Prototype in 10 business days: from manual to working
From manual process to working prototype in 10 business days, on your own systems. Day 11 is your decision.
Automating a manual process does not have to take months. In 10 business days, we turn one manual process into a working prototype on your own systems. It runs on real data, a person approves, and everything is logged. On day 10, it is live for a defined group. Day 11 is your decision. Optimi is an AI process optimization agency in The Hague for companies with 10 to 100 employees.
Who the prototype is for
It is for companies that have done the AI process scan and want to build the first opportunity. That is the normal route. It is also for companies with one clear, repeating process they want to tackle right away: orders from email into the ERP, purchase invoices into the accounting software, a CMR (the international road waybill) that has to become an invoice. Then you can book the prototype on its own.
From manual process to working prototype in 10 business days
The promise is fixed, and we mean it exactly as written. This is what the timeline looks like:
- Days 1 to 2 · Intake and scope. We record what goes in, what has to come out, which exceptions exist, and who approves. Access to the systems is arranged.
- Days 3 to 8 · Build. Connections to email, ERP, or accounting; the step that pulls data out of a PDF or email; the approval screen; logging. A 15-minute check-in every day with the process owner.
- Day 9 · Testing. On real, historical data, together with the people who do the process. Anything that deviates goes on the exception list.
- Day 10 · Live. For a limited part, for example one client group or one supplier. A handover note comes with it.
- Day 11 and beyond. Your decision: extend it, scale up through workflow automation, or hand the management over to Managed AI.
What you have in hand on day 10
- A working prototype in your own environment.
- An approval step: green goes through, orange is reviewed by a person.
- A manual of 1 to 2 pages.
- A measurement of hours before and after for the part that is live.
- The list of exceptions we ran into and how they are handled now.
- A proposal for scaling up.
What "working" means
Working means it runs on real data, with approval, with logging. No slides, no demo on sample files. If the prototype does not meet those three conditions on day 10, it is not finished yet, and we will tell you so.
Lead time and what it asks of you
10 business days. In week 1, we need about 2 hours a day from one process owner: someone who knows how the process really runs and can say what good looks like. In week 2, it is less, except on the test day.
What you can count on
- It runs on your systems, including Microsoft 365, Exact, AFAS, Twinfield, and common TMS/WMS packages.
- AI models run in EU regions, in our managed environment. Your data is not used to train models.
- The code and configuration are yours.
- Every automation has a person at the controls.
- We work in line with the GDPR and the principles of ISO 27001 and ISO 42001.
Keep reading
- Example case: Wholesale: order entry from email and PDF
- Example case: Transport: from CMR to invoice
- Article: Document automation: process PDFs and invoices without retyping
- Article: From Excel and email to a workflow: choose the first step
01, RESULT
- A working prototype in your own environment, on real data.
- An approval step
- a person approves before anything enters the system.
- A short manual of 1 to 2 pages.
- A measurement
- hours before and after for the part that is live.
- A list of the exceptions we found and how they are handled.
- A proposal for scaling up, with the next processes included.
Process diagram
02, HOW WE WORK
-
Days 1 to 2 · Intake and scope
input, output, exceptions, and who approves. Access to your systems arranged.
-
Days 3 to 8 · Build
connections, the extraction or AI step, the approval screen, and logging. A 15-minute check-in every day.
-
Day 9 · Testing on real, historical data, together with the people who do the process.
-
Day 10 · Live for a limited part, for example one client group. A handover note comes with it.
-
Day 11 and beyond · Your decision
extend it, scale up through workflow automation, or hand it over to Managed AI.
FAQ
FAQ
What exactly does "working" mean?
Three things at once: it runs on real data from your systems, every result passes an approval screen before it is processed, and everything is logged. A demo on sample files is not a prototype to us. On day 10, the prototype processes real orders, invoices, or requests for a defined group.
Can we do it without the scan?
Yes, if the process is clear: one input (for example, an email with a PDF), one destination (for example, an order in Exact), and a known approver. Then you can book the prototype on its own. If you are unsure which process pays off most, the scan is the better first step.
What if the process has more exceptions than expected?
That happens almost every time, which is why we test on historical data on day 9. Exceptions we cannot handle in 10 days go to a queue in the approval screen, where a person handles them as they do now. You get the list of exceptions, so you know what the next iteration fixes.
Which systems does this work with?
Microsoft 365 and Exchange, Exact, AFAS, Twinfield, SnelStart, King, and Unit4 (widely used Dutch business software), common TMS and WMS packages, Salesforce, HubSpot, and SharePoint, among others. Without an API, we work through files, mailboxes, or a small bridge. AI models run in EU regions, in our managed environment.
Who owns the prototype afterward?
You do. Code, configuration, and documentation sit in your environment or are handed over to you. You can continue without us, continue with us through workflow automation, or leave the management to us through Managed AI. No lock-in.
