Video summary

AgentforcePharma Grid: Salesforce That Thinks Like a Human 🤯

Main summary

Key takeaways

Technology

Summary of technological concepts & product features (AgentforcePharma Grid / Salesforce Agentforce Grid)

Problem addressed (pharma inventory transfers)

Inventory transfers are more than logistics—they must account for:

  • Compliance
  • Timing
  • Risk
  • Stock availability

Bad decisions can lead to:

  • Regulatory violations
  • Expired medications
  • Stockouts

Core idea

AgentforcePharmaGrid embeds “decision intelligence” directly into the Salesforce UI (described as “embedded directly into your Salesforce UI”), providing:

  • Constraint evaluation
  • Recommendations
  • Guardrails (blocking when required)

Key configurable parameters / rules introduced

  • IST (Inter-Store Transfer): Moving medication from one pharmacy location to another.
  • Safety stock: Minimum inventory that must remain on hand. If inventory drops below this threshold, the system flags low stock.
  • DEA Schedule II controlled substances: Enforces a hard stop—it will never automatically move Schedule II drugs (e.g., Adderall). Users attempting transfer receive a compliance message to prevent human error.
  • Cold chain requirement: For temperature-sensitive meds (e.g., insulin, Mounjaro), the system checks that both source and destination stores have refrigeration.
  • Transfer log: A historical audit trail of transfers for DEA review/reference.

The “six scenarios” demonstrated (recommendation + blocking + fallback logic)

  1. Happy path (simple recommended transfer)

    • If a store is out of stock, the grid identifies the best source based on capabilities such as cold-chain handling.
    • Example outcome: selects a store that satisfies cold-chain needs.

  2. Schedule II compliance block (hard stop)

    • For Schedule II drugs (e.g., Adderall XR 30mg), the system blocks transfer even if the user clicks the action.
    • The user sees a restriction/compliance message, and the transfer does not start.

  3. No suitable source → fallback distributor recommendation

    • If no store can meet the need (including safety stock constraints), the system does not fail silently.
    • Instead, it recommends a fallback distributor order and displays escalation guidance (e.g., contact a district manager to expedite).

  4. Expiry-based filtering (avoid near-expiry inventory)

    • When multiple stores can supply, the system excludes risky/near-expiry stock.
    • Example outcome: selects the source with longer shelf life, then logs the transfer and updates quantities.

  5. Schedule IV allowed transfers (rules-based permissions)

    • Not all controlled substances are blocked. Schedule IV transfers are allowed under defined rules.
    • Example: Xanax 0.5mg transfers successfully because the Schedule II blocking gate does not apply.

  6. Multiple-source optimization (best match selection)

    • If multiple sources are valid, the system optimizes rather than selecting randomly.
    • Example outcome: chooses the store with the highest available quantity (“highest quantity wins” / best based on quantity and efficiency).

Tutorial / setup steps shown (how to try the demo)

Accessing the demo

  • Use a QR code to navigate to the repo documentation.
  • In the repo docs, use the “live demo” link.

Provisioning a scratch org to run the code

  1. Fill a form (requires GitHub username and a fork of the repository).
  2. Provide:
    • Fork URL
    • Dev Hub URL (retrieved via instructions in the video)
  3. Click launch demo and choose scratch org validity duration.
  4. Wait about 10–15 minutes for scratch-org creation.
  5. Use the emailed credentials (login URL + username + password).

Demo data seeding & automation

  • Demo data is seeded into the scratch org (including a GIF for illustration).
  • A prompt-builder template step is created manually because metadata direct deployment isn’t available.
  • The presenter notes they automated activation and subsequent steps.

Grid configuration in Salesforce

  • Use a separate app called Agentforce Grid.
  • Create/configure an Agentforce Pharma Grid (example name: Inventory Transfer Ops).
  • Configure grid columns including:
    • Medication name, store name, safety stock, quantity, expiry date, status, and status color
    • A recommendation column using a prompt template such as:
      • “Enter store transfer inventory recommendation.”
    • A JSON output column (visibility configurable)
    • An invocable action named “transfer optimize” executing “Execute Inter-Store Transfer”
  • Operational detail:
    • Avoid “execute now” behavior for all rows—initiate the action per-row from the grid.

Reviews / analysis / results highlighted

  • The demo validates each scenario by showing:
    • Grid outputs (recommendation JSON, blocked/compliance messages, and “no suitable source”)
    • Transfer initiation behavior (sometimes requiring multiple clicks; noted as potentially improvable later)
    • Transfer logs and inventory record updates, including expiry-based selection
  • The presenter emphasizes the distinction between:
    • Optimization + decision intelligence (multi-constraint selection)
    • vs. simple automation

Main speakers / sources (as stated)

  • Speaker: Jayavardhan Raju (“Jay”) — Salesforce application architect, 15+ years experience.
  • Primary product/platform source: Salesforce Agentforce Grid / AgentforcePharma Grid (pharma use-case repository + documentation, accessed via QR/repo).

Original video