AI agent for office and back-office automation
An agent that takes over routine office paperwork between your systems: it collects requests, routes them for approval, drafts letters and contracts from your templates, files documents and sends managers a daily summary of what is open.
- 9 new requests collected from email and formsDone
- 6 requests sent to managers for approvalDone
- 3 letters drafted from templatesAwaiting approval
- A payment request is waiting 4 days for approvalAsks a person
An example. Your agent is built around your own work.
How it works
-
Take the request
Leave, purchase, payment and document requests arrive by email, a form or a chat message. The agent reads each one, finds what is missing and asks the sender for it.
-
Check and route
It checks the request against your rules (budget limits, leave balance, the right approver) and sends it to the person who must approve, with the details they need in one message. Late approvals are chased.
-
Prepare the paperwork
Once approved, it fills your letter, contract or proposal template, attaches the sources it used, and saves the file in the right folder. Anything that goes outside the company waits for a person to approve or sign.
-
Report
Managers get a daily or weekly summary of what is open, approved and overdue. Each action the agent took is logged with who approved it.
New to AI agents? How they work, in plain words
Real examples, setup steps and tool connections for this agent
An example conversation
An example of how a conversation runs. The grey lines show what the agent does in your systems meanwhile.
- CustomerHi, I need a purchase approved: 4 monitors for the design team, about 1,100 in total. Quote attached.
- Read the quote PDF from Sana's email: supplier Lumen IT, 4 x 27-inch monitors, total 1,096. Checked the supplier list: Lumen IT is an approved supplier. Checked the design team's budget in the shared sheet: 2,400 left this quarter.
- AgentThanks, Sana. The supplier is approved and the budget covers it. Purchases over 1,000 need the finance manager. I have sent her the request with the quote, and I will let you know when she answers.
- Sent an approval request to the finance manager on Teams with the quote, the amount and the remaining budget. Set a reminder for 4 pm.
- CustomerAlso, can you prepare the service agreement for the new client, Brightwell? Same terms as the Ardent one.
- Opened the template and the Ardent agreement. Filled in the Brightwell name, address and start date from the CRM. Kept the Ardent payment terms (30 days). The liability clause differs from the template.
- AgentThe Brightwell draft is in the Contracts folder. It is based on the Ardent terms, with one difference you should look at: the liability clause is not the same as in our standard template. I have not sent it anywhere. Who should review it?
- CustomerSend it to Omar for review, then we sign.
- Shared the draft with Omar in Google Drive with edit access and a comment pointing to the clause. Added a task to the legal checklist.
- AgentSent to Omar. The monitors are approved by finance, so I have created the purchase order and emailed it to Lumen IT.
How it is built
- Start with one task you can measure
- We pick one routine job with a clear input and output, such as requests and approvals or proposals from a template, and note how long it takes today. That figure is how you tell later whether the agent really saves time.
- Inputs from the systems your people already use
- The agent reads email, forms, shared drives, calendars and your ERP or HR system through their official connections (Microsoft 365, Google Workspace, and the APIs of your other software). Read and write permissions are kept separate, and each folder or system is opened only if the job needs it.
- The language model drafts, rules decide
- A language model reads messy requests and drafts letters and summaries from your templates and documents. Rules decide the things that must be exact: who approves what, the amount limits, required fields, which client a document belongs to.
- Approvals are built into the steps
- The agent sends the request to the right person on Teams, Slack or email and waits for the answer. It writes, sends or files only after that answer. Documents for signature go through your e-signature tool, and the agent never signs.
- Sources are attached to every draft
- Each fact in a draft, such as a price, a date or a clause, comes with the document it was taken from, so a reviewer can check it in seconds.
- Everything is logged
- Each action is recorded with the person, the time and the documents used. This makes the audit trail, and it makes the monthly hours-saved comparison honest.
- Where it runs
- In the cloud or on your own servers. Staff keep using their normal tools. The agent's messages and drafts appear in them.
What you can connect to it
- Microsoft 365 (Outlook, SharePoint, OneDrive, Teams)
- Reads requests from mail, files documents in SharePoint, and sends approval requests and reminders on Teams.
- Google Workspace (Gmail, Drive, Docs, Calendar)
- Reads mail, drafts documents from templates in Docs, files them in Drive and books meetings in Calendar.
- Slack
- Sends approval requests with the details attached and records the answer.
- DocuSign or another e-signature tool
- Sends a finished document for signature to the named signers. The agent does not sign.
- ERP and accounting (Odoo, SAP, QuickBooks, Xero)
- Creates purchase orders and invoices from approved requests, and reads budgets and supplier lists.
- CRM (Salesforce, HubSpot and others)
- Takes client names, addresses and deal details to fill proposals and agreements.
- HR or timesheet system
- Reads leave balances and timesheet hours for requests and for invoices based on time.
- Shared drives and document templates
- Holds your letters, contracts and checklists the agent starts from.
- Google Sheets or Excel
- Holds budgets, approval limits and simple request logs where there is no other system.
Connects to your tools. You stay in control.
- Works with what you have
- Microsoft 365, Google Workspace, Teams, Slack, e-signature tools such as DocuSign, your ERP or accounting software, CRM, HR system, shared drives, document templates, spreadsheets.
- A person approves what matters
- Payments, orders, discounts or anything you choose wait for approval. Every action is logged.
- Can run on your own servers
- The agent, its data, its logs and even the language model (open models such as Llama, Qwen or Mistral) can run on your servers, so documents and customer data stay inside your network. How that works
What it does not do
- It does not sign contracts or approve spending. People approve and sign; the agent prepares and routes.
- It does not send documents to clients or suppliers without an approval step you have set.
- It does not read your whole drive. It gets only the folders and systems the job needs, and you can review that access.
- It does not decide matters that need judgement, such as disputes or exceptions to policy. Those go to a person with a summary.
- It does not promise hours saved in advance. You measure that on your own process after a few months.
Questions
Do we need a new office automation system?
No. The agent works across the email, folders and software you already use. Requests and approvals arrive in the tools your staff already use.
Who signs letters and approves requests?
Your managers. The agent prepares and routes; people approve and sign. Anything that goes to a client or changes money waits for a named person.
How is this different from n8n or Power Automate?
Those tools run fixed flows that someone designs step by step. An agent also reads messy input, like emails and scans, and decides within your rules. Where a fixed flow is enough, we build that instead. If you already have flows in Power Automate, the agent can work alongside them.
What office tasks can an AI agent actually take over?
Repetitive ones with a clear input and output: collecting and routing requests, drafting letters and contracts from templates, filing documents, preparing invoices from timesheets and sending status summaries. We start with the one that costs your team the most hours.
How do we know it is saving time?
We note how long the task takes before the agent starts and compare after a few months, counting review time too. A process that looks faster but needs more checking is not a gain, so the whole cycle is measured.
Is it safe for the agent to see our documents?
It sees only the folders and systems you give it, with read and write rights kept separate. It can run on your own servers, and every access is logged. The permissions can be reviewed whenever you like.
What if a request is unusual?
The agent sends it to a person with a short summary and the documents attached. Exceptions are expected, and handling them is the staff's job; the routine path is the agent's.
How much does an agent like this cost?
It depends on the task and the systems it connects to. After a first call we give you a clear estimate of cost and time.
Want this agent built around your work?
Tell us in a few lines what the job looks like today and we will set up an online intro call.
Request sent
We will contact you to set up an intro call.