Ageing Card Requests: A Shared Dust Agent Finds the Bottlenecks
Card requests sit blocked in approval queues without the procurement team knowing. A Dust agent published for the whole procurement team maps every pending request in a single question.
The Problem
Card requests are how teams ask to spend: a virtual card for a purchase or a subscription, or a top-up. (The Spendesk MCP reads card requests only, not purchase order requests.) In practice, they frequently sit blocked in approval queues for days or even weeks — without the procurement team or the requester knowing exactly where the hold-up is.
The problem stems from a lack of centralised visibility. Each approver sees their own queue, but nobody has a cross-functional view: how many requests have been pending for more than a week? Who is the bottleneck? Which departments are accumulating unprocessed requests? Answering these questions today requires exporting data, joining multiple tables, and manually building a consolidated view.
In the meantime, operational teams escalate by email or Slack, generating noise and straining the relationship between procurement and business units. Urgent requests are not visually prioritised. And when a €50,000 request has been sitting in an absent manager's queue for three weeks, nobody knows.
A shared agent for the whole procurement team
With the Spendesk MCP connected once, workspace-wide, by an admin, a Dust agent configured for this analysis is immediately usable by the whole procurement team — no need for each buyer to rebuild the same consolidated view in a spreadsheet.
Show me all card requests that have been pending for more than 7 days. Sort by oldest first.
I identified 14 card requests pending for more than 7 days (`get_requests` covers card requests only; purchase-order requests are not included):
| Request | Requester | Department | Amount | Pending Since | Awaiting Approver |
|---|---|---|---|---|---|
| REQ-2026-0891 | Lucas Bernard | IT | €48,000 | 22 days | Thomas Renard |
| REQ-2026-0834 | Camille Vidal | Marketing | €12,400 | 19 days | Marie Dupont |
| REQ-2026-0867 | Florent Masse | Engineering | €8,900 | 17 days | Thomas Renard |
| REQ-2026-0902 | Sandra Klein | HR | €6,200 | 16 days | Paul Girard |
| REQ-2026-0915 | Marc Jouve | Sales | €5,800 | 14 days | Marie Dupont |
| REQ-2026-0923 | Inès Moreau | Finance | €4,400 | 13 days | Thomas Renard |
| REQ-2026-0941 | Romain Faure | Product | €3,100 | 12 days | Paul Girard |
| REQ-2026-0958 | Léa Dumont | CS | €2,800 | 11 days | Sophie Blanc |
| … | … | … | … | … | … |
Would you like me to prepare a summary of follow-up messages to send to each approver, sorted by urgency?
Going Further: 2 Additional Use Cases
Approver With the Most Unprocessed Requests
Identifying which approver holds the most pending requests helps determine whether a bottleneck is structural (scope too wide) or circumstantial (absence, busy period).
Which approver has the most unprocessed card requests pending?
Approver ranking by number of pending requests:
| Approver | Pending Requests | Total Amount | Average Wait Time |
|---|---|---|---|
| Thomas Renard | 6 requests | €74,100 | 15.3 days |
| Marie Dupont | 4 requests | €21,800 | 12.1 days |
| Paul Girard | 3 requests | €14,200 | 9.4 days |
| Sophie Blanc | 2 requests | €5,600 | 8.2 days |
| Others (5 approvers) | 4 requests | €6,100 | 7.1 days |
Would you like me to identify the request types that systematically go to Thomas Renard, to evaluate where delegation is possible?
Requests Over €10,000 Blocked for More Than 14 Days
Large blocked requests represent both an operational risk (project delays) and a financial risk (unformalised commitments that may lead to off-process purchases).
Are there any card requests over €10,000 that have been waiting for more than 14 days?
I identified 4 card requests above €10,000 pending for more than 14 days:
| Request | Description | Amount | Department | Pending Since | Approver |
|---|---|---|---|---|---|
| REQ-2026-0891 | IBM Maximo licences | €48,000 | IT | 22 days | Thomas Renard |
| REQ-2026-0834 | Q3 display campaign | €12,400 | Marketing | 19 days | Marie Dupont |
| REQ-2026-0867 | Kubernetes infrastructure | €8,900 | Engineering | 17 days | Thomas Renard |
| REQ-2026-0902 | Leadership training 2026 | €6,200 | HR | 16 days | Paul Girard |
Would you like me to generate an escalation email for each of these approvers with the full request detail?
Automating the Monday follow-up
Once this analysis has been validated manually, a scheduled Trigger can re-run it every Monday morning and post the result directly to the #procurement channel via @Dust — the whole team sees the blockers without anyone having to ask the question themselves.
Key Takeaways
- A Dust agent published on the workspace saves every buyer from rebuilding the same consolidated view.
- Card requests pending more than 7 days signal an approval bottleneck —
get_requests(withwithApproval) +get_userssurface them in a single question. Purchase-order requests are not returned by this tool. - A single approver can hold a disproportionate share of pending requests — visibility on this bottleneck enables action (delegation, reminder, prioritisation).
- Requests over €10,000 blocked for more than 14 days create bypass risk (off-process purchases) and project delivery risk.
- A weekly Trigger turns this check into an automatic routine shared by the whole team.
Related skills