Senior AI Agent Developer – Claude AI / CRM Automation

Position: Senior AI Agent Developer / AI Automation Engineer Focus: Claude AI Agents, Browser Automation, CRM Automation & Automotive Sourcing Engagement: Contract / Project-Based with Potential for Ongoing Work Experience Level: Senior / Expert About the Project We are looking for a highly experienced AI developer to build and deploy Claude-powered AI agents that can perform real operational tasks inside our existing systems. This is not a chatbot project. Our goal is to create AI agents capable of completing repetitive tasks currently performed by human employees, particularly around vehicle sourcing, data entry, vehicle valuation, CRM management, and initial seller outreach. We already have established workflows and internal tools. We need someone who can translate those workflows into reliable AI agents. Initial Use Case: Private-Party Vehicle Sourcing Agent One of the first agents will function as a vehicle sourcing/farming agent. The agent should be capable of performing a workflow similar to the following: 1. Search for Vehicles Search approved private-party vehicle listing sources for vehicles matching predetermined buying criteria. Examples may include private-party marketplaces, classified sites, and other approved vehicle listing sources. The agent should be able to identify potentially attractive vehicles while avoiding obvious duplicates or previously reviewed listings. 2. Capture Vehicle Information Extract relevant information from the listing, including when available: Year Make Model Trim Mileage VIN Asking price Seller name Phone number Email Location Listing URL Vehicle description Photos Other relevant listing information 3. Create or Update the Vehicle in Our CRM The agent should interact with our CRM and: Search for an existing vehicle or seller Avoid creating duplicate records Create a new record when appropriate Populate the correct fields Attach or reference the original listing Add relevant notes and sourcing information Update the appropriate status or workflow 4. Book Out / Value the Vehicle The agent should then use our existing vehicle booking/valuation tools to determine relevant vehicle values. This may require: Entering VIN or vehicle information Selecting the correct trim/options Entering mileage Retrieving valuation information Returning that information to the CRM Comparing asking price against predetermined acquisition criteria The objective is for the agent to perform as much of the initial valuation process as possible without requiring human intervention. 5. Initial Seller Outreach If a phone number or other approved contact method is available, the agent may initiate the first outreach using predetermined messaging and business rules. The agent should be capable of: Sending an approved initial message Recording the outreach in the CRM Monitoring for a response where appropriate Updating the lead status Escalating the opportunity to a human buyer when required We are not looking to have AI independently negotiate or make unrestricted purchasing decisions initially. There should be clear guardrails determining when the AI can act and when a human needs to take over. 6. Human Handoff If the agent cannot contact the seller—for example, if the listing does not contain a phone number—the agent should automatically create a task or notification for a team member. Example: "Qualified vehicle identified. No direct phone number available. Manual marketplace outreach required." The employee can then open the listing, contact the seller manually, and continue the process. What We Want to Build This initial sourcing agent is only one use case. Longer term, we want to create multiple specialized AI agents capable of handling operational workflows across our company. Potential agents could include: Vehicle sourcing agents CRM data-entry agents Vehicle valuation agents Lead validation agents Seller outreach agents Follow-up agents Document collection agents Data cleanup agents Management/reporting agents Quality-control agents The ultimate objective is to develop an AI-powered operational workforce where agents handle repetitive tasks and employees focus on negotiation, decision-making, exceptions, and closing transactions. Technical Requirements We are specifically looking for someone with significant experience building Claude-based agents and agentic workflows. Strong candidates should have experience with several of the following: Claude / Anthropic APIs Claude Agent SDK or equivalent agent frameworks Tool use / function calling MCP (Model Context Protocol) Browser automation Computer-use agents Playwright or similar browser automation frameworks API integrations CRM integrations Webhooks Structured data extraction Authentication/session management Multi-step agent workflows Agent memory/state management Human-in-the-loop workflows Error handling and retries Logging and monitoring Permission and security controls LLM evaluation/testing Production deployment of AI agents Experience integrating agents with CRMs and third-party SaaS applications is especially valuable. Automotive industry experience is a major plus but is not required. Reliability Is Extremely Important We are not interested in impressive demos that only work 70% of the time. These agents will interact with real business systems and eventually perform tasks currently handled by employees. The developer must understand how to build systems with: Defined permissions Guardrails Validation before actions Duplicate prevention Error recovery Human escalation Detailed activity logs Monitoring Testing Cost controls Secure credential management We need to be able to understand what an agent did, why it did it, and where a workflow failed. Ideal Candidate The ideal person has already built AI agents that interact with real software applications and perform multi-step business processes. We would especially like to speak with someone who can show examples such as: "I built an agent that logs into a system, finds information, evaluates it, enters information into another system, communicates with a customer, updates the CRM, and escalates exceptions to an employee." That type of experience is significantly more relevant to us than traditional chatbot development. What We Would Like to See When applying, please provide: Examples of Claude AI agents you have built Description of what those agents actually did Whether they operated through APIs, browser automation, computer use, or a combination Experience with Claude tool use / MCP Examples of CRM automation Experience with Playwright or similar browser automation How you handle failures and retries How you prevent agents from taking incorrect actions How you implement human approval or escalation Links, screenshots, demos, GitHub repositories, or architecture examples when available Please explain what you personally built rather than simply listing AI technologies you have worked with. Initial Project The first milestone will likely be: Build a working Claude AI vehicle sourcing agent that can identify a private-party vehicle, collect the relevant information, enter it into our CRM, retrieve vehicle valuation information, determine the appropriate next action, initiate approved outreach when possible, and otherwise assign the opportunity to a human team member. Once this workflow is stable, we expect to expand the system into additional agents and business processes. We are looking for someone who can help us build the underlying architecture correctly so that we can continue adding AI agents over time.