Back to Blog

How to Set Up an AI Approval Process: Step by Step

Henri Jung, Co-founder at Superkind
Henri Jung

Co-founder at Superkind

Superkind logo on an abstract black silk texture

An AI approval process decides which tasks your AI employee handles on its own and which ones pause for a person to say yes. Friday afternoon, 16:20. A payment reminder is ready, a quote needs to go out, and three invoices are waiting to be booked. One person in the team checks every single one before it leaves - even the ones that never change. The bottleneck is not the work itself; it is the sign-off. This guide shows you how to set up an approval process where the right tasks wait for your OK and everything else runs, step by step, with Superkind’s ready-made AI team for approvals.

Key takeaways

You decide what needs approval - set rules per task type, amount or recipient. Sensitive steps wait for your OK; routine work runs on its own. One rule change is one sentence to Superkind.

Get a ready-made AI team for approvals - Superkind comes with an approval team (Invoice Clerk, Quote Guard, Onboarding Pilot) that processes, checks and routes work automatically. Start for free with unlimited users, or build your own.

Keep control, keep data in Europe - every approval, rejection and automated action is logged. Superkind is hosted in Europe, GDPR compliant, never trains on your data, and you set who sees what.

Explain it once, it keeps running - a rule like “Quotes above EUR 10,000 need management sign-off” applies to every future quote until you change it.

Your AI team for approvals

Invoice #4471 from Meier Elektrotechnik
Meier Elektrotechnik GmbH
Please find attached invoice #4471 for the November delivery.
Invoice ClerkAI employee
New invoice in inbox
Invoice matched to PO-2289. Amount EUR 2,340 net. Booked in lexoffice, cost centre 410.
AmountEUR 2,340 net
Matched toPO-2289
Cost centre410 - Operations
Runs automatically, no approval needed
Example: Invoice Clerk at work in Outlook

What is an AI approval process?

An AI approval process is a workflow where an AI employee completes routine tasks automatically and pauses at defined checkpoints for a person to approve, reject or change the result before it takes effect. The AI does the preparation, routing and logging; the person makes the call on the steps that carry risk. One example is Superkind: an AI employee that books invoices, drafts quotes and runs onboarding steps in your tools - with approval wherever you set it, hosted in Europe.

The difference to a manual sign-off chain: the AI does not just forward a request. It reads the data, applies your rules (amount thresholds, recipient lists, margin floors), prepares a concrete proposal with full context, and routes it to the right approver. What comes back is not a vague request but a decision you can make in one click.

How it works with Superkind: your AI team for approvals

You write to Superkind by email, in Teams, in Slack or in the web app like a colleague, and Superkind does the task in your programs - with 1,700+ integrations, including HubSpot, lexoffice, Personio and Outlook. Hosted in Europe, GDPR compliant.

With Superkind you don't start from zero: it comes with ready-made AI employees for your approval workflows. They run on their own in the background, and you decide where they ask you first. Use them right away, adapt them, or build your own.

Invoice Clerk

AI employee for invoice processing

Invoices pile up in the inbox, and each one needs matching, booking and sometimes a second pair of eyes. The Invoice Clerk picks up every new invoice from Outlook, matches it to the order in your system, books the standard ones in lexoffice on its own, and sends anything above your threshold to you for approval before it posts.

Works inOutlookExcel+ 1,700 more
Invoice #4471 from Meier Elektrotechnik
Meier Elektrotechnik GmbH
Please find attached invoice #4471 for the November delivery.
Invoice ClerkAI employee
New invoice in inbox
Invoice matched to PO-2289. Amount EUR 2,340 net. Booked in lexoffice, cost centre 410.
AmountEUR 2,340 net
Matched toPO-2289
Cost centre410 - Operations
Runs automatically, no approval needed
Example: Invoice Clerk at work in Outlook

Quote Guard

AI employee for quote approvals

Your sales team sends quotes all day, and the ones above a certain value need sign-off before they leave. The Quote Guard checks every new quote against your pricing rules and margin floors, lets the clean ones go, and holds anything outside the rules until someone from management approves.

Works inMicrosoft TeamsHubSpot+ 1,700 more
QuotesSales
Quote GuardAI employee
New quote created in HubSpot
New quote for Baumann AG reviewed. 12 line items, total EUR 48,500 net. Margin 18 % - above your 15 % floor. Pricing rules passed.
CustomerBaumann AG
TotalEUR 48,500 net
Margin18 % (floor 15 %)
Management sign-offOver EUR 25,000, so it waits for your OK.ApproveReject
Example: Quote Guard at work in Microsoft Teams

Onboarding Pilot

AI employee for employee onboarding

A new colleague starts next Monday and twelve onboarding tasks need to happen: welcome email, laptop order, calendar invites, tool access, Slack channels. The Onboarding Pilot runs through the checklist on its own, but waits for IT to approve before granting system access or sending credentials.

