Custom software & automation for companies: a guide

Custom software has stopped being a luxury of large companies. AI shortens the build - developers supported by AI complete tasks 55% faster - so the barrier to an application shaped to your processes is lower than ever. But a cheap start is not the whole bill: maintenance consumes an estimated 60-90% of software's total cost, and unsupervised generated code can be leaky - in Veracode's study 45% of AI-generated code failed security tests. This guide sorts out how to use all this wisely.

Each topic is summarized below; where we already have a full post with data and sources, we link to it. Further posts in this cluster are in the works and will be linked here.

Off-the-shelf system or custom application?

Off-the-shelf wins when the process is standard and not your edge - accounting, e-mail, a basic CRM. A bespoke application pays off where the process is your competitive advantage and a boxed product forces compromises: you pay per-seat licences for 80% of features you never use, and you still end up bending your company to fit the tool instead of the other way round.

Where do I start with process automation?

With a process map and one candidate: repetitive, high-volume, with a measurable cost of error. Before you build a system, check the integration quick wins - wiring together the tools you already have (ERP, CRM, spreadsheets, e-mail) often returns in weeks. Read the full post: how to start automating →

What does custom software cost?

The build price is the smaller part of the bill - most of the lifecycle cost is maintenance. So when you compare a "cheaper and faster" quote with one that includes specification and architecture, you are not comparing prices - you are comparing where you will pay: up front, or for years after launch. Add the alternative maths: per-seat licences summed over 3-5 years often exceed the cost of an application that is yours.

Why do IT projects blow their budgets and deadlines?

Most often because nobody wrote down exactly what was to be built - scope grows mid-flight, and every change discovered late costs many times more than the same decision made on paper. The fix is not a rigid contract but a specification: needs, features and architecture settled before any code exists. Read the full post: why the spec matters →

Software built with AI - opportunity or risk?

Both, depending on the process. AI radically speeds up the build, but "vibe coding" - accepting code without understanding it - moves the cost to production: 84% of developers use AI, yet only 3.1% highly trust its output. The proven setup is spec-driven: specification first, then AI builds to the plan and a human controls the result. Read the full post: AI software without failures →

How do I measure the return on automation?

Hours times rate is only the start - the full bill includes fewer errors, shorter cycle time and throughput without extra headcount. The measurement rule is the same as for AI: a baseline before you start, business outcomes instead of activity, the right horizon. Read the full post: how to measure ROI →

How we help

We build applications and automations shaped to your processes - from simple fixes to complex systems. We always start with a specification, and the competence and documentation stay with you:

• Process analysis - what to automate first, and what to leave alone.
• Specification and architecture - a plan you could take to any contractor.
• AI-assisted build - faster than classic, but with engineering control and tests.
• Maintenance and monitoring - because most of the cost and value comes after launch.

Frequently asked questions

Off-the-shelf system or custom application? Off-the-shelf wins when the process is standard and not your edge - accounting, e-mail, a basic CRM. Custom pays off where the process is your competitive advantage and no boxed product fits without compromises.

Where do I start with process automation? With one process that is repetitive, high-volume and has a measurable cost of error. Often the first step is not building a system but wiring your existing tools together - a return in weeks, not months.

What does custom software cost? Less than it used to - AI shortens the build - but the purchase price is not the whole bill. Most of the lifecycle cost of software is maintenance, so architecture and specification matter more than the build rate.

Key takeaways

• The barrier to custom software is lower than ever - AI shortens the build, but does not excuse you from engineering.
• Compare lifecycle cost, not build price - most of the bill is maintenance.
• Start with a process, not a system - integration quick wins before a big build.
• The specification is the cheapest place for decisions - skipping it is deferred cost, not savings.
• Measure business outcomes - errors, cycle time and throughput, not just hours saved.