The conversation with the CFO of a mid-sized SMB usually starts in the same place. He lists the tools he already has: an ERP to invoice, a CRM to sell, a gateway to collect, a spreadsheet to reconcile, a channel to notify the customer. All of them work. None of them fail. And yet operations cost more every quarter. When we ask what's excess, the answer is rarely a tool. The answer is a person.
That person doesn't fail, isn't redundant, doesn't do their job badly. They simply act as a human bridge between programs that don't talk to each other. Copying a number from the payment into the ERP. Forwarding an email to the accountant. Manually checking whether the payment came in. Filling the Friday spreadsheet because the system report doesn't say what leadership needs. Each of those tasks takes three minutes. Together they eat up a quarter of the workday. And none of them appears in the job description.
Why it doesn't show up on your P&L
The cost of the human bridge doesn't have a budget line. There's no monthly invoice that reads "operating between systems". There's a payroll cost. And that payroll is justified because "operations are what they are". The problem is that operations are what they are because nobody has done the exercise of looking at how much of that operation exists for things and how much exists to connect things.
In a company of twenty people, with a CFO, a back-office, and two administrative staff, we've measured consistent figures: between 18% and 35% of the admin team's time is spent moving data from one system to another. Data the system could move by itself if someone bothered to connect it. In euros, for that headcount, we're talking about €22,000 to €48,000 a year in gross cost. It will never show up as "human bridge" on the balance sheet. It shows up as "administrative staff", as if it were an immutable constant.
The human bridge has three symptoms
When we visit a company for the first time, we don't ask about the tools. We ask about the symptoms. There are three, and they usually appear together:
- There's one person who knows how everything works, and if that person goes on holiday the team has to resort to WhatsApp to keep orders moving.
- Several parallel spreadsheets exist where the same information lives in different formats, and the only way to know which one is up to date is to ask someone.
- The reports leadership sees are prepared by hand on Friday afternoon and always arrive with a four- or five-day lag from the operational reality.
If all three symptoms are present, there's a human bridge. If two are, probably too. If just one, it's worth investigating before automating anything.
Why it's fragile
A system only the builder understands is not a system. It's a dependency with an open-ended contract.
The human bridge doesn't fail on an ordinary Monday. It fails on the Monday in August, the day the key person is sick, the week an urgent order breaks the routine. It fails when someone new joins the team and nobody has time to explain how things actually work beyond what the org chart says. And above all, it fails when the person holding up the bridge leaves.
At that moment, the company discovers that operations didn't live in the programs but in a single head. Rebuilding that knowledge takes three to six months, with service outages, complaints, and customers who sense the internal disorganisation even if they can't name it.
What to do before automating
The intuitive reaction to the human bridge is to replace it with software. In 80% of cases it's the right one. In the remaining 20%, it's not. Before touching anything, three questions are worth answering:
- Is the process the person runs correct, or is it just what was set up on day one and never revisited? Automating a badly framed process only repeats it faster.
- Is there a baseline number? That is: how much time is spent on that work today, how often, with what error margin? Without that number, you can't prove later that improvement happened.
- Who's going to maintain the system once delivered? If the answer is "we'll see", the system shouldn't be built yet. Automation shifts the dependency, it doesn't remove it.
The human bridge is not a people problem
It's worth stating clearly, because the debate usually gets twisted at this point. Removing the human bridge doesn't mean firing anyone. It means that the person who today copies and pastes stops copying and pasting, and can dedicate themselves to the work they were actually hired for. In the vast majority of companies we work with, headcount doesn't shrink. It gets reallocated. The admin team goes from operating between systems to operating on customers, suppliers, new lines of business.
That reallocation is, in practice, the only way an SMB can grow without doubling the team. Without it, any increase in volume translates into a proportional increase in human bridge, until the model breaks.
How it's spotted from the outside
You don't need a three-month audit to know if there's a human bridge. Twenty minutes of conversation with two people from the admin team reveal it. Just ask how an order reaches the ERP, how the customer is notified when their payment goes through, and what happens when it doesn't. If any of those three answers contains the word "we notify", "I forward", or "I pass along", there's a human bridge. And there's probably room to remove three or four hours of weekly work per person without changing a single program.
That's the conversation any udae project starts with. No canned diagnosis. No quote up front. First we need to understand how the business works, and only then do we propose anything.