Everyone says AI. Nobody says where.
The board asks about it, vendors call, the news is full of it. Nobody has pointed at one place in your company and shown what it would save.
AI and digitalisation
Everyone talks about AI. Nobody says where to start.
We find where it would pay off in your operation and start with one step.
Documents · Requests · Company data · Decisions · Rules
Most companies have seen a demo and still cannot point to the place where it would pay off. We start from the work your people do by hand, and we name the few candidates worth doing first.
DocumentsRequestsWhat people knowRule or model?The answer, prepared
Is this your case?
The tools exist. The question is where they fit here.
The place where it pays off sits inside the daily work: who reads what arrives, who decides, and what gets retyped.
The demo answered questions about a document.
It read a file and answered well. Your quotes, your exceptions and your approvals live somewhere else, and nobody showed how they would reach it.
Someone pastes rows into a chat window.
A person copies data out of a system, pastes it into a chat window and copies the answer back. It works, and it stops the day that person is away.
You suspect a plain rule would do it.
Half the cases follow the same instruction every time, and someone still reads each one. Nobody has checked whether a rule would settle them.
What it costs today
The pilot ends. The reading and retyping stay.
A pilot that never reaches production costs the time it took and leaves the work where it was. Meanwhile decisions wait for a report someone assembles by hand, and experienced people spend part of the day reading, sorting and retyping what arrives. Those hours are the real price, and they repeat every week.
E-mails and requestsDocumentsERP / CRMWhat people know
Where the decision is made
A prepared answerA record already filled in
The first step is to find the one decision that happens often enough to be worth automating, and to name the data it needs.
Your options
Several of these need no AI at all.
A standard tool you already pay for, a rule with no model in it, or one small connection can each be the right answer. We say so when they are.
Use what you already pay for
Fits when
the task sits inside one product you already own
Several systems you use already have an assistant, a document reader or an import that does this. We check that first and show you what it covers.
A rule, with no model in it
Fits when
the input is regular and the decision follows a written rule
Sorting, copying, checking and sending can run on plain automation. It is cheaper to run, and you can see exactly why it did what it did.
Reading what arrives as text
Fits when
the input is e-mails, orders, drawings or documents
A model can read what comes in, pull out the fields and prepare the record or the answer. A person confirms the cases it was unsure about.
An assistant over your own systems
Fits when
the question crosses orders, stock and dispatch
One connection point publishes your systems to an assistant under your permissions, so it answers from your data. The systems stay as they are.
Wait
Fits when
the data or the owner of the decision is not there yet
If the answer would sit on data nobody maintains, or on a decision nobody owns, starting now creates work without a result. We will say so.
How we work on it
We learn the operation first. The tool comes after that.
Your operation has rules that never made it into a manual. We learn them from the people who apply them every day.
We learn the work first
We sit with the people who read, decide and retype, and we write down the rules they apply, including the exceptions. That part decides whether anything here works.
One candidate, written down
We agree on one task, what goes in, what comes out and how we will know it worked. The scope and the estimate are written before anything is built.
In front of real users early
The first version goes to the people who do the job, on their real cases. We change it with them while it is still small.
We stay with it
We take responsibility for the result in daily use, and for what happens when the inputs change. The code, the data and the rules stay yours.
Relevant work
Work that started from one task someone did by hand.
Retail, Czech Republic and Slovakia
Routine legal work handled by automation
Hundreds of work hours saved a year
Datart runs consumer electronics retail across two countries. We went through the routine steps in the legal department with the people doing them, and moved the repeated part into automation.
Automotive tier-2
A quote that stopped waiting for an engineer
Half a day per frame to about 15 minutes
Costing one wire frame meant an engineer reading the 3D model for half a day, and one project can carry twenty of them. The model now goes in and a priced quote comes out, on the company's own calculation rules.
Customer requests
Model example: incoming requests sorted and filled in
Model example, no client numbers
A request arrives by e-mail. The system reads it, decides which product line it belongs to, fills in the record, and puts the cases it was unsure about in front of a person. This is a model example, drawn to show the shape of a first step, and it carries no client numbers.
The same question opens all of these. Which decision happens often enough to be worth automating, and what does it need to read before it can be made.
We start with the work
The questions we will ask in the first meeting.
- Which task did someone do three times today that reads the same way every time?
- When a request comes in, who decides what it is, and what do they look at before deciding?
- Which of your rules is written down, and which one lives with one person?
- What has to be right every single time, and what can a person still check afterwards?
Why Moravio
A software partner for deciding where to start
Build or buy advice
We check first whether a product you already pay for can do it. Custom development starts where your rules, your data or the exceptions require it.
Built around what you keep
ERP, CRM and the tools that do their job stay in place. What we add reads from them and writes back to them.
A rule before a model
Where a plain rule settles the case, we build the rule. A model is used where the input arrives as text that a person has to read.
Yours to own
The code, the data and the rules stay with your team, and the system can run on your infrastructure.
The first step
One process, followed all the way through.
We take one place where people read, sort or retype, and we follow it from what arrives to what someone does with it. You need nothing prepared and no specification. You talk to the people who design and build the system, so nothing is lost on the way to a delivery team.
01
What arrives
We look at what comes in over a normal week. E-mails, orders, documents, questions, and who opens them first.
02
Who decides what
We follow one item to the person who decides, and we write down the rule they apply and the cases where they set it aside.
03
What data exists
We check what the systems already hold, what is only in a spreadsheet, and what lives in someone's head.
04
What to start with
We put the candidates side by side with what each one needs and what it would save, and we name the one to do first.
What you bring
One real week of requests, orders or documents. Nothing prepared, no specification.
What you leave with
A short list of candidates, what each one needs as input and where its limits are, the one to start with, and a scope and estimate for it.
FAQ
Frequently asked questions
That is the first part of the work. We sit with the people who do the job and write down the decisions, the data and the exceptions before anything is designed. We have done it in retail, in automotive supply and in transport, and the operation was different every time.
No. The ERP, the CRM and the tools that do their job stay where they are. What we add reads from them and writes back, and a replacement only comes up where a tool forces the wrong way of working.
The first version goes in front of the people who do the job, on their real cases, while it is still small. We change it with them. Nothing is switched off until the new part has earned trust in daily work.
You leave the first step with a written scope, an estimate, the options we rejected and the reason, and the risks. IT gets the technical picture in their own terms: permissions, where the data goes, and who owns the code.
For many tasks the answer is yes, because the inputs already exist as e-mails, documents and records. Where a task would need data nobody keeps today, we say so and put it aside until that data exists.
That is your decision, and we set it up to your rule. Processing can stay inside your environment, and what a tool is allowed to read and to change is set by your permissions and recorded.
Yes, and it is usually the sensible scope. One task shows you the real inputs, the exceptions and the value before anything larger is agreed.
Start with the situation
What do you need but you can’t buy off the shelf?
Send us the messy version. We will help make the next decision clear.