Works inSlackGmail+ 1,700 more
#onboarding
Onboarding PilotAI employee
Routine - every new hire, 5 days before start
Onboarding for Lisa Berger (start date 25 Nov) is running. 9 of 12 steps done: welcome email sent, laptop ordered, calendar invites created, Slack channels joined.
Steps done9 of 12
WaitingIT access approval
Start date25 Nov
IT access for Lisa BergerSystem access waits for IT approval.ApproveReject
Example: Onboarding Pilot at work in Slack

Every employee is ready to use right away. Or build your own on the platform.

Safe by default

Approval decisions on invoices, quotes and employee data stay in Europe, logged and under your control.

What you need

Checklist before you start

  • A list of the tasks you want to hand over: invoices, quotes, onboarding, payment reminders, emails to customers
  • For each task: who approves today, what threshold or rule triggers the approval, and where the data sits (Outlook, HubSpot, lexoffice, Personio, Excel)
  • The channels your team uses: Outlook, Gmail, Microsoft Teams or Slack
  • An admin who can connect the systems and confirm the permissions
  • A free Superkind account at app.superkind.ai - the 1,700+ integrations connect to your tools without code

Step by step with Superkind

Step 1: List the tasks and sort them into “auto” and “approval”

Write down every repeating task your team does: invoice booking, quote creation, payment reminders, onboarding steps, email replies. For each one, ask: “If I saw this result, would I change anything?” If the answer is almost never, it runs on its own. If it sometimes needs a correction - or if the stakes are high (money leaving the account, data going to a customer, access being granted) - it gets an approval step.

  • Auto examples - matching invoices under EUR 500 to purchase orders, sorting incoming emails by topic, creating calendar entries for new hires
  • Approval examples - quotes above EUR 10,000, payment reminders to key accounts, granting system access to new employees, emails to external contacts with pricing

Step 2: Connect your systems

Sign up at app.superkind.ai and connect the programs your tasks live in: Outlook or Gmail for the inbox, HubSpot or Pipedrive for deals and quotes, lexoffice or sevdesk for invoices, Personio for HR, and Teams or Slack where approvals happen. An admin confirms which areas Superkind can read and write. No API keys, no code.

Step 3: Tell Superkind the rules

Write Superkind a message in Teams or Slack with your approval rules, in plain language. One sentence per rule.

TeamsDaniel to Superkind

Invoices under EUR 1,000: book in lexoffice on your own. Above EUR 1,000: show me the booking first and wait for my OK. Quotes above EUR 10,000: Nadine approves before it goes to the customer.

Superkind confirms each rule and applies it from the next task. You can add, change or remove rules at any time with another message.

Step 4: Run the first task and check the result

Take a real task - an invoice that just arrived, or a quote your sales team needs to send - and hand it to Superkind.

OutlookDaniel to Superkind

Book the attached invoice from Meier Elektrotechnik.

Superkind reads the invoice, matches it to the purchase order, checks your rules (amount EUR 2,340 - above EUR 1,000, so it waits), and sends you the proposed booking with the full context in Teams. You approve or reject with one click. If you approve, it books in lexoffice. If you reject, it asks what to change.

Step 5: Adjust the thresholds

After a week, check the log: which approvals did you wave through without changing anything? Those are candidates for auto. Which auto tasks had issues? Those need an approval step. Write the new rules to Superkind.

SlackNadine to Superkind

Raise the invoice threshold to EUR 2,000. Everything under that books without approval.

Gartner recommends this iterative approach: start with more approval, loosen as trust builds, and tighten again when new task types are added.1

Set it up as a routine

Once the rules are tested, the approval process keeps running on its own. New invoices are booked or held, new quotes are checked or sent, and new hires get their onboarding steps - all without someone starting the work each time.

  • Schedule example - “Every morning at 8:00, check the invoice inbox and process everything. Anything above EUR 2,000 waits for Daniel.”
  • Trigger example - “When a new quote is created in HubSpot, check the margin and amount. Above EUR 10,000, Nadine approves.”
  • What Superkind learns - every approved, rejected or changed task feeds back into the Company Brain. The rules get sharper, the proposals more accurate, and the knowledge stays with the company even when people change roles.

Data protection and EU hosting

Approval workflows handle sensitive data: invoice amounts, customer names, employee records, pricing. When you use Superkind, that data stays in Europe. The EU AI Act (Regulation (EU) 2024/1689) requires human oversight for high-risk AI systems4 - Superkind’s approval feature gives you exactly that, with every decision logged and traceable.

