Configra — AI deal & configuration builder for Revenue Cloud
A Salesforce RLM managed package with built-in Einstein AI: describe a deal in one sentence and Configra creates the Account, Contact, Opportunity and Quote; describe products in plain language and it builds the configuration lines — all without polluting the catalog. Don't lose the deal because the catalog is missing one SKU.
- Year
- 2026
- Role
- Salesforce ISV Partner — Founder & Developer
- Tech stack
- Revenue Lifecycle ManagementEinstein Models API (GPT-4o mini)Natural-language / Generative AIPrompt-injection hardeningPlatform Events (empApi / CometD)GenAiFunction (Agentforce action)k-anonymity / DSGVOCustom PermissionsApexLightning Web ComponentsManaged 2GP PackageAppExchangeEphemeral Custom ObjectsFLS/CRUD enforcedPMD-clean
Highlights
- AI deal builder — one plain-language sentence creates the Account, Contact, Opportunity, Quote and configuration (native Einstein Models API, GPT-4o mini — no external API key)
- AI line builder — 'describe it in plain language' turns free text into structured config lines, with a deterministic multilingual fallback so it never hard-fails
- Prompt-injection hardened — the LLM never touches prices or discounts; those are parsed deterministically from the rep's verbatim text. OWASP- and Unicode-aware sanitizer (CSV/formula injection incl. fullwidth, XSS, zero-width/bidi, surrogate-safe caps), strict-JSON, FLS/CRUD-enforced
- Privacy-by-design Catalog Gap Intelligence — cross-rep aggregation enforces a hard k-anonymity floor (k≥5, NIST/ENISA), bucketed prices, and never stores deal/account identifiers (DSGVO data-minimization)
- Real-time Platform Event UX — an inline-edit guard publishes a Platform Event that an LWC catches over empApi/CometD to redirect the rep straight to the right Configra edit form
- Agentforce R&D — built a custom GenAiFunction action ('Create Configra Configuration') + an internal Employee Agent, then kept out of the shipped package so customers need no Agentforce license
- Salesforce ISV Partner — Managed 2GP package on AppExchange (in Security Review)
- Quote-native configuration: deal-specific products without polluting the catalog
- Multi-line configurations with real-time net pricing and a per-line pricing widget
- Reusable template library (private or shared) + catalog gap detection
- Convert-to-RLM-product for proven write-in patterns
- Seven UI languages: EN, DE, ES, FR, IT, PT, JA
- Security Review-ready: PMD-clean, FLS/CRUD enforced
The problem
When a customer asks for a product or bundle that isn't in the catalog, reps are stuck with bad options: bloat the catalog with one-off SKUs, wait on an IT ticket, skip the formal quote, or fall back to untraceable spreadsheets. Each choice either pollutes Revenue Cloud's catalog — the governed source of truth — or kills the momentum of the deal.
The approach
Two Einstein AI entry points sit on top: a Home-page deal builder turns one sentence into a full Account → Contact → Opportunity → Quote → configuration, and an on-quote natural-language builder turns free text into structured lines — both on the native Einstein Models API (GPT-4o mini, no external API key), with a deterministic multilingual fallback so they never hard-fail and prompt-injection hardening where the LLM never touches monetary fields (prices and discounts are parsed deterministically from the rep's verbatim text). Configra models in-progress configuration in ephemeral, session-scoped custom objects designed to be thrown away, so the durable RLM catalog never takes part in the scratch work. Reps configure multi-line products on the quote with real-time net pricing and a per-line pricing widget, save proven patterns to a reusable template library, and — when a write-in keeps recurring — promote it to a real RLM product. Catalog gap detection surfaces what's worth productizing. The package is built clean: PMD-clean, with FLS/CRUD enforced and ready for AppExchange Security Review.
The outcome
A Salesforce ISV Partner solution, distributed as a Managed 2GP package on AppExchange and currently going through Security Review. Designed so reps can close faster — describe a deal or its lines in plain language and the built-in Einstein AI builds them — without write-in friction, while the Revenue Cloud catalog stays clean and governable. The UI ships in seven languages (EN, DE, ES, FR, IT, PT, JA).