Currently shipping AI-native work at Assurant

Agentic AI, applied to enterprise Salesforce.

Autonomous SDLC agents. Headless 360 via Salesforce Hosted MCP Servers. External Client Apps authenticating Claude as a first-class agent. Vibe-driven development with Claude Code and Agentforce Vibes. AI agents that perform real enterprise work — not demos.

3
Autonomous SDLC agents shipped
(Dev · QA · DevOps)
2 days
to rebuild a 2-month OmniStudio app
using Claude Code
2
Salesforce MCP Servers in production
(SObject Reads + Apex REST)
18+ yrs
enterprise Salesforce context
behind every agent

Most Salesforce orgs aren't using AI well — yet.

Agentforce is shipping. Salesforce Hosted MCP Servers are now generally available. Claude Code and Agentforce Vibes are reshaping how Salesforce work gets done. But most enterprise teams are still running 2022-era patterns: manual code reviews, hand-written test classes, weeks-long migrations.

I'm bridging the two worlds — applying agentic AI to the slow, manual parts of the Salesforce SDLC, while building Agentforce skills, MCP-backed agents, and Headless 360 architectures that integrate with real enterprise Salesforce data.

The premise: deep OmniStudio and CPQ roots across Rogers, Telus, Verizon, McKesson, and Windstream form a foundation that domain generalists cannot replicate. The agents I build know what enterprise Salesforce actually looks like — because I've shipped at that scale for a decade.

“Pioneering use of Claude Code and Agentforce Vibes for vibe-driven development — directing AI agents to handle implementation while personally focusing on architecture, review, and governance.”
— Technical Architect, AI & Agentic Systems · Jun 2025–present

Six ways I bring agentic AI into your enterprise.

Each offering is grounded in production work I'm shipping today — not a theoretical capability deck.

Autonomous Salesforce SDLC Agents

Architect Dev/QA/DevOps agents that ingest Azure DevOps work items, execute end-to-end on Salesforce orgs, and emit structured resolution summaries. Human approval gates between stages enforce governed execution at every critical step. Currently in production at Assurant.

● Live at AssurantClaude CodeAzure DevOpsAgent Script

Headless 360 via Salesforce MCP Servers

Configure Salesforce Hosted MCP Servers — SObject Reads for natural-language SOQL access, Apex REST for custom service invocation — and provision an External Client Application (ECA) that authenticates Claude (or any external AI agent) for direct conversational access to live org data. No browser, no manual login.

● Live at AssurantMCP ServersECAOIDC

Vibe-Driven Development

Direct Claude Code and Agentforce Vibes to handle implementation while you focus on architecture, review, and governance. Validated on Salesforce: rebuilt a 2-month OmniStudio app as a production-quality LWR application in 2 days, resolving every known pain point of the original.

● Doing this nowClaude CodeAgentforce VibesLWR

AI-Accelerated Salesforce Migrations

Convert legacy stacks at 3–5× human pace with AI-driven code translation: AngularJS → LWC, Aura → LWC, Visualforce → LWC, OmniStudio → LWR. Code-quality guardrails, regression tests, and deployment plans included. Built on real migration experience at Telus and Assurant.

Claude CodeLWC / LWRSFDX

Agentforce Skills, Actions & Custom Agents

Build Agentforce skills and custom actions that integrate with your Salesforce data, Apex services, and back-office systems via MuleSoft. Agents that actually do work — not chatbots that summarize. Backed by deep OmniStudio and CPQ context.

AgentforceApexMuleSoftFlow

AI Strategy & Pilot Engagements

For enterprises new to agentic AI: identify the highest-ROI use cases in your Salesforce ecosystem, run a controlled pilot, and prove value before scaling. No transformation slides — just a working agent in production by the end of the engagement.

StrategyPilotROI

A four-step engagement.

Designed for senior buyers — short discovery, fast prototype, governed rollout.

Discovery

Free 30-minute call. Understand the goal, the constraint, the timeline. No deck.

Prototype

One- to two-week vibe-driven POC. You see a working agent, not a slide pitch.

Production

Iterative delivery with weekly demos. Human approval gates, code reviewed, instrumented.

Handover

Documentation, training, runbooks, governance model, and optional ongoing support.

Tools I actually ship with.

No vendor pitches. These are the tools running in production today.

Salesforce-native AI

  • Agentforce
  • Agentforce Vibes
  • Agent Script
  • Salesforce Hosted MCP Servers
  • External Client Apps (ECA)
  • Named Credentials · OIDC / SSO

Foundation models & tooling

  • Claude (Sonnet, Opus)
  • Claude Code
  • MCP Tooling
  • LLM Orchestration
  • Azure DevOps Integration

Application & delivery

  • LWC / LWR
  • Apex · Salesforce Flows
  • MuleSoft
  • JavaScript / Next.js
  • Copado / Gearset

Got a Salesforce + Agentic AI problem in mind?

Tell me what you're trying to do. I'll tell you honestly whether agentic AI is the right answer — and if so, how I'd ship it.

krishna.gonnabathula@noveltyss.com →