™Business
Custom AI applications vs off-the-shelf tools
Buy when your process resembles everyone else's. Build when the difference between your process and the standard one is the point.
By Tomryn ·
This is usually framed as a cost comparison, which is the least useful way to look at it. The real question is whether the way you do the work is incidental or is part of why customers choose you.
The case for buying
- It works this month, not next quarter.
- Somebody else maintains it, absorbs model changes and handles security updates.
- The cost is predictable and visible.
- For genuinely standard work (note taking, transcription, general drafting), a product will beat anything custom.
The case for building
- Your process is not the standard one and forcing it into a product's shape loses something that matters.
- The work depends on internal information that a general tool cannot see.
- It must integrate with systems the product does not support.
- Data handling requirements rule out sending the information to a third party.
- The same subscription is being paid for dozens of seats to get one feature.
Costs each side hides
Bought tools hide the cost of adaptation: reshaping how people work to match the software, per-seat pricing as headcount grows, and the difficulty of leaving once the data lives there.
Built tools hide the cost of ownership: someone has to maintain it, models and APIs change, and the person who understands it may leave. Budgeting only for the build is the most common mistake.
A reasonable default
Buy the general capability, build the specific bit. Most organisations should be using off-the-shelf tools for the common tasks and reserving custom work for the one or two workflows that are genuinely theirs. It is also entirely reasonable to start with a bought tool, learn what you actually need, and build later with much better information.