Internal tools & automation

Internal tools and process automation that give your team hours back

Copying data between systems, chasing approvals, building the same report every Monday. We replace that work with internal tools and automations that run reliably in the background, with a person in control wherever judgement is needed.

  • Admin panels and back offices built around your process
  • Scheduled jobs with a dashboard that shows what ran and what failed
  • AI for the tedious parts, people for the decisions

How an automation runs

  1. Trigger

    A schedule, an incoming email, a webhook or a button in the admin.

  2. Gather

    Pull the data from the systems involved, with every source recorded.

  3. Process

    Apply your rules, or let AI classify, summarise or draft where rules fall short.

  4. Check

    A person approves anything that sends, spends or can't be undone.

  5. Log and report

    Every run is logged, failures alert someone, and results land where the team works.

The pattern behind the automations in our own products.
What we can build

What we build

Tools for the work behind the scenes, built around how your team actually works.

Admin panels and back offices

Search, edit and manage your data with screens designed for your team, instead of a generic database tool.

Approval flows

Requests, reviews and sign-offs with notifications and a clear audit trail of who approved what, and when.

Scheduled jobs

Syncs, reminders, clean-ups and reports that run on time, with a dashboard that shows every run and every failure.

Email and document handling

Incoming email matched and sorted, lifecycle emails triggered by what users do, and documents assembled from your own data.

AI-assisted work

Research, classification, summaries and drafts done by AI and checked by a person before anything goes out.

Reporting

Dashboards and scheduled reports that pull from all your systems, so nobody has to build spreadsheets by hand.

Proof, not promises

Automation that runs our own businesses

Our products run a long list of scheduled jobs: syncing inboxes and marketplace orders, refreshing tokens, sending lifecycle emails, sweeping abandoned checkouts and checking backups. Admin back offices cover billing, users, AI jobs and sales, and a job dashboard shows every background task and its status.

Even the dull admin is automated. Postcrest syncs its Stripe invoices, payouts and fees to Moneybird from a bookkeeping dashboard, and prepares the evidence for payment disputes so a person can review it before it's sent.

  • A job dashboard showing every background task, its status and its failures
  • Lifecycle emails triggered by what users do, with one-click unsubscribe
  • Abandoned checkouts released automatically, so stock never stays reserved
  • Stuck AI jobs detected, failed and refunded without anyone stepping in
  • A weekly automated check of the database backups

A few of the jobs running in our products

  1. Every 15 seconds: publishing queue

    Scheduled social posts picked up and published on time.

  2. Every 5 minutes: inbox sync

    Mailboxes mirrored, replies matched and classified.

  3. Every 10 minutes: orders and checkouts

    Marketplace orders imported, abandoned checkouts released.

  4. Every 15 minutes: token refresh

    OAuth tokens renewed before they expire.

  5. Weekly: backup check

    An automated check of the database backups.

Scheduled jobs from Postcrest, Twelve Lumens and Salestreak.
between publishing queue runs
15 s
between inbox syncs
5 min
background job types in one product
28
of outgoing sales emails approved by a person
100%
Who it's for

Who we automate for

Operations teams

Planning, scheduling and order handling that now depend on spreadsheets and people remembering things.

Customer service teams

Incoming requests sorted, enriched and routed, with drafts ready for a person to send.

Finance and administration

Invoices, payments and bookkeeping synced automatically, with exceptions flagged instead of hunted down.

Sales teams

Research, follow-ups and CRM updates handled in the background, with a person deciding what goes out.

How it works

How we automate a process

Start with the task that costs the most time, prove it works, then do the next one.

  1. 1

    Find the hours

    We map where your team spends time on repetitive work and which tasks are worth automating first.

  2. 2

    Decide what stays human

    Decisions, exceptions and anything customers see get a review step.

  3. 3

    Build the smallest useful tool

    Start with the one task that saves the most time, live early and improved together with your team.

  4. 4

    Connect your systems

    Integrations with the tools you already use, so nobody types the same data twice.

  5. 5

    Log everything

    Every run is visible: what happened, when and why, with an alert when something fails.

  6. 6

    Expand

    Automate the next task once the first one has proven itself.

Tooling we use

Internal tools run on the same stack as our products: admin screens in Rails, background jobs with a dashboard, and AI only where it beats a simple rule.

  • Ruby on Rails
  • Hotwire
  • PostgreSQL
  • Solid Queue
  • Mission Control
  • Gmail API
  • Resend
  • OpenAI
  • MCP
  • Sentry
Cost & timeline

What does an internal tool or automation cost?

Automating one well-defined task is a small project. These factors decide the estimate:

Number of processes
One task, or a chain of steps across several teams.
Systems involved
Each system you read from or write to needs an integration.
Exceptions
Processes with many special cases need more rules, more testing and more review steps.
AI or rules
Simple rules are cheap and predictable. AI handles messy input but adds running costs and a review step.
Users and permissions
Who can see, change and approve what, and how that's recorded.
FAQ

Questions about internal tools and process automation

What does business process automation cost?

It depends on the number of processes, the systems involved and how many exceptions there are. After a free intro call you get a written scope and estimate, usually starting with the task that costs your team the most time.

How long does it take to automate a process?

That depends on the process and the systems involved. We start with one task, get it live early and expand from there. The estimate after the intro call includes a timeline per phase.

Which processes are worth automating?

Repetitive, rule-based work that happens often: copying data between systems, sending reminders, building reports, sorting incoming requests. Rare work that needs a lot of judgement is usually better left to people.

Should we use AI for this?

Only where it beats a simple rule: reading messy email, classifying requests, summarising or drafting. For everything predictable, plain code is cheaper and more reliable.

Why not use Zapier or Make?

For small, simple flows they're fine. When a process is business-critical, high-volume or full of exceptions, a built tool with tests, logs and proper error handling is more reliable, and often cheaper over time.

Will our team stay in control?

Yes. We add review steps for anything that sends, spends or can't be undone, and every run is logged, so you can see exactly what happened.

Where does your team lose the most time?

Tell us about the repetitive work in your week. You'll get an honest view on what's worth automating and what it would take.

martin@balk.dev

Typically a reply within one business day.

Useful to mention

  • What you want to build, in a few sentences
  • Who is going to use it
  • Any deadline or launch date
  • Tools and systems it should connect to