Automation starts with a diagnosis, not a tool.
Most automation projects pick the tool before anyone has measured where the time actually goes. We work the other way round: map the workflow first, tie every inefficiency to a specific person, a time estimate and a business impact — and only then build.
Book a call (30 min) →A tool that fixes the wrong process gets quietly worked around.
AI projects rarely fail because the tool was bad. They fail because the tool landed on a process nobody had examined: a licence is bought, training is held, and three months later nobody opens it.
So the first step is not choosing a platform. The first step is establishing where the time goes, whose time it is, and what that costs. Once that exists on paper, the choice of tool stops being a matter of taste and becomes something you can check.
What the diagnostic found elsewhere.
Four findings from engagements that became our case studies. None is an estimate from a meeting — each was measured during the diagnostic and tied to a specific person and a specific step in their day.
An hour a day · managed print, Belgrade
After every service call the same data was entered by hand in three places: a paper field report, a ticket, and four documents in the ERP. One person, every day, adding no new information.
An hour a day · managed print, Belgrade
Answering one question about one of 2,500 devices meant opening three systems side by side and cross-referencing them by hand. A daily bottleneck for service, procurement and field decisions alike.
45 minutes a day · policy organisation, UK
Every morning began with a manual scan of the media for stories relevant to the research. Underneath it sat years of journalist contacts, mapped to policy beats and unused.
Two to three days · policy organisation, UK
Planning a single event started from nothing every time: searching for speakers with no shared record, writing the blurb from scratch, then typing individual invitations by hand.
Two to three weeks. Fixed scope.
The diagnostic is a structured engagement, not a workshop. We interview the whole team, map the workflow end to end, and analyse the systems and data flows that hold it together. Every inefficiency we find is tied to a specific person, a time estimate and a business impact — not to a general claim that there is room for improvement.
Workflow map
How the work actually moves through the company, including the steps nobody wrote down.
Systems map
Which systems hold which data, where data is duplicated, and where it is retyped by hand.
Three to five opportunities
Automation opportunities in priority order, each with an estimated return on investment.
What “tied to a specific person” means.
An analysis that ends up in a drawer usually says there is “room for optimisation”. A sentence like that can be neither disputed nor budgeted for. A finding is only usable when it carries four things, and each has to survive being shown to the person who does that work.
Who
A role, not a department. “Field engineer”, not “technical services”. If a finding cannot be put to one person to confirm or dispute, it has not been checked.
Which step
The exact point in the workflow, not the whole process. Not “order handling”, but “retyping the order number from an email into the ERP”.
How much time
A measured estimate per repetition, multiplied by frequency. That is where the weekly and annual figure comes from — one finance can check.
What it means for the business
A late quote, an invoice error, a missed deadline. The hours are the cost; the consequence is why somebody approves a budget.
Diagnose or Pilot — how to choose.
Design and build follow the diagnostic, and both are set out on the services page. What matters here is the first decision, because that is the only one being made now.
Take the diagnostic if you do not know where the time is going, or if you need a number to defend a budget with. You get the measurement and the order to act in, with no obligation to continue with us.
Take Pilot if the decision is broadly made and what is missing is proof that it works in your business specifically. The whole loop on one process, six weeks, the success criterion written down before we start — and the design and delivery fee comes back in full if we miss it. The risk of the first step sits on our side.
Frequently asked.
- How long does business process automation take to a first result?
- The diagnostic runs two to three weeks, design one. Build time depends on the scope of the work. If you need one process running to a known deadline, Pilot covers the whole loop — diagnosis, design and delivery of a single opportunity — in six weeks of fixed scope.
- Do we have to replace the systems we already use?
- Not necessarily. We work with the tools you already pay for and connect them to each other. Introducing a new platform is a decision made at the design stage, through the build-versus-buy recommendation, and it is argued in the document you receive.
- How do you measure whether the automation paid off?
- Every opportunity from the diagnostic comes with a return estimate derived from the time specific people spend on specific steps. In Pilot the success criterion is written down together before the work begins, and the result is measured against it. Miss the criterion and the design and delivery fee is returned in full.
- What if we already know what to automate?
- Then you skip the diagnostic. The Delivery Sprint is a fixed-scope engagement of two to four weeks for one or two specific automations that are ready to build. You can also bring your own diagnosis and start at the design stage.
- Do you work with small companies or only large ones?
- Company size matters less than whether the process repeats. An hour a day lost across five people and an hour a day across fifty are measured the same way; only the figure that comes out differs. We have worked with family firms and with large systems.
- What happens once the build is finished?
- Your choice, and it is made at the end rather than the beginning. We can hand the system over and step back, or stay on as your AI operations team under a monthly engagement. The scope of those options is set out on the services page.
Start here
Start from the measurement, not the tool.
On a thirty-minute call we go through the process costing you most, and the systems holding it together. From that we can see whether a diagnostic makes sense and what its scope would be. If it does not, we will tell you that too — and if it does, the document you get stays yours whatever you decide next.
Book a call (30 min) →or write to ai@meteorit.rs