Platform / Operations / Digital Employee
Operations · AI

Your best finance hire never sleeps.

A real AI agent — not a chatbot. It reads your invoices, pings your vendors, approves what matches your rules, and reconciles in your GL. Trained on the financial data of 7,000 companies.

24/7
Always on
8,400+
Actions / month / agent
97.2%
First-try correctness
ISA · AP SPECIALIST · 3:14 AMFound 14 invoices from last 4 hours. Running 3-way match now.
ISAMatched 13 of 14. One mismatch: AWS usage invoice is $12,840 but PO ceiling is $10,000. Flagging for @rachel in #ap-queue.
RACHEL · 9:02 AMApproved the AWS one, increase the cap to $15K.
⚙ tool_call: update_budget(gl="6410", ceiling=15000)
ISADone. 13 invoices released for payment, AWS paid on card (1.5% cashback = $192). Budget updated. You saved 38 min.
What it is

Not a chatbot. Not a macro.

An actual agent

Long-running, multi-step. Reads invoices. Drafts emails. Calls our API. Asks you when unsure. Learns from corrections.

Scoped by role

AP Specialist, Collections Analyst, Treasury Ops, Controller. Each with its own tools, rules, and limits.

With full audit

Every decision logged: prompt, tool call, result, diff. Replayable, reviewable, reversible.

Meet the team

Four specialists to start.

All four come off-the-shelf, scoped to your data, ready in minutes.

01 · AP
Isa — AP Specialist

OCRs invoices, matches to POs, routes approvals, schedules payments, reconciles to GL.

02 · AR
Marcus — Collections Analyst

Writes dunning emails in your tone, fields customer replies, escalates polite-to-firm, applies cash.

03 · Treasury
Nora — Treasury Ops

Forecasts cash daily, moves between accounts, buys FX, alerts on breach of min balance.

04 · Close
Avi — Controller

Runs the month-end close checklist. Reclasses. Flips journal entries. Signs off once reconciled.

05 · Tools
Full tool access

Every Payouts API is a tool. Plus your ERP, your Slack, your email, your calendar.

06 · Memory
Per-company memory

Learns vendor preferences, GL codes, approval patterns. Improves week-over-week.

07 · Limits
Hard spend limits

Per-agent, per-day, per-vendor caps. Cannot be overridden without human multi-sig.

08 · Review
Inbox of decisions

Every ambiguous call surfaced as a 20-second review. Approve, reject, correct, learn.

09 · Slack
Lives in Slack

@isa, @marcus — or DM. Attach an invoice, ask a question, reply with a fix.

10 · Browser
Browser-use skills

Logs into vendor portals to download invoices. Navigates state tax sites to file.

11 · Policy
Policy engine

Write plain-English rules ("wires over $50K need 4-eye"). Agent enforces automatically.

12 · Your own
Build your own

Finance, ops, or procurement. Point at your data, write the scope, give it tools.

Onboarding an agent

From signup to working in one afternoon.

1
Pick a role

Isa, Marcus, Nora, Avi — or a custom-scoped one.

2
Set rules

Approval thresholds, tone, limits. Rules in English, enforced in code.

3
Shadow mode

Agent drafts actions for 2 weeks, you approve. It learns you.

4
Autopilot

Flip the switch. Agent runs inside limits. Reports weekly.

Questions

Digital Employee FAQ.

What if the agent makes a mistake?
+
Every action is reversible. Payments below the 4-eye threshold can be clawed back via ACH reversal. Any action above threshold requires human co-signing anyway. You also see every decision in an audit log — you can replay any of them.
Does it use my data to train?
+
No. Your data stays in your tenant. Agents are fine-tuned per-tenant on per-tenant memory. We do not train foundation models on customer data.
Which model do you use?
+
We use Anthropic Claude as the primary planner with specialized smaller models for extraction and classification. Which specific model is an implementation detail — you get consistent behavior, we handle upgrades.
Can I build my own agent?
+
Yes. Define a role, give it tools from our MCP server, set policy. Full SDK + docs on the developer page.

The team you wish you could hire.
Starting at $99/mo.