Expert guide

Classify Emails and Tickets in Microsoft Access with AI

From a shared inbox nobody owns to a prioritised queue linked to your customers and orders

Free Cost Audit

Call us at +1 (385) 386-3860

Short answer: capture new emails from a shared Outlook folder into an Access tblTickets table with VBA, send each subject and body to an AI model that returns category, priority, customer or order reference and a short summary, then match the reference to your customer table and show the result on a queue form. Keep a confidence field and a review view for uncertain tickets. We build this as part of AI workflow automation.

Shared inboxes such as info@, support@ and orders@ are where work goes to wait. Nobody knows who has replied, urgent messages sit behind newsletters, and the only report is "the inbox feels full". If your customers, orders or jobs already live in Access, the inbox can feed the database directly, with AI doing the first read of every message.

The Ticket Table

FieldTypeSource
TicketIDAutoNumberGenerated
EntryIDShort Text (255)Outlook's unique message ID, prevents duplicates
ReceivedOn, FromAddress, Subject, BodyDate/Time, Text, Text, Long TextOutlook
Category, PriorityShort TextAI, editable by staff
ReferenceShort TextAI-extracted order, invoice or job number
CustomerIDNumberMatched from FromAddress or Reference
SummaryShort Text (255)AI
ConfidenceShort TextAI: high, medium, low
AssignedTo, Status, ClosedOnText, Text, Date/TimeWorkflow

Step 1: Capture Emails From Outlook

Public Sub ImportSupportEmails()
    Dim ol As Object, ns As Object, fld As Object, itm As Object
    Dim db As DAO.Database, rs As DAO.Recordset

    Set ol = CreateObject("Outlook.Application")
    Set ns = ol.GetNamespace("MAPI")
    ' Shared mailbox folder: adjust names to your mailbox
    Set fld = ns.Folders("support@yourcompany.com").Folders("Inbox")

    Set db = CurrentDb
    Set rs = db.OpenRecordset("tblTickets", dbOpenDynaset)

    For Each itm In fld.Items
        If itm.Class = 43 Then                       ' 43 = olMail
            If DCount("*", "tblTickets", "EntryID='" & itm.EntryID & "'") = 0 Then
                rs.AddNew
                rs!EntryID = itm.EntryID
                rs!ReceivedOn = itm.ReceivedTime
                rs!FromAddress = itm.SenderEmailAddress
                rs!Subject = Left$(itm.Subject, 255)
                rs!Body = Left$(itm.Body, 30000)
                rs!Status = "New"
                rs.Update
            End If
        End If
    Next itm
    rs.Close
End Sub

For busy mailboxes, restrict to recent items with fld.Items.Restrict("[ReceivedTime] >= '" & Format(Date - 2, "ddddd h:nn AMPM") & "'") instead of scanning the whole folder every time.

Step 2: The Classification Prompt

Classify this customer email for a building-supplies distributor.
Return exactly one line:
category | priority | reference | summary | confidence

category: Order status, New order, Billing, Returns, Complaint,
          Technical question, Supplier, Spam, Other
priority: High if delivery is today or overdue, a site is waiting,
          a safety issue, or a legal or cancellation threat.
          Medium for anything needing action this week. Low otherwise.
reference: any order (SO-#####), invoice (INV-#####) or PO number; else blank
summary: max 15 words, what the customer needs
confidence: high, medium or low

Subject: {subject}
Body: {first 3,000 characters of body}

Call it with the AskAI function from connecting MS Access to the OpenAI API, split the pipe-separated result with Split(), and write each part to its field. Truncating the body avoids paying to classify long email threads and signatures.

Step 3: Match to Customers and Orders

-- By sender address
UPDATE tblTickets AS t INNER JOIN tblContacts AS c
    ON t.FromAddress = c.Email
SET t.CustomerID = c.CustomerID
WHERE t.CustomerID Is Null;

-- By extracted order reference
UPDATE tblTickets AS t INNER JOIN tblOrders AS o
    ON t.Reference = o.OrderNo
SET t.CustomerID = o.CustomerID
WHERE t.CustomerID Is Null;

Once tickets link to customers, the queue form can show the customer's open orders, last invoice and account manager next to the email. That context is what makes the queue faster than the inbox.

Step 4: Rules That Override AI

  • Key accounts: raise priority one level when the customer's tier is Key.

  • Repeat contact: raise priority when the same sender has an open ticket older than 48 hours.

  • Spam with high confidence: auto-close, but keep the record.

  • Low confidence: route to a Review view regardless of category.

These rules live in queries, not in the prompt, so they are consistent and easy to change.

Step 5: The Queue Form

  • Continuous form sorted by Priority then ReceivedOn, filtered by Status <> Closed.

  • Colour by priority using conditional formatting.

  • Buttons: Assign to me, Reply in Outlook (opens a reply to the original via EntryID), Close.

  • A Category combo box so staff can correct AI; log corrections in a separate field to measure accuracy.

Measuring Whether It Works

MeasureHow to calculateWhat good looks like
Category accuracyShare of tickets where staff did not change the categorySteady improvement as definitions are refined
Time to first responseFirst reply time minus ReceivedOn, by priorityHigh priority handled the same day
BacklogOpen tickets by age bucketFew tickets older than a week
Volume by categoryCount per weekShows where root-cause fixes would cut volume

Category volume is often the most valuable output. If a third of emails are "Where is my order?", an automatic dispatch notification removes them entirely. Pair this with customer feedback theme analysis for a complete picture.

Scaling Up

Running Outlook automation from one desktop is fine for a small team. For larger volumes, a scheduled service reads the mailbox through Microsoft Graph and writes to the back end or SQL Server, so nothing depends on one PC being on. If tickets are part of a broader customer system, our Access CRM is a good base, and Access VBA development covers the integration work.

FAQ

AI ticket classification in Access FAQ

Capture, categories, priority and accuracy

Yes. A routine reads new messages from a shared Outlook folder, saves sender, subject, body and received time to an Access table, and sends the text to an AI model that returns category, priority, customer reference and a one-line summary. Staff then work the queue from an Access form instead of the inbox.

Between five and twelve categories that map to who handles the work, such as Billing, Order status, Technical issue, Returns, Complaint, Sales enquiry, Spam and Other. Categories that do not change who acts on the ticket add confusion without value.

Define priority with explicit rules the model can apply, for example 'High if the customer reports a system down, a safety issue, a legal threat or a missed delivery today'. Combine AI priority with data rules, such as raising priority for key accounts, and let staff override it.

For clear categories it is usually accurate enough to route automatically, with a confidence field and a review queue for uncertain cases. Track how often staff change the category in the first weeks; if corrections are frequent, refine the category definitions rather than the model.

If tickets need to link to orders, customers, jobs or assets that already live in an Access database, keeping them in the same database avoids double entry and makes reporting straightforward. Larger teams with SLA reporting, portals and many agents may be better served by a helpdesk product integrated with the database.