In this guide, you will learn how to set up a workflow in Build Concierge that automatically reviews BigChange job cards and flags those that require manual review.

Martin & Sons are a plumbing company serving commercial and domestic customers across the Midlands.

On completion of a job, plumbers fill out a job card in BigChange with details of the work carried out and notes about any follow-up work required.

Currently, these job cards are reviewed manually by the support team before being sent to customers to ensure they're clear, professional and to flag any further work.

By building this workflow in Build Concierge, the whole process can be automated.

Let's see how to build it!

This is a more advanced workflow to set up, but once it's live, it runs entirely in the background with no ongoing input required.

To begin, make sure you're logged into your Build Concierge account.

Click the Workflow Builder icon.

Click Create new workflow.

Give the workflow a name, such as Job Card Review.

Enter an optional description to provide additional context or instructions.

Click Create to open the Workflow Editor and begin building the workflow.

Drag and drop an Incoming Email Starting action into the empty space.

This will trigger the workflow each time the auto-generated email is received containing a job card as a PDF attachment.

Click the node to open the Properties panel.

Select the email address where job card emails are being received.

To learn more about how to add your own inbound email addresses, rather than using the default email address associated to the Incoming Email node, read the article entitled Email Management.

Click Save.

To learn more about the Incoming Email node, read the guide entitled Incoming Email Starting action.

Add an AI Prompt node.

This will identify which job the job card belongs to by extracting the job reference from the PDF attachment in the incoming email.

Connect the AI Prompt node to the Complete branch of the Incoming Email node.

Configure the node by clicking on it to open the Properties panel.

Under the Prompt field, write a prompt to return the job reference as an output that can be used as a variable later in the workflow.

Make sure to add the variable Text from PDF Attachments into the prompt.

Here is our example prompt, which you can copy and paste into your own workflow:

From the following information, extract me the reference, this should start ESI-

[Text from PDF Attachments]

Finally, under the Output values section, click + Add,

In the Field Name field, name the output Job Reference.

In the Data type field, keep the Text option selected unless your job references are numerical.

Click Save.

You can learn more about the AI Prompt node in the guide entitled AI Prompt action.

Add another AI Prompt node to your workflow, to review the content of the job card itself.

Connect the nodes and open the Properties panel.

In the Prompt field, add a prompt that asks the AI to review the job card in the PDF attachment against defined quality criteria and to return a structured result for each check.

Below is our suggested prompt, which you can copy, paste and edit to fit your needs:

You are an assistant reviewing an electrical safety inspection jobcard completed by a building services engineer.

Your task is NOT to rewrite or correct the text.

Your task is to REVIEW the jobcard and decide whether it:

- Can be sent to the customer without manual review

- Requires manual review before sending

- Requires follow-up work or clarification

You must assess the jobcard exactly as written.

REVIEW CRITERIA:

1. Clarity & Language

- Is the wording clear enough for a customer to understand?

- Does spelling, grammar, or shorthand reduce clarity?

2. Technical Certainty

- Are supply type, earthing, bonding, and protection clearly confirmed?

- Are there statements of uncertainty or assumptions?

3. Safety & Compliance Risk

- Are there indications of potential safety issues?

- Are any potentially serious defects mentioned?

- Are protective measures (e.g. RCDs) incomplete or unclear?

4. Follow-Up Indicators

- Is further investigation required?

- Are remedial works recommended?

- Does the engineer's overall safety judgement include uncertainty?

RULES:

- Do NOT rewrite or correct the jobcard text

- Do NOT invent defects or conclusions

- Base your assessment only on the provided content

- Use professional, neutral language

OUTPUT FORMAT (MANDATORY):

Return the following sections:

1. Review Outcome

   - One of:

     - "OK to send to customer"

     - "Manual review required"

     - "Manual review required + follow-up work indicated"

2. Reasons for Review Outcome

   - Bullet-point list referencing specific issues found

3. Risk Flags

   - Bullet-point list using categories:

     - Language / Clarity

     - Technical Uncertainty

     - Safety / Compliance

     - Follow-Up Required

4. Recommended Next Action

   - One short paragraph describing what should happen next

     (e.g. engineer clarification, remedial quote, supervisor review)

INPUT:

The following text is the completed jobcard. Review it as-is.

[Begin Jobcard]

attachment text

[End Jobcard]

To add the Text from PDF Attachments variable to the prompt, click + Add a variable.

Then select the Text from PDF attachments output from the Incoming Email node.

Define the outputs, which will determine whether any follow-up action is needed in relation to the job card.

Click + Add to add additional outputs.

In this example, we added four outputs:

• Review outcome

• Reason for review outcome

• Risk flags

• Recommended next action

Click Save.

Add an Extract Text node, which will check for the manual review value in the Review Outcome output.

Connect the Extract Text node to the Complete branch of the AI Prompt node.

Click the node to open the Properties panel.

Map the Review outcome variable into the Text to extract from field.

Click + Add to add a regular expression to extract the manual review outcome.

Name the output Review Outcome.

Enter the Regex which will extract the manual review required output from the Review outcome output.

Since the output from the AI Prompt is manual review required, the regex to use is:

(manual review required).

Click Save.

From here, if the phrase is found, meaning that manual review of a job card is required, the workflow will continue to the next step of the workflow to find the job in BigChange.

If it isn't found, the workflow ends with no further action.

To learn more about the Extract Text node, read the guide entitled Extract Text action.

Next, add the Find a Job in BigChange action.

This will look for a job and, if found, will output the Job ID needed to add a flag.

Select the Job Reference in the Search Field.

Click + Add a variable.

Map the Job Reference variable that was extracted by the first AI Prompt node.

Press Save.

To learn more about the Find a Job in BigChange node, read the guide entitled Find a Job in BigChange action.

Finally drag and drop an External Action node into the empty space.

This is what will be used to add a flag to a found job.

Connect the External Action node to the Found branch of the Find a Job in BigChange node.

In the URL field, add the relevant call, which you can find on BigChange API calls page.

As part of the call, add the Job ID variable output by the Find a Job in BigChange node.

As part of the call, also add the Reason for review outcome variable, which was defined as an output in the AI prompt node.

When manual review is required, as identified in the workflow, the reasons provided by the AI Prompt node for such outcome will now be added as a flag to the job.

In the Request Method field, select GET.

Click Save.

To learn more about the External Action node, read the guide External Action action.

Now with this workflow in place, Martin & Sons no longer need to manually review every completed job card.

Instead, each PDF is automatically read, assessed against defined quality criteria and only flagged in BigChange when follow-up or amendments are required.

Now, it's your turn, try building this workflow yourself and save your team hours of work.