How does a Dock agent run a sourcing analysis against Workday Strategic Sourcing?
The agent watches for the close of a Workday Strategic Sourcing event, pulls every supplier bid via the API, and writes one row to a Dock sourcing_analyses table. The row carries the recommended supplier, the modeled annual savings, the assumptions, and the diff against the incumbent. The procurement lead approves or rejects in Dock. Only on approval does the agent write the award back to Workday Strategic Sourcing. SAP Ariba follows the same shape.
The architecture
Workday Strategic Sourcing and SAP Ariba stay the system of record for the raw sourcing data: the RFx structure, supplier responses, bid line items, attachments, and event timeline. Dock is the system of record for what the agent interpreted from that data. Each Dock row carries a pointer back to the platform record (workday_sourcing_event_id or ariba_event_id), the agent identity that produced the analysis, the recommendation, the reviewer, and the timestamp. The agent re-fetches the platform event whenever it needs the current bid state.
The Dock surface
A single table, sourcing_analyses, holds one row per closed event.
| event_id | platform | category | incumbent_cost | recommended_supplier | modeled_savings | confidence | agent | reviewer | status |
|---|---|---|---|---|---|---|---|---|---|
| WSS-4471 | workday | Contract logistics, US Midwest | $4.2M | Penske Logistics | $612K (14.5%) | High | argus.proc | Lena O. | approved, awarded |
| WSS-4498 | workday | SaaS observability renewal | $890K | Datadog (incumbent) | $134K (15.0%) | High | argus.proc | Lena O. | approved, awarded |
| ARIBA-2210 | ariba | Print and mailing, EMEA | EUR 1.1M | Quadient | EUR 198K (18%) | Medium | argus.proc | pending | awaiting review |
Each row links back to the source event. The full bid sheet and supplier attachments stay in Workday or Ariba; Dock holds the interpretation.
A worked workflow
A logistics RFP closes in Workday Strategic Sourcing. The agent argus.proc fetches eleven supplier responses, normalizes line items, models three-year TCO against the incumbent, and writes one row to sourcing_analyses with a recommendation, a confidence score, and a memo on which assumptions could flip the answer. Lena, the procurement lead, opens the row, queries two assumptions, and approves. The approval triggers a consent gate: the agent writes the award back to Workday Strategic Sourcing and drafts the supplier notification for Lena to send. The Dock row now reads approved, awarded and links to the executed award.
Why this matters
Sourcing analyses are where procurement value gets won or lost. Hackett Group's 2025 Digital World Class benchmark found top procurement teams generate 2.03x greater cost savings as a percentage of spend and run sourcing cycles 24% shorter than peers (Hackett Group, 2025). Bid collection is not the bottleneck. Workday Strategic Sourcing already handles that with RFx, reverse auctions, and savings tracking built in (Workday, 2026). The bottleneck is analyst hours spent normalizing bids and writing the memo.
Dock makes the agent a first-class participant in that interpretation layer without letting it act unsupervised. Every recommendation is a row, every approver is named, every award is a reviewed write back to the platform of record. This is the same pattern we use for accounting close, legal review, and the broader procurement cluster.
When auditors ask who approved the Penske award, the answer is one row with a reviewer, a timestamp, and a link to the Workday event. That is the bar for agent audit and compliance.
Try the procurement pattern
Set up the pattern in your own Dock workspace and watch a Workday or Ariba event flow through agent analysis, reviewer approval, and platform write-back.
FAQ
Does the agent need write access to Workday Strategic Sourcing? Read access is sufficient until the moment of award. The award write is gated by a procurement-lead approval in Dock, after which the agent's identity is the principal making the call. See agent identity lifecycle for how that identity is provisioned.
What about SAP Ariba events?
Same shape. The Dock row carries ariba_event_id instead of workday_sourcing_event_id, and the agent uses Ariba's API for the read and award write. The interpretation layer is identical.
Where does the savings number come from? The agent models it from incumbent contract pricing pulled from your contract repository plus the winning bid line items. The assumptions are stored alongside the row so the reviewer can challenge them. Numbers that feed savings dashboards live in Dock for finance.
Can the agent award without a reviewer? No. Award is a consent-gated operation in Dock. The agent can recommend, model, and draft. The procurement lead approves. That separation is the whole point of the architecture.
