your starting point
Ask yourself these three questions:
- Does my leadership team agree on the top one or two AI priorities? If no, start with AI Strategy & Advisory.
- Is our biggest pain in a back-office or administrative process rather than on the shopfloor? If yes, start with Business Process & Administrative Automation.
- Do we have reasonable data available and a specific operational problem? If yes, consider Process Automation, Supply Chain Intelligence or Energy AI.
None of these starting points is permanent. The goal of the first engagement is not to solve every AI challenge in your business. It is to build a small, credible foundation of results that makes the next step easier to justify and easier to execute.