Automate B2B Account Provisioning
Budget / SalaryHourly project
TypeFreelance project
LocationRemote
Posted2 hours ago
I want a Claude-powered workflow that watches every new ticket in our help-desk system, spots those asking for B2B tool log-ins, and acts without human touch. The flow is straightforward:
• Claude reads the ticket body, identifies the company or contact being referenced, then pings our separate customer database through its API to see whether a live B2B account already exists.
• If the account is present, Claude simply fetches the credentials (or a reset link) and emails them to the sales rep tied to that customer.
• When there is no clean match, Claude must go ahead and create the user inside the B2B tool, generate fresh credentials, and notify the same sales rep that a brand-new account has been issued.
Key points you’ll need to cover:
– Integrate with our non-Zendesk help-desk platform (REST and webhook access available).
– Query the dedicated customer database to confirm existing accounts.
– Connect to the B2B tool’s user-management API for account creation, password resets, and credential retrieval.
– Compose clear, templated emails (or internal ticket updates) to the sales rep listing the credentials.
– All logic has to run under Claude, so prompts, system instructions, and error handling need to be included in code and thoroughly documented.
Deliverables
1. Source code or serverless functions that tie Claude, the help-desk API, the customer database, and the B2B tool together.
2. A short configuration guide so I can point the workflow at my production endpoints.
3. Test logs showing each decision branch: existing account found, new account created, and the resulting sales-rep notification.
Acceptance criteria
• Tickets are classified correctly at ≥95 % accuracy.
• No duplicate accounts are created for the same customer.
• Sales reps receive credentials within two minutes of ticket arrival.
If you have prior experience wiring LLMs into customer-service pipelines, especially with Python or Node and RESTful APIs, your solution will fit right in. I’m ready to review your plan and set up access keys as soon as we agree on the approach.
• Claude reads the ticket body, identifies the company or contact being referenced, then pings our separate customer database through its API to see whether a live B2B account already exists.
• If the account is present, Claude simply fetches the credentials (or a reset link) and emails them to the sales rep tied to that customer.
• When there is no clean match, Claude must go ahead and create the user inside the B2B tool, generate fresh credentials, and notify the same sales rep that a brand-new account has been issued.
Key points you’ll need to cover:
– Integrate with our non-Zendesk help-desk platform (REST and webhook access available).
– Query the dedicated customer database to confirm existing accounts.
– Connect to the B2B tool’s user-management API for account creation, password resets, and credential retrieval.
– Compose clear, templated emails (or internal ticket updates) to the sales rep listing the credentials.
– All logic has to run under Claude, so prompts, system instructions, and error handling need to be included in code and thoroughly documented.
Deliverables
1. Source code or serverless functions that tie Claude, the help-desk API, the customer database, and the B2B tool together.
2. A short configuration guide so I can point the workflow at my production endpoints.
3. Test logs showing each decision branch: existing account found, new account created, and the resulting sales-rep notification.
Acceptance criteria
• Tickets are classified correctly at ≥95 % accuracy.
• No duplicate accounts are created for the same customer.
• Sales reps receive credentials within two minutes of ticket arrival.
If you have prior experience wiring LLMs into customer-service pipelines, especially with Python or Node and RESTful APIs, your solution will fit right in. I’m ready to review your plan and set up access keys as soon as we agree on the approach.
Apply on Freelancer →
Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.