A guide to custom AI software for business leaders
What custom AI software is, when it's worth building instead of buying, how a project usually runs, and what to ask a firm before you hire one.
Insights
Articles on building AI software
We write about the questions clients ask us most: when to build instead of buy, how to scope a first project, what forward-deployed engineers do all day, and why so many pilots never make it to production.
- Custom AI
- Strategy
- Delivery
- Forward-deployed engineering
- Custom AI software is written for one organization's problem. The AI is often a small part of it; most of the work is integration, interface and testing.
- It's worth building when the problem depends on your own processes or data and nothing on the market handles it well.
- Projects go best when the builders spend time with the people who have the problem and test an early version with them.
All insights
Build or buy AI: how to decide
When an off-the-shelf AI product is the better choice, when custom AI software is, and how to settle the question for a specific problem.
How to scope an AI use case
A practical way to scope an AI project: watch the work, choose one problem, write requirements and success measures, then test a prototype.
What is a forward-deployed engineer?
A forward-deployed engineer works inside the client's organization, learning the problem from the people who have it and building software where it's used.
What makes AI software production-ready
A checklist for taking AI software into daily use: integration, security, evaluation, safeguards, human review, monitoring, documentation and ownership.
Why AI pilots stall before production
Why so many AI pilots never reach daily use, and what tends to get them over the line: a real problem, honest data, agreed measures and proper engineering.