PitchAI Agent Engine Try it out

Automate work with agents. Process by process.

For example

  • Read invoice emails each night and enter the details in our banking system.
  • Update our CRM from customer emails.
  • File incoming documents in the right project folders.
  • Create follow-up tasks for open quotes.
  • Compile a weekly project update from our mail and systems.

Read inputs from any system and write outputs to any system.

How does it work?

Describe your task once. Set the schedule. Watch it run automatically.

Example run

  1. 01Describe your task

    Read my emails, find the invoices, and enter all invoice details into our accounting system. Where possible, match each invoice to its bank statement.

    Every night · Runs automatically

  2. 02The agent gets to work.

    • Read 400 emails
    • Read the PDF attachments
    • Identified and organized 300 invoices
    • Entered all invoice details in the accounting system
    • Matched available bank-statement entries
    • Investigated an invoice amount that differs from its email
    • Reported the difference and highlighted two new suppliers

  3. 03See the finished result

    Agent Engine✓ Completed

    I read 400 emails and found 300 invoices. All 300 have been entered into your accounting system.

    298 were recurring invoices from known suppliers. I also matched invoices to the available bank-statement entries.

    Two are from new suppliers and are highlighted for you to check. One of these has an amount that differs from its email; I recorded the invoice amount and flagged the difference.

What else can you hand over?

Finance & documents

  • Match invoices to bank transactions.
  • Find the receipts missing from our expense report.
  • File supplier invoices by month.
  • List unpaid invoices with their payment status.
  • Put invoice amounts and VAT in a spreadsheet.
  • Check incoming invoices for duplicates.
  • File signed contracts with the right customer.
  • Compare supplier quotes in one table.

Customers & projects

  • Add new customer enquiries to our CRM.
  • Update deal stages from customer emails.
  • Turn meeting notes into tasks with owners.
  • Draft a follow-up email for each open quote.
  • Add agreed deadlines to the project plan.
  • Collect the documents for our next customer meeting.
  • Summarize this customer’s emails before our call.
  • Update the project record when a document arrives.

Planning & operations

  • Put confirmed appointments from emails into our calendar.
  • Remind me before our contracts expire.
  • List this week’s completed project tasks.
  • Find tenders that match our services.
  • List the documents required for this tender.
  • Update our price list from supplier documents.
  • Record delivery dates from order confirmations.
  • List the missing documents in this project file.

How it works 01

Give the task. See it done.

Describe the work in ordinary office language. The completed result tells you exactly what changed.

01
You ask

Invoices, every morning

Every morning, read the finance mailbox, inventory every invoice, add each one to Exact Online, and record its invoice date, due date, and status.

Done

Exact Online is up to date

I found 14 invoices. I added 13 to Exact Online and recorded the invoice date, due date, and status for each one. 1 duplicate is ready for review.

02
You ask

Meeting actions, assigned

After every team meeting, read the notes, add every agreed action to Microsoft Planner, assign its owner, and put each deadline in Outlook.

Done

The follow-up is in place

I added 7 actions to Microsoft Planner, assigned 4 owners, and placed 5 deadlines in Outlook. 2 actions without a date are ready for review.

03
You ask

Signed quotes, reflected in CRM

Every Friday, check the sales mailbox for signed quotes, update the matching HubSpot deals, attach each quote, and schedule the next follow-up.

Done

The sales timeline is current

I updated 9 HubSpot deals, attached 6 signed quotes, and scheduled 8 follow-ups. 1 unmatched quote is ready for review.

It is as simple as that.

Model-agnostic

Works with every AI and every agent.

The Agent Engine is not tied to one AI company or one agent. For every task it picks the model that suits it best: fast and economical for administration, strong at code for engineering, creative for writing. When a better model arrives tomorrow, you simply put it to work.

AI models

  • OpenAI

    GPT · Codex

  • Anthropic

    Claude · Claude Code

  • Google

    Gemini · Gemini CLI

  • DeepSeek

    DeepSeek-V3 · R1

  • Mistral AI

    Mistral · Mistral Code

  • Alibaba

    Qwen · Qwen Code

  • Moonshot AI

    Kimi

  • Zhipu AI

    GLM

  • xAI

    Grok

  • Meta

    Llama

Agents

  • Codex
  • Claude Code
  • Gemini CLI
  • OpenCode
  • GitHub Copilot
  • Cursor
  • Kiro
  • Cline

Your task definition stays the same, whichever model or agent does the work.

Processes 02

The office work that keeps returning.

Start where your team copies, checks, chases or rebuilds the same context every week.

Process 01 Good first process · instant with the Event Inbox (Silver)

Inbox follow-up

Check for the authoritative reply, update the working record, prepare the next action and stop the reminder when the loop is closed.

