AI and machine learning can support software development and add capabilities to business applications. Useful projects start with a specific problem, suitable data and a way to evaluate results. Human oversight, privacy and security remain part of the design.
A practical example
A support assistant might help draft replies to common product questions. Start with a bounded set of questions and reviewed reference material. Test incorrect assumptions, missing information and cases that should be escalated. Compare the result with the existing process. Expand the role only when the team can explain both its usefulness and its failure modes.
Before adding AI to a product
| Focus | Practical approach | Check before proceeding |
|---|---|---|
| Problem | Define the decision or task that AI should support. | Compare against a simpler rule-based approach. |
| Evaluation | Use representative examples and an agreed success measure. | A convincing demo is not evidence of reliable performance. |
| Oversight | Provide review, escalation and monitoring. | Plan for incorrect or inappropriate outputs. |
When is a rule-based system better than AI?
When the task has clear, stable rules and the simpler system meets the need. Evaluate reliability, maintenance and cost before assuming a model is necessary.
Useful next steps
For a related perspective, read Responsible AI Development: Data, Impact and Accountability, Using ChatGPT at Work: Drafting, Summarising and Planning, Choosing AI Tools: Inputs, Outputs and Human Review.
If you need help applying this to your business, explore our AI development and automation or discuss your project.
Sources and further reading
By Tech Me Today. Originally published November 16, 2023. Expanded on .
Explore all articles →
