This post part of a series on undervalued practices in ML/AI:
- Part 1: Getting started with ML/AI
- Part 2: Hiring and Team Structure
- Part 3: Planning projects
- Part 4: Project Execution (this post)
- Conclusion (this post)
Over the past four posts, I’ve walked through a number of practices that will help a company make the most of their ML/AI investment and potential, while dramatically reducing their risk. For example:
- Develop an organization-wide understanding of what ML/AI really is, and what is possible with it in this company.
- Develop concrete plans of what to do, couched with realistic expectations.
- Take the time to understand your data.
- Scope and time-box your experiments.
These all strike me as fairly reasonable steps. Following them will put you on the right path for the long-term, and will pay large dividends over time.
Why, then, do I say they are “undervalued” practices? These should be industry-standard approaches, but most companies miscalculate the long-term payoff of early discipline. They want to get straight to the “exciting” part of ML/AI – building models, say – and they fall for the siren song of cutting corners.
That leads me to the core lesson of this series:
All in all, the most undervalued assets in ML/AI are honesty and discipline in the leadership team.
Cutitng corners offers an attractive start, because company leadership gets to claim they’re “doing AI” from the first day. But it costs them over time. They wind up re-doing work, restructuring their ML/AI efforts, and possibly even replacing data scientists who get bored and leave. Being first in your peer group to buy into AI is not the same as being the first to achieve effective AI.
These practices I’ve outlined work best for a new company, but they’re still suitable for an existing effort. What if your company has already taken the shortcuts and you’re feeling the pain? It’s time to pause your ML/AI work and draft draft a plan for a reset.
To implement this you may have to make some difficult changes: restructure or downsize your team, scale back on your ML/AI ambitions, and swallow some pride as you admit that it’s time to chart a new course. As painful as this may be, it’s better done sooner than later.