Skip to content
Discuss your project

Automate a manual process without breaking the parts that work.

The process stops depending on one person at their desk. Routine steps move records, check conditions, and send notifications across systems, while approvals, exceptions, and unusual cases land with a named owner.

You know this is you when

  • The same steps are done in the same order every day, by hand.
  • Data is copied from one system to another by a person.
  • A checklist or a standard operating procedure exists, and people follow it from memory.
  • The process stops when one particular person is away.
  • Errors come from the routine steps, not the difficult ones.

Why it is happening now

Manual processes scale by adding people. At some volume that stops working, or the errors start reaching customers, or the person who holds the process leaves. The routine steps in most processes can be handled by software: moving records between systems, checking conditions, routing, notifying, drafting. The judgment steps should stay with people, with the software putting the right information in front of them.

Documents and requests pass through a model; confident results post to the system of record and are sampled, uncertain ones go to a review queue where a person decides.Documents, requestsModel stepextract · classify · draftSystem of recordReview queueconfidentuncertainperson decidessampled for accuracy
  1. Documents, requests
  2. Model stepextract · classify · draft
  3. confidentSystem of recordsampled for accuracy
    uncertainReview queueperson decides
Nothing posts without a threshold, a review path, and a sample. The exception queue has a reason and an owner.

Your options

The choices, stated plainly.

Building is not always the answer. These are the options we would put on the table, including the ones that do not involve us.

01

Configure automation in the tools you have

Choose it when

The process lives inside one product with a workflow engine, and the rules are simple.

What it costs

Limited to what that product can see. Cross-system steps still need a person.

02

A no-code automation platform

Choose it when

A few simple flows between well-known apps, with someone willing to own the flows.

What it costs

Fragile at volume, hard to test, and silent when a step fails. Fine for small flows.

03

A custom workflow with review and exceptions

Choose it when

The process crosses systems, has rules that matter, or needs a record of what happened.

What it costs

A scoped project. The routine steps run themselves; exceptions land with a person, with the reason.

What we would do first

The first steps, in order.

  1. 01

    Map the process as it actually runs

    We walk through it with the people who do it and record every step, every system, and every place a person copies data or makes a decision.

  2. 02

    Separate routine from judgment

    Routine steps are automated. Judgment steps stay with a person, with the software preparing the decision.

  3. 03

    Build the first flow with exceptions visible

    We build one process end to end, tested against real records, with a queue for anything the automation cannot handle.

  4. 04

    Measure and extend

    The process owner sees volumes, exceptions, and time saved. Then we move on to the next process.

What it looks like after

The routine steps run without anyone touching them. The person who used to do them reviews exceptions, handles the judgment calls, and sees the volumes. Nothing fails silently, and the process no longer depends on one person being at their desk.

Dental laboratory

Practices submit cases online, and the coordinator handles only the incomplete ones.

Dental laboratories typically receive cases by courier slip, email, and phone, and a coordinator types each one into the lab system, assigns a technician, and calls practices to confirm due dates. An automated intake flow gives practices an online form, creates the case in the lab system with its due date calculated, routes the work by case type, and notifies the practice at each stage. Cases missing a prescription detail land in a queue for the coordinator. Remakes and invoicing are sensible processes to automate next.

The path from here

What we would build, and for whom.

Questions

Before you decide.

What companies in this situation ask us first.

Which steps should stay manual?

The ones that need judgment: approvals above a threshold, unusual cases, anything with a customer relationship at stake. The software prepares those decisions rather than making them.

Where does AI fit?

In the reading and drafting steps: extracting fields from documents, classifying requests, drafting responses for a person to approve. Each step has human review and an exception path.

What happens when a step fails?

It lands in an exception queue with the reason and an owner, and it can be retried once fixed. Nothing fails silently.

Can we start small?

That is the only way we start: one process, end to end, measured, then the next.

Describe the process.

Tell us the situation and the systems around it.
We’ll come back with what we would do first, and what it would take.