Source
Shared inbox and open follow-up
Work
Match the reply, update the record and prepare the next action
Returned
Updated case, reviewed follow-up and closed reminder
Checkpoint
Approval before an external reply

Returned with the workResult, source and completion evidence

Integrate with any of your systems

Contact us and share your ideas. Our developers will build your connector within a week.

  • Exact Online
  • AFAS
  • Twinfield
  • Moneybird
  • SnelStart
  • e-Boekhouden.nl
  • Yuki
  • Billit
  • WinBooks
  • Octopus
  • Horus
  • Teamleader
  • Odoo
  • Nmbrs
  • Loket.nl
  • SD Worx
  • Securex
  • Liantis
  • Acerta
  • ING
  • Rabobank
  • ABN AMRO
  • bunq
  • KBC
  • Belfius
  • BNP Paribas Fortis
  • Mollie
  • Adyen
  • Pay.
  • Visma eAccounting
  • Microsoft 365
  • Google Workspace
  • Salesforce
  • HubSpot
  • SAP
  • Oracle NetSuite
  • Sage Accounting
  • QuickBooks
  • Xero
  • Slack
  • Zoom
  • Dropbox
  • Box
  • Docusign
  • Shopify
  • WooCommerce
  • Trello
  • Asana
  • monday.com
  • Airtable
  • Jira
  • Confluence
  • GitHub
  • GitLab
  • Notion
  • Miro
  • Figma
  • Canva
  • Adobe Acrobat
  • ServiceNow
  • Workday
  • Zendesk
  • Freshdesk
  • Intercom
  • Mailchimp
  • ActiveCampaign
  • SendGrid
  • Twilio
  • Stripe
  • PayPal
  • Square
  • Amazon Web Services
  • Microsoft Azure
  • Google Cloud
  • Snowflake
  • Databricks
  • Tableau
  • Power BI
  • Smartsheet
  • Wrike
  • PostgreSQL
  • MySQL
  • MongoDB
  • Redis
  • Microsoft SQL Server
  • Oracle Database
  • MariaDB
  • Elasticsearch
  • SQLite
  • ClickHouse
  • DigitalOcean
  • Cloudflare
  • WordPress
  • Webflow
  • PrestaShop
  • Adobe Commerce (Magento)
  • Zoho CRM
  • Pipedrive
  • FreshBooks
  • Basecamp

Privacy

Building a privacy boundary for your agents.

We are building edge anonymisation to share less personal data with AI models. It is now in design and validation.

Edge anonymization in development · Agent Engine Silver and Gold

The planned boundary would replace identifying details with neutral references inside a PitchAI-controlled environment. Authorised results would return through that boundary to restore the details. The privacy page names the AI model providers.

Explore the privacy roadmap
  1. Example input

    Jan de VriesNL91 ABNA 0417 1643 00

  2. Planned privacy boundary

    in design and validation

  3. Planned model input

    Person_1IBAN_1

  • We don't train models on your data

    PitchAI does not use your data to train AI models.

  • Hosted in the EU

    PitchAI runs the agents and stores your data on its own servers in Germany and Finland. The AI model providers we use are named on the privacy page.

Pricing

Start small. Put more work behind you.

Invoices entered. Documents filed. Follow-ups created. Choose the capacity and support that fit your business. Prefer a managed package? See Managed AI Service.

Pay as you go

Try your first process, then pay for the work it completes.

Freeto get started

€5 free trial credit

Try it out
  • Use your trial credit to try it and get it working
  • No monthly subscription
  • No upfront setup fee
  • Pay only for successfully completed work

Self-employed

Keep recurring administration moving while you run your business.

€400per month

600 credits per month

Start with your admin
  • Monthly credits for your recurring work
  • Support when you need help
  • Advice on putting agents to work

Small & medium business

Bring your team’s processes together with a PitchAI consultant.

€2,000per month

4,000 credits per month

Plan your workflows
  • Help setting up workflows successfully
  • Consultant help with clear task descriptions
  • Advice on reducing costs and the connectors you need

Enterprise

Automate step by step around your company, systems and capacity needs.

Let’s talkcustom quote

Capacity tailored to your business

Discuss your operations
  • Reserved capacity
  • Dedicated direct support and consultants
  • Service-level agreement
  • Dedicated nodes, or a dedicated node you operate
  • Structured company and process intake to plan automation and lower total cost

Only successful work uses credits. Agree the completed result and its credit cost before the process runs.

Managed AI Service

Prefer that we run it for you?

