Skip to main content
The best way to understand what AI Agents can do for your factory is to see them handling real problems. The four examples below are drawn from tasks that manufacturing teams perform every day inside Bold — tasks that are necessary, repetitive, and entirely delegatable. Each example explains the problem it solves, how to configure the agent, and what you can expect once it is running.

​
Email-to-Order

The problem: A customer sends an email listing product references, quantities, and delivery terms. Someone on your team reads it, opens Bold, and manually creates the order line by line. It takes time, introduces transcription errors, and happens multiple times a day. How to configure it:
1

Write the instructions

“When I paste a customer email into this conversation, identify the product references, quantities, and delivery terms mentioned. Create a new order in Bold using those details. If any reference is ambiguous or not found in the product catalogue, flag it and ask me before creating the order.”
2

Set permissions

Read access to the product catalogue and customer records. Create access for Orders.
3

Set scheduling

On demand — the agent runs whenever a team member pastes an email into the conversation.
Expected outcome: Paste the email, review the agent’s interpretation in seconds, confirm. No manual transcription, no typos, and the order is in Bold before the customer has finished their coffee.

​
Daily Factory Report

The problem: Your plant manager needs a morning briefing: what was produced yesterday versus the plan, any anomalies, and how each operator performed. Building that view manually means opening dashboards, exporting data, and writing a summary — every single morning. How to configure it:
1

Write the instructions

“Every morning at 07:00, query yesterday’s production data and compare it against the planned targets. Identify any anomalies — lines that stopped unexpectedly, significant deviations from plan, or operations that were left open. Calculate productivity by operator. Send a structured email summary to the plant manager with these three sections: Production vs. Plan, Anomalies, and Operator Productivity.”
2

Set permissions

Read access to Production, Timesheets, and Maintenance modules.
3

Set scheduling

Daily, at 07:00 (or whichever time works best before your team’s morning standup).
Expected outcome: The plant manager receives a clear, structured email every morning without opening a single dashboard. Decisions get made faster because the data is already waiting in their inbox.
Use the powerful intelligence model for this agent — it needs to reason across multiple data sources and identify patterns, not just retrieve a single record.

​
Bulk Recipe Editing

The problem: A product change or process improvement requires updating the same manufacturing step across dozens of recipes. Doing it one by one is a full afternoon’s work and carries the risk of missing a recipe or making an inconsistent change. How to configure it:
1

Write the instructions

“I will describe a change I want to apply to a set of recipes. Identify all recipes that match my description, apply the change consistently across all of them, and give me a summary of what was modified before saving.”When you run it, you might say: “This product is made the same way as that one, except the Band Saw step is replaced by CNC and the last step is removed.”
2

Set permissions

Read and edit access to Recipes.
3

Set scheduling

On demand — you run it whenever you have a batch of recipe changes to apply.
Expected outcome: Describe the change once. The agent applies it across every matching recipe and shows you a summary to review before committing. What used to take an afternoon now takes minutes.
Always ask the agent to summarise the changes it plans to make before it saves them. This gives you a final review step without slowing down the process.

​
Forgotten Clock-Outs

The problem: At the end of a shift, some operators forget to clock out. Their operations remain open in Bold, polluting your production data and forcing someone to chase them down the next morning to correct the timesheets manually. How to configure it:
1

Write the instructions

“At 22:00 each weekday, check for any operations that are still marked as active. For each one, automatically pause the operation and add a note: ‘Clock-out error — paused automatically at end of shift.’ Do not modify any operations that were deliberately left open with a supervisor note.”
2

Set permissions

Read and edit access to Operations and Timesheets.
3

Set scheduling

Daily at 22:00 (adjust to match the end of your last shift).
Expected outcome: Clean production data every morning. No chasing operators, no manual timesheet corrections. The audit trail shows exactly what the agent did and why.

​
Identifying Your Own Use Cases

The four examples above are starting points. The real question is: what does your team repeat inside Bold every day? Any task that fits this pattern is a strong candidate for an agent:

It happens regularly

Daily, weekly, or every time a certain event occurs — if there’s a rhythm to it, an agent can run on that rhythm.

It follows a consistent logic

The steps are roughly the same each time, even if the specific data changes with each run.

It lives inside Bold

The task reads from or writes to Bold data — queries, reports, record updates, process closures.

It takes time away from higher-value work

If a skilled team member is spending an hour a day on it, that’s an hour that could go elsewhere.
When you spot a candidate, write down what the task involves — even a rough description is enough. Bring it to your Account Manager and they will help you turn it into a working agent. After your first one is running, you will find it natural to build the next ones independently.