Other ways to handle approvals

MethodEffortRisk of errorsBest for
SuperkindLow - rules in plain languageLow - logged and checkedTeams that want approval without building workflows
Manual sign-offHigh - every task by handMedium - depends on the personVery small teams, rare tasks
Email chainsMedium - forwarding and waitingHigh - lost in the inboxAd-hoc approvals
Workflow toolsHigh - build and maintain flowsLow - if maintainedIT teams with time to build

Common mistakes

  • Approving everything - if every task waits for a person, the AI saves no time and the approver rubber-stamps after a week. Start with the tasks that actually need a human decision.2
  • Approving nothing - letting everything run unsupervised from day one builds no trust and catches no early mistakes. Start with more approval, then loosen.1
  • No clear owner - an approval step without a named person becomes a dead end. For every rule, name who approves and who covers when that person is out.
  • Ignoring the log - the audit trail shows which tasks were approved, rejected or changed. Review it monthly to find rules that are too tight or too loose.
  • Setting rules once and forgetting - the team, the thresholds and the systems change. Review the approval rules every quarter, or when a new task type is added.

6 tips for a working approval process

1. Start with one process, not ten

Pick the task that causes the most bottlenecks today - often invoice booking or quote sign-off - and get that running before adding more.

2. Make the approval concrete

The approver should see the full proposal (invoice, booking, quote draft) with all relevant data, not a vague “Please approve.” Concrete proposals lead to fast decisions; vague ones lead to delays.6

3. Set a time limit

If an approval sits for more than 24 hours, escalate it. Otherwise the queue backs up and the AI employee stops making progress.

4. Use the channel the approver already lives in

If Nadine works in Teams all day, send the approval there. If Daniel checks Outlook first thing in the morning, send it by email. Superkind works in all of them.

5. Check data protection before you connect

Make sure the approval tool is GDPR compliant, hosted in Europe, and offers a DPA. With Superkind this is covered. For other tools, check the vendor’s data processing location and whether they train on your data.

6. Document the rules where the team can find them

Every approval rule should be visible to the team, not hidden in one person’s head. In Superkind, every rule is stored in the Company Brain and can be read and changed by anyone with the right permissions.

Conclusion

An AI approval process is not about removing people from the loop. It is about removing the tasks that never needed a person in the first place, and making the ones that do faster and better documented. The fastest way to start: take the next invoice or quote that lands in your inbox, forward it to Superkind with your approval rule, and see the first result in minutes. If you want to understand how Superkind fits with sovereign, European AI or what AI tools work for teams in Europe, read those guides next.

Frequently Asked Questions

An AI approval process is a workflow where an AI employee handles routine tasks on its own and pauses at defined checkpoints for a person to approve, reject or change the result before it goes out. You set the rules: which actions run freely, which ones wait. The AI prepares, routes and logs; you make the call on the steps that matter.

You tell Superkind which steps need sign-off - for example, quotes above EUR 10,000 or emails to external contacts. Everything else runs automatically. When an approval is due, Superkind sends the proposal to the right person in Teams, Slack or Outlook with the full context. One click to approve, one to reject. The result, the decision and the timestamp are logged. With 1,700+ integrations - including Pipedrive, Personio and Outlook - Superkind connects to the systems where your approval data already lives.

Tasks with low risk and clear rules are good candidates: matching an invoice to a purchase order, sorting incoming emails, creating a calendar entry, posting a status update in a channel. The test is simple: if the same person would approve it every time without changing anything, it can run on its own.

It depends on the tool. Superkind is hosted and processed in Europe, is GDPR compliant, aligned with the EU AI Act (Regulation (EU) 2024/1689) and does not train on your data. A data processing agreement (DPA) is available through the Trust Center at trust.superkind.ai. Every approval, rejection and automated action is logged with a timestamp for your audit trail.

You start for free at app.superkind.ai with up to 150 hours of work, no credit card and no sales call. Paid plans start at EUR 50 per month for 50 hours of work for the whole company with unlimited users. No price per seat. As of November 2026.

Yes, at any time. Write Superkind a new rule in plain language - "From now on, payment reminders under EUR 500 go out without approval" - and it applies from the next task. No code, no workflow editor. The old rule, the change and the person who made it are logged in the audit trail.

Related Articles

Henri Jung, Co-founder at Superkind
Henri Jung

Co-founder of Superkind, the AI employee teams write to in Outlook, Teams and Slack. Henri helps companies, from small teams to the Mittelstand, hand their routine work to AI - without new tools, consultants or an in-house AI team.

Send Superkind your first approval rule.

Start for free in Outlook, Teams or Slack. Or let us show you in 30 minutes how it runs in your systems.