CRM AI Agent Development for Software Company
- AI Agents AI Development CRM Consulting and SI
- Less than $10,000
- Nov. 2025 - Mar. 2026
- Quality
- 5.0
- Schedule
- 5.0
- Cost
- 5.0
- Willing to Refer
- 5.0
"The entire team was amazing, collaborative, and expeditious in completing the project."
- Other industries
- San Jose, California
- 1-10 Employees
- Online Review
- Verified
Synebo developed an AI agent in Salesforce for a software company. The team built a GPT-powered assistant that helps users ideate, draft, and refine components and generate parent records for the client.
Synebo's work reduced manual tasks and provided an efficient way to interact with the client's Inbound and Outbound pipelines. The team delivered a flexible and scalable solution for current and future needs. Synebo's project management was excellent, and they delivered key milestones on time.
The client submitted this review online.
BACKGROUND
Please describe your company and position.
I am the Founder of ST8MNT LLC
Describe what your company does in a single sentence.
ST8MNT has created a suite of Saleforce native apps that manage Statements of Work. Users can create, manage and track performance (ST8MNT Inbound; ST8MNT Outbound), keep track of agreement renewals, amendments and terminations (AGRE2MNT) and ideate ST8MNT content using our AI assistant, NAVIG8R.
OPPORTUNITY / CHALLENGE
What specific goals or objectives did you hire Synebo to accomplish?
- Needed an AI agent in Salesforce to assist users preparing, ideating, updating, and recommending new content and content changes in Salesforce object records (ST8MNT Inbound and ST8MNT Outbound). NAVIG8R can create new records and associate dependencies between them.. The goal was to reduce manual work and help the new users quickly realize the benefits of ST8MNT and NAVIG8R..
SOLUTION
How did you find Synebo?
Online Search
Why did you select Synebo over others?
- Great culture fit
- Good value for cost
- Company values aligned
How many teammates from Synebo were assigned to this project?
2-5 Employees
Describe the scope of work in detail. Please include a summary of key deliverables.
ST8MNT Navigator – Detailed Process Description
1. High‑Level Summary
ST8MNT Navigator is a GPT-powered assistant surfaced in Salesforce via Agentforce. It assists users in ideating, drafting, and refining SOW components (Deliverables, Resources, Change Orders) and can also generate parent ST8MNT records. Navigator is launched via a Global Action and does not auto-detect record types. Users provide context, Navigator generates structured outputs aligned to ST8MNT fields, and—when requested—invokes Salesforce Flows to write suggestions back into records.
2. User Interaction & Entry Points
• Entry Point: Global Action available in the Salesforce header and on relevant Lightning record pages.
• Record Awareness: Navigator does not auto-detect Inbound/Outbound; users declare context in the prompt (e.g., “Inbound SOW, Vendor X”).
• Modes:
– Drafting Mode: Generate Deliverables, Resources, Change Orders, or a parent ST8MNT outline.
– Insert Mode: Send the generated text to a Salesforce Flow for field-level writes into draft records.
– Review Mode: Summarize current SOW content and highlight gaps (read-only).
3. Functional Flow (Step‑by‑Step)
3.1 Launch & Context Capture
1) User clicks the ‘Open ST8MNT Navigator’ Global Action.
2) Navigator opens in an Agentforce panel and presents conversation starters (e.g., “Draft a milestone deliverable”).
3) User provides context explicitly (e.g., app type, project name, scope drivers, constraints).
3.2 Ideation & Drafting
4) Navigator generates structured content aligned to ST8MNT field definitions (e.g., Deliverable Name, Type, Scope, Acceptance Criteria, Dependencies, Schedule, Price).
5) User can refine via follow‑up prompts within the same interaction (no persistent session memory beyond the current interaction).
3.3 Insert to Salesforce (via Flow)
6) User clicks ‘Insert to Salesforce’.
7) Agentforce triggers a named Salesforce Flow with input variables (e.g., recordId, objectType, mapped fields from Navigator output).
8) Flow creates or updates draft ST8MNT records (parent or child objects) and returns success/failure to Navigator.
9) Navigator displays a success toast and offers a deep link to the created/updated record.
3.4 Review & Summarize
10) On demand, Navigator calls a Summary Flow/Apex to pull current draft data (read-only) and produces a human-readable summary of what’s in the SOW and what’s missing.
11) Summary text is written back to a logging object and optionally shown as a ‘SOW Summary’ section on the record.
4. Data Interaction Model
• Input: User-provided context in plain language; Navigator does not auto-read record fields.
• Output (Text): Structured sections for Deliverables/Resources/Change Orders/Parent ST8MNT, formatted to match field-level schemas.
• Writeback (Optional): Through a Salesforce Flow (invoked by Agentforce) that maps Navigator outputs to fields, validates values, and creates/updates draft records.
• Read-Only Summaries: A separate Flow/Apex composes record summaries for Navigator to present without direct data writes.
5. Agentforce Integration & Architecture
• UI Host: Agentforce panel launched by a Global Action.
• Invocation Pattern:
– Generate: Navigator runs prompts and returns text to the panel.
