Back-office AI agent that handles requests, email and internal tasks
The agent takes over the flow of incoming email, requests and documents. It identifies the request type, extracts the data, enters it into your systems and replies or hands over to a person. You see every request and what was done with it.
What hurts
A shared inbox receives hundreds of messages: freight requests, insurance claims, tenant queries. Someone reads each one, routes it across systems and forwards it to colleagues. When that person is on holiday, the queue grows.
Data from emails and attachments is copied by hand into the CRM, accounting and spreadsheets. A wrong contract number or amount is found too late. Checking who handled a request and when is hard.
You could hire another administrator, but the flow grows with the business. You need custom AI development that handles routine requests on its own. And leaves the complex cases and decisions to people.
What we build
- Connection to email, forms, chat channels and portals where requests arrive
- Request classification by type, urgency and owner
- Data extraction from emails and attachments: PDFs, scans, spreadsheets
- Data entry into your CRM, SAP, Odoo, accounting or Google Sheets
- Template replies and email drafts for human review
- Step execution on Temporal with retries and deadline tracking
- An exception queue for people, with the reason the agent stopped
- A processing log for every request
What you get
- A working agent connected to your inbox and systems
- Agent and integration code in your repository
- A flow dashboard: request types, handling time, exceptions
- Rule documentation and a guide for the back-office team
- A test suite built on real emails and documents
How we measure the result
- Share of requests handled without a person
- Time from a request arriving to its entry in the system
- Number of errors in transferred data
- Size of the exception queue and its causes
Systems and stack
Industries
Questions
What requests can the agent handle?
Routine requests with a repeatable structure: freight orders, insurance claims, tenant queries, invoices, document requests. Unusual or disputed cases go to a person with a short summary. During discovery we review your flow and tell you what to automate first.
How much does a back-office agent cost?
From $6,000. The price depends on the number of request types, sources and target systems, and on document quality. The cheapest start is one request type from one mailbox. You get a fixed estimate 2–3 days after discovery.
What about poor-quality scans?
The agent reads scans and photos, but AI sometimes makes mistakes on unclear documents. So for amounts, bank details and contract numbers we add a check: a match against your database or confirmation by a person. Doubtful documents go to the exception queue, not into accounting.
Which systems does it work with?
Gmail, Outlook, corporate mail, HubSpot, Pipedrive, SAP, Odoo, Google Sheets and any system with an API. For systems without an API we set up file exchange. We connect a dedicated mailbox or a forwarding rule, never the personal mailboxes of your staff, and sign a GDPR DPA before access.
How long until launch?
Usually 4–8 weeks. First the agent handles one or two request types on part of the flow under human review. Then we add more types. After launch we provide SLA support and refine rules based on the exception queue.
Related services
More in this section
Describe the task. We reply within one business day, the first call is with the founder.
Describe the task. The first call is with the founder.
We reply within one business day. 30 minutes to tell you whether it is worth doing and what it costs.