One package of agent software, agent usage, advice and setup, for one fixed price per month: Bronze €500, Silver €2,000, Gold €5,500, usage included. Voice, the Event Inbox, edge anonymization, the SLA, the Wiki and Managers start at Silver.

Compare the packages
BronzeSilverGold
Price per monthincluding usage€500€2,000€5,500
Agent usage includedat cost price≈ 10 per working day (≈ 200 task runs a month)≈ 50 per working day (≈ 1,000 a month)≈ 200 per working day (≈ 4,000 a month)
Always onruns in the PitchAI Agent Engine Cloud✓✓✓
Reliablemonitoring, retries, updates✓✓✓
Ready-made connectorsGoogle, Microsoft, WhatsApp and hundreds more, set up by PitchAI✓✓✓
Custom connectorsbuilt by PitchAI for apps and websites in your business✗4 custommore, no fixed limit
Local connectorsecurely to your own computers and network✓✓✓
Setup, monitoring, advice and support✓✓✓
Training and enablement2 key peoplea full team, about 5 to 10 people30+ people
The best model, automaticallyor choose your own✓✓✓
Own cloud compute✓✓✓
Secure sign-in with SSOMicrosoft✓✓✓
WebsiteWork Inbox✓✓✓
iOS and Android app✓✓✓
Set up oncefixed times, schedules and triggers✓✓✓
Suggestionsagents find work other agents can take over✓✓✓
Wikia shared company wiki where agents read and record decisions, knowledge and ways of working✗✓✓
Managersmanager agents direct other agents and keep them working, for one-off projects✗✓✓
SLAagreements on availability✗✓ 99% uptime✓ 99.9% uptime
Guaranteed data storage in the EU✓✓✓
Dedicated nodes in the cloud✗✗✓
Priority support with a fixed response time✗✗✓
Fixed contact person✗✗✓
Quarterly strategy session✗✗✓
Audit logs for compliance✗✗✓
Custom version of the Agent Engineyour own version, adapted to your requirements and rolled out separately✗✗✓
Event Inboxtasks start on an event, such as a particular email✗✓✓
Voicecall one voice agent that directs all agents✗✓✓
Edge anonymizationagents work on personal data that stays on your own computer · in development, available once live✗✓✓
Advice and changes1 hour a month; 2 hours at setup4 hours a month; 4 hours at setup10 hours a month; 8 hours at setup
Fixed meeting✗every 2 weeksweekly, plus a quarterly board session
Response time1 working day4 working hours1 hour, 06:00–18:00

¹ Usage of the language model is charged at cost price, without mark-up: when tokens get cheaper, you get more work for the same money. The number of task runs varies with the complexity of a task and with how often the cache is hit (usually 95% or more). One task run is, for example, one morning planning or one round of WhatsApp messages. We count 20 working days a month.

In Bronze, tasks run at fixed moments, for example every 15 minutes. Picking up a message the moment it arrives is the Event Inbox in Silver. Prices per month, excluding VAT, usage included.

Start without risk: 5 hours, never more than €500
A small first decision
01

Zero initial cost

There is no setup fee for the standard first-process start.

02

One process

Agree a clear beginning, end and completion rule before work starts.

03

Pay for completed work

Review the result and its evidence. Only completed and validated work is billable.

Practical questions 06

Before the first process.

A useful process has a recognizable finish, a current owner and a clear line for human authority.

01Which process should we start with?

Choose a repeated process with a clear trigger, a recognizable finish and a current owner. Good first candidates involve checking, reconciling, preparing, updating or reminding.

02Do we need to replace our current software?

No. PitchAI connects the agreed process to the approved systems it needs. Your team keeps its working tools and records.

03Can PitchAI connect to our systems?

PitchAI supplies the connection needed for the agreed process. Some systems use an existing API. Others need a focused connector. We scope that route before the first run.

04Will the Engine send emails or submit documents by itself?

Only when that authority has been explicitly agreed for the process. Sending, submission, signature, purchase and sensitive changes can remain human checkpoints.

05What happens when information is incomplete?

The process records the missing fact or exception and routes it for review. An unresolved case stays open.

06What exactly gets validated?

The returned result is checked against the completion rule agreed before the run. That may be an updated record, a reconciled document set, a prepared follow-up, a closed reminder or another observable result.

07Can we stop after the first process?

Yes. The first process stands on its own. Keep it, change it, stop it or expand when the result justifies the next decision.

08What if we later need a broader custom integration?

That can be scoped separately after the process and its value are understood. The standard first-process start can begin without it.

Your first process 07

Start with the process your team keeps doing by hand.

Tell us what starts it, where the work happens and what a correct result looks like. We will define a practical first run with no setup fee.

Define your first process One process. One completion rule. A result you can review.