– Insert: Agentforce calls a named Screenless Flow with input variables; the Flow performs DML and responds.
• Mapping Layer: A lightweight mapping utility (custom metadata) maintains field-name alignment across Inbound/Outbound and future apps.
• Error Handling: Flow returns structured errors (e.g., missing required fields); Navigator surfaces them and offers fix prompts.
• No Auto-Detection: Record type selection is user-driven (keeps behavior predictable and controllable).
6. Security & Privacy
• Data Access: Navigator itself operates with read-only access to Salesforce data; writes occur only via Salesforce Flow invocations with org-side permissions.
• Authentication: Flows run in user context or an integration user context depending on your security model.
• Data Residency: Prefer keeping interactions within Salesforce/Agentforce where possible; any external LLM calls should avoid sending secrets/PII and can be governed by prompt/response minimization.
• Governance: Maintain an allowlist of fields eligible for AI-assisted writeback; audit writes through standard Salesforce field history and custom logs.
7. Telemetry & Logging
• Log each interaction: prompt hash, timestamp, userId, target object, Flow outcome, and record links.
• Store summaries on a ‘Navigator Interaction’ custom object related to the parent ST8MNT record.
• Admin insights: dashboard cards for usage by app (Inbound/Outbound), completion rates, and error types.
• Compliance: Configurable retention window (e.g., 90 days).
8. Implementation Plan
Phase 1 – Enable & Prototype (Week 1)
• Global Action + Agentforce panel scaffold
• Initial prompt templates (Deliverables, Resources, Change Orders, Parent ST8MNT)
• Screenless Flow v1: create draft child records (no updates)
Phase 2 – Field Mapping & Insert (Weeks 2–3)
• Custom Metadata for field mappings per record type/object
• Screenless Flow v2: create/update draft records; return IDs & errors
• Basic logging object + usage dashboard
Phase 3 – Summaries & Hardening (Weeks 4–5)
• Summary Flow/Apex: compile read-only SOW summaries
• Error handling & validations, guardrails for required fields
• Admin controls: enable/disable writeback per profile/permission set
9. Acceptance Criteria
• Users can launch Navigator via Global Action from anywhere in Salesforce.
• Navigator generates field-aligned text for the selected ST8MNT object(s).
• ‘Insert to Salesforce’ successfully creates/updates draft records through Flow with clear success/failure messages.
• Interaction logs and summaries are captured and reportable.
• No auto-detection of record type; user selection governs behavior.
10. Non‑Goals (v1)
• No persistent cross-record session memory.
• No direct Apex REST endpoints for writes (all writes via Flow).
• No automatic ingestion of existing record fields into prompts (context is user-provided).
11. Open Decisions & Recommendations
• External vs. Internal Inference: If using an external LLM, implement data minimization and mask sensitive values; consider using guardrails and an allowlist for fields eligible for writeback.
• Flow Context: Prefer running the insert/update Flow in USER context with Permission Sets that mirror app entitlements; use an Integration User only where necessary.
• Mapping Strategy: Use Custom Metadata Types to manage field mappings by app/object; expose a simple admin UI for maintenance.
• Future: Optionally add Platform Events for decoupled notifications (e.g., ‘NavigatorDraftCreated’) to trigger downstream automations.
RESULTS & FEEDBACK
What were the measurable outcomes from the project that demonstrate progress or success?
1) Reduced manual work for ST8MNT record creation or updates.
2) An efficient way to interact with ST8MNT Inbound and ST8MNT Outbound.
3) Flexible, scalable solution for current and future needs.
4) Delivered as an AppExchange managed package for easy installation and maintenance across orgs.
Describe their project management. Did they deliver items on time? How did they respond to your needs?
Synebo’s project management was excellent. They delivered key milestones on time and maintained clear, consistent communication throughout the engagement. They were highly responsive to my needs, adapting quickly to feedback and helping refine requirements as NAVIG8R evolved. Their ability to balance structure with flexibility made the development process smooth, collaborative, and well-aligned with my expectations. I will definitely utilize their resources in the future!
What was your primary form of communication with Synebo?
- Virtual Meeting
- Email or Messaging App
What did you find most impressive or unique about this company?
What stood out to me the most was their ability to translate a conceptual AI-driven product like NAVIG8R into a practical, working solution within Salesforce. They understood both the technical and user experience aspects, and were able to bring structure and clarity to something that initially started as a high-level idea.
Are there any areas for improvement or something Synebo could have done differently?
Nothing that comes to mind - the entire team was amazing, collaborative, and expeditious in completing the project.
RATINGS
-
Quality
5.0Service & Deliverables
"High-quality work throughout: deliverables were well-designed, reliable, and aligned closely with requirements."
-
Schedule
5.0On time / deadlines
"Consistently met deadlines with clear communication and proactive updates on progress and milestones."
-
Cost
5.0Value / within estimates
"Strong value for the investment, with high-quality output, thoughtful input, and a true partnership approach."
-
Willing to Refer
5.0NPS
"I would unreservedly recommend Synebo for any Salesforce development.