Beyond The Buzz: Practical AI & Copilot Strategies for Manufacturers
Key Takeaways
✅ AI initiatives often fail when the problem
isn’t clearly defined upfront
✅ Success with Copilot depends on a process-first
approach, not a product-first mindset
✅ High-impact use cases are practical,
repeatable workflows with measurable gains
✅ Governance and data security are foundational
advantages of Copilot in Microsoft environments
Related Resources
+ Full Transcript
Practical AI and Microsoft Copilot Strategies for Manufacturers
Moderator: Everybody, welcome. Thanks for joining us today. We’re going to tackle practical AI and Microsoft Copilot strategies for manufacturers.
I’m Subzy, Marketing Director here at 2W Tech, and I’ll be moderating today’s conversation.
Today’s presenters are Mark Jamieson, President and CEO, and Megan Dorne, one of our Solutions Engineers.
If you’re not familiar with 2W Tech, we’re a technology solutions and managed services provider specializing in solutions for the manufacturing and distribution industries. Our roots are in Epicor ERP, and we’ve since expanded across Microsoft, managed IT, cybersecurity, cloud, data, and AI solutions.
Now I’m going to turn things over to Mark to get us rolling.
Why Do AI Projects Fail?
Mark: There’s been a lot of buzz around AI. And I don’t mean just a little buzz. There’s a lot, a lot of buzz.
We’re having conversations with clients about questions like: How can I actually use AI? What’s the practical application? Where does it provide real value?
Megan has been working with our clients on AI initiatives, and one thing we’re finding is that a lot of organizations are jumping into AI, only to see those projects stall.
Megan, why do you think that’s happening?
Megan: In many cases, it’s not because the technology is bad. It’s because the organization wasn’t clear enough about the process or problem AI was supposed to solve.
Businesses invest time, budget, and energy into an AI proof of concept, but sometimes they eventually walk away because they aren’t realizing the value they expected.
A common issue is that organizations focus on the AI product before they’ve spent enough time understanding the business process.
When the problem isn’t clear, the context is scattered, and the workflow hasn’t been properly defined, simply adding an AI tool doesn’t necessarily improve it.
Mark: And that becomes especially important when you look at where business information actually lives.
Why Is Business Data Important for AI?
Megan: A lot of the information manufacturers and other businesses rely on is unstructured.
You have information sitting in emails, Microsoft Teams conversations, physical documents, PDFs, meeting notes, and other sources.
Mark: Exactly. Think about the difference between the “what” and the “why.”
Structured systems such as your ERP, CRM, and manufacturing applications often tell you what happened.
But the context explaining why something happened may be sitting in an email, Teams meeting, memo, document, note, or even tribal knowledge that isn’t formally captured anywhere.
A useful AI strategy needs to look across the enterprise, including structured and unstructured business data.
That’s why we’re finding that AI success isn’t primarily about the product you buy. It’s about the process you’re trying to improve and the information needed to improve it.
Organizations sometimes buy an AI tool and expect it to be magic.
And as we know, there ain’t no such thing as magic systems.
How Should Manufacturers Start Building an AI Strategy?
Megan: Instead of starting with the product, start with the friction.
Where are the bottlenecks? What are you trying to measure? Where are employees losing time? What efficiency or effectiveness problem are you trying to solve?
Once you’ve defined that, you can evaluate the available AI tools and determine which solution provides the right value.
Then you can measure the results, refine the process, and scale what works.
With Microsoft Copilot, for example, the value doesn’t come simply from owning the license. The value comes from how Copilot is applied to actual workflows and business problems.
Mark: When we engage in an AI initiative, we start by planning.
What are you trying to accomplish? What’s the strategy? Where does the data live? Do you have the information you need? Who has access to it?
Microsoft Copilot is valuable because it operates within the Microsoft environment and works with existing Microsoft 365 permissions and governance.
What Makes a Workflow Ready for an AI Pilot?
Megan: Before choosing a tool, we like to pressure-test the workflow.
There are several questions that help determine whether a workflow is ready for an AI pilot:
1. Is the process repeatable?
Does this task happen frequently enough that improving it will create meaningful value?
2. Who owns the process?
Is there a clear person or team responsible for the workflow?
3. What does the process cost today?
How much time, labor, rework, or other resources does the current process require?
4. What does success look like?
What outcome are you trying to improve, and how will you know whether the AI pilot worked?
5. What problem are we actually trying to solve?
Are we clear on the friction point, business need, or inefficiency we’re addressing?
If you can’t confidently answer those questions, the workflow may need more refinement before it’s ready for an AI pilot, no matter how impressive the AI demonstration looks.
Mark: Everybody wants to get to the fun stuff. Nobody wants to do the hard stuff.
Someone once told me preparing for AI is like cleaning out your garage. It’s messy and dirty, but once you’re finished, it’s worth it.
Practical AI Use Case #1: Automating Manufacturing Timesheets
Megan: One manufacturing example involved very manual employee timesheets.
