Baling wire between the applications

Manuel Pattyn · October 2026

Short answer. Small apps that catch the gaps of bigger systems work like baling wire: it holds, as long as nobody pulls on it. They automate work that shouldn't exist and make the cause permanent. A patch is fine, as a temporary fix with an end date and a plan for what goes underneath.

We recently saw this at an installation company with a few hundred vans. Each app neatly catches a gap the bigger systems leave open. Together they form a layer of baling wire between the applications.

Three apps, three gaps

The company has an ERP for the jobs, a separate calendar app for the technicians, a time-registration app, two track-and-trace systems, a fuel card portal and an Excel file where the planning really happens. An AI agency proposed three applications to close the gaps in between.

The planning agent. Two planners spend every day (re)planning some 450 to 500 maintenance visits a month. They export from the ERP to Excel, plan in Excel, copy the visits into the calendar app and email customers to confirm. The refrigeration team still plans on paper, in folders. The agent takes over the retyping and emailing. Striking: there's no new screen. The agent works in the existing Excel file, which has to move to the cloud for it, and changes the status colours in it.

The time and route checks. Technicians register their trips and working hours manually. Every van already has track-and-trace. HR compares both every day, with the planning next to it, to calculate wages correctly: several hours a day. The app does that comparison automatically and emails HR an Excel with the deviations. A check on fuel fills and consumption comes on top.

The chatbot. An assistant for customer questions, with an option to let customers solve faults themselves first. Calculated on 50 to 200 conversations a month.

Each of the three is a good idea. But together they take the existing landscape as a given and build a layer on top of it.

What the patches cost

For one company in the group, the three proposals add up to about €65,600 to implement and €52,500 a year. Over three years that's well over €220,000, excl. VAT.

Not in those totals: the rollout to the other companies in the group (another 200 technicians, €9,000 a month extra in the first proposal), a twelve-month contract per application, each with its own renewal date, and the licences of all the systems underneath. Those stay. The patches come on top.

The vendor is honest about it. With the prices, it names two risks: what using the AI models will cost, and how much support and maintenance it will take to keep the integrations with all the tools working.

Why baling wire doesn't hold

Nine integrations to seven systems. The planning agent talks to the ERP, the calendar app, the Excel file and the mailbox. The control app to the time-registration app, two track-and-trace systems, the planning and the fuel card portal, whose integration wasn't yet confirmed at the time of the proposal. Every update of any one of those seven can break an integration.

The Excel file becomes a production system. An AI agent that changes status colours in a spreadsheet is clever work. But it keeps the company's most important planning in a file anyone can accidentally sort, change or copy.

The data stays bad. For 18 to 20% of customers there's no contact detail in the system at all: no phone, no email. An agent that emails customers doesn't email those customers. So the planners stay, now for the exceptions.

Double registration gets double-checked. Technicians manually register what track-and-trace already knows. The control app compares both and emails HR a list. That beats hours of manual work. But the question nobody asks: why is it registered twice?

What's still analogue has to go digital first. The refrigeration team plans on paper. In the proposal, that planning is listed as "after digitalisation". The patch can only stick to a surface that doesn't exist yet.

First the core, then the smart layer

AI isn't the problem here. The order is: core, data, automation. In a central system such as Odoo, the maintenance job, the planning and the technician's calendar are the same record. The customer confirms via a link in the email and the status changes by itself. The technician starts and stops the job in the app, with track-and-trace as a source instead of a check afterwards. Cleaning up contact details is a project with an end, not an app, and after that the system enforces that no new customer is created without a phone number or email.

Honestly: a standard planning module doesn't optimise as cleverly as a custom AI agent. For the summer months when half the team has to switch to service, such an agent can really make the difference. But as the last layer, on one integration instead of nine. Not as glue.

Five questions before you sign

The example is anonymised. Amounts come from real proposals, excl. VAT, and are rounded. Also read: Odoo with custom work or six tools side by side →

If you can't answer the first question without "because that's just how it is", you're buying baling wire.

Let's talk →

← All posts