Some information was handwritten, some came from punch clocks, and someone was spending two to three hours manually reading and rekeying that information.
The organization began scanning the different timesheets into a central location. AI could then extract the information from those scanned sheets and structure the data in Excel.
Instead of manually entering every record, an employee could review the output and focus primarily on exceptions or information that couldn’t be read confidently.
That reduced the process from approximately two or three hours of manual entry to roughly 20 to 30 minutes of review and validation.
Mark: I’ve used a similar workflow with handwritten notes. I can take handwritten notes, bring them into Copilot, and turn them into structured text that I can then review and refine.
That’s a good example of AI helping convert unstructured information into something usable without requiring someone to manually retype everything.
Practical AI Use Case #2: Improving Manufacturing Shift Handoffs
Megan: Another opportunity involves manufacturing shift handoffs.
The problem usually isn’t that employees aren’t trying to communicate. It’s that every handoff can look different.
Important information may live in notes, chats, meetings, system records, or someone’s memory.
AI can help bring that information together and organize the handoff around consistent categories such as decisions, blockers, owners, risks, due dates, and next steps.
That information can then be summarized into a structured format, making it easier for the next shift to quickly understand what happened and what requires attention.
We’ve seen workflows like this reduce a lengthy shift handoff to a much shorter review process.
Mark: A variation I use personally is something I call “Missed Promises.”
I have a process that reviews my communications and identifies things I’ve promised people I’d get done but haven’t completed yet.
It flags those outstanding commitments for me so they don’t disappear into email or Teams.
That same idea applies to personal productivity, department communications, manufacturing handoffs, and team workflows.
Practical AI Use Case #3: Preparing for Customer Meetings
Megan: AI can also significantly reduce the time required to prepare for customer conversations.
Whether you’re responding to a service ticket, preparing for a sales call, or meeting with an existing customer, you can ask Copilot to help summarize the relevant history.
That can include prior meetings, emails, Teams conversations, open items, tickets, and other information you have permission to access.
Instead of manually searching through multiple places, you can create a concise briefing and link back to the relevant source material.
Mark: This works internally too.
Before a one-on-one with your manager, you can summarize what you accomplished during the week, identify blockers and outstanding commitments, and even surface accomplishments you may want to discuss.
The same concept works for project reviews, quarterly customer meetings, department meetings, and recurring business reviews.
Practical AI Use Case #4: SOP and Engineering Change Reviews
Megan: Another popular manufacturing use case involves standard operating procedures, engineering documentation, and change reviews.
Quality and engineering teams can spend significant time reviewing documents, comparing versions, validating changes, and making sure employees are using the correct approved documentation.
AI can help compare approved documents against proposed revisions, identify changes, summarize differences, and prepare recommended edits for human review.
The source matters. You want your AI workflow looking at the approved production documents, not an uncontrolled draft sitting somewhere else.
That’s another reason good content organization, permissions, and governance matter before you introduce AI.
Mark: You can also take a meeting transcript where a team discussed how a business process should work and use AI to generate a first draft of an SOP.
Then a person reviews it, corrects it, and approves it.
That’s important: you generally don’t want to publish raw AI output without human review.
AI can accelerate the creation and revision process, but people remain responsible for validating the final result.
How Can AI Help with Compliance and Document Governance?
Megan: This also has implications for compliance and auditability.
Organizations dealing with requirements such as CMMC can use structured document repositories, naming conventions, templates, permissions, and approval processes to help AI work with the right information.
Prompts can also specify required formats, guidelines, and source locations so employees receive information that aligns more closely with established organizational processes.
AI can also help check references and links within documentation, identify outdated information, and make the document review process more efficient.
What If Microsoft Copilot Isn’t the Entire Solution?
Megan: Copilot isn’t the answer to everything.
Sometimes the workflow requires automation or information from systems outside Microsoft 365.
Maybe requests are getting lost in email. Maybe you need information from Jira, an ERP, CRM, finance application, quality system, or another business platform.
In those cases, the goal is to look at the entire workflow: intake, routing, assignment, notification, tracking, approvals, and reporting.
Copilot might be part of that solution, but additional automation, agents, integrations, or business-specific AI may also be needed.
Mark: We’ve been able to make interesting progress here with Resolve IQ and integrations into other business systems.
For example, you might ask AI to prepare you for a meeting with a customer and then also include relevant structured data, such as the customer’s accounts receivable status.
You don’t necessarily need a massive integration project across every application.
Start with the workflow, identify the data it requires, and connect what you need to solve that specific business problem.
How Do You Measure AI ROI?
Megan: When evaluating AI, we want to be careful not to treat projected savings as guarantees.
Instead, start with real inputs from your organization: labor costs, processing frequency, current task duration, error rates, rework, and other measurable factors.
For example, if a document-processing workflow saves two hours every pay period, you can use the organization’s actual labor costs and number of annual pay periods to calculate the potential savings.
The same approach works for shift handoffs, customer meeting preparation, SOP reviews, quality processes, approvals, and other repeatable workflows.
Mark: That’s the key to AI readiness. You have to define what a win looks like.
Don’t simply say, “We’re going to implement AI.”
Establish a baseline, define the desired result, measure what happened, and determine whether the workflow actually improved.
What Does AI Readiness Look Like?
Megan: AI readiness doesn’t mean waiting until every possible risk disappears.
It means understanding who owns the process, where the relevant information lives, how access is controlled, who should participate in the pilot, and how success will be measured.
We look at strategy, data, security, processes, applications, identity, licensing, training, and measurement.
For a Microsoft Copilot pilot, we’ll typically identify a specific group of users, train them, establish testing procedures, and help them apply Copilot to everyday work habits.
Then we measure the results and determine whether the use case should be refined, expanded, or scaled.
Mark: One process, not one product.
It’s about the process. It’s not about the product.
The methodology is pretty simple: baseline it, prove it, scale it.
Microsoft Copilot and AI for Manufacturing: Webinar Q&A
What Makes a Process a Good AI Pilot Candidate?
Megan: Look for something repeatable, measurable, and owned by a specific person or team.
You need to understand what the process costs today and what a better outcome would look like.
Manual processes that happen frequently, require repetitive data entry, involve multiple people, or create consistent bottlenecks can be good candidates.
The timesheet example is a good illustration because the task happened frequently, consumed a measurable amount of time, and had a clearly defined improved state.
Mark: I’d add one more thing: What’s most painful?
Find the process that’s creating real friction.
My first significant Copilot initiative came from email because my inbox was crushing me. I was receiving hundreds of emails a day, but only a fraction actually required action.
I gradually created prompts and workflows to understand the incoming email, organize it, identify what needed action, and summarize lower-priority information.
That’s created significant personal productivity gains because it solved an actual problem I had, rather than starting with a technology and searching for something to do with it.
How Do You Protect Sensitive Data and Prevent Oversharing with Microsoft Copilot?
Megan: Microsoft Copilot works with the permissions users already have. It doesn’t automatically give users permission to information they couldn’t previously access.
But that also means organizations need to make sure their existing permissions are configured correctly.
Review SharePoint permissions, sensitive content locations, external sharing settings, identity controls, and overall information governance.
An AI pilot should include security testing in addition to functionality testing.
Don’t only ask, “Does this workflow produce the result we expected?”
Also ask, “Can someone who shouldn’t see this information retrieve it?”
Mark: And be careful about putting sensitive company information into unapproved public AI tools.
Before connecting business systems or uploading sensitive information to an AI service, understand where your data is going, how the service handles it, what permissions apply, and what your organization’s policies allow.
What Should You Do If Your First AI Initiative Failed?
Mark: Congratulations. You’re not alone.
A common problem is starting with the product instead of the business process.
If your first AI initiative stalled, go back and ask: What exactly were we trying to accomplish? Was the workflow ready? Was the data available? Did users understand how to use the tool? Did we define success?
And don’t start too big.
Start with a smaller workflow such as meeting preparation, SOP review, document processing, or another repeatable task where you can clearly measure the result.
Megan: Training matters too.
You can’t simply deploy an AI tool and assume everyone will immediately understand how to use it.
Employees need examples, prompting guidance, practical use cases, and some time to build AI into their everyday habits.
If someone tries AI a few times, gets poor results because the prompt wasn’t well constructed, and doesn’t understand how to improve it, they’re likely to stop using the tool.
That’s why adoption, training, and change management need to be part of the AI strategy.
How Should Businesses Think About AI Costs and Usage?
Mark: As AI adoption grows, organizations also need to think about usage and cost governance.
Small productivity prompts may represent relatively modest usage, while larger enterprise AI workloads, custom agents, integrations, and advanced models can introduce additional consumption and cost considerations.
Businesses should monitor AI usage, understand which models and services they’re consuming, and establish appropriate cost controls as adoption expands.
Megan: This becomes part of AI administration and governance.
IT administrators should understand which AI capabilities are being introduced, how they’re being used, what controls should be applied, and how usage and costs will be monitored over time.
How Do You Keep an AI Initiative from Stalling?
Megan: Don’t try to boil the ocean.
Don’t start with the broadest possible deployment.
Pick a specific workflow or task. Define the problem. Establish a baseline. Pilot the solution. Measure the outcome. Refine it.
Once you’ve proven that workflow, then scale.
The timesheet example is a good illustration. They addressed one portion of the process first. Now that it’s working, they can look at the next opportunity for automation.
Mark: Consistency and effort over time are how you get there.
Start small, establish some success, create a roadmap for what comes next, and keep improving.
Don’t try to boil the ocean.
Final Takeaway: Start with the Process, Not the AI Product
Moderator: That’s all the questions we have for today.
We’ll follow up with the webinar recording as well as a curated collection of practical Microsoft 365 prompts covering applications such as Outlook, Excel, Word, and other tools to help you begin experimenting with practical AI use cases.
Thanks again for joining us.
Mark: Thanks, guys.
Megan: Thank you.