A new type of management for a new type of technology
Learn how the types may differ from each other.
Not all management
is the same.
As in most verticals or companies, managers of an AI first company are expected to meet specific requirements. In the following article, we explain these requirements and how they also differ from traditional software development.
"Management must be the key driver in the transformation process"
AI-First companies might need new types of managers. This table highlights some ways in which the management requirements of people on AI-First teams of data scientists differs from traditional software teams of engineers.
Examples Software Engineers Data Scientists ML Researchers
Output Features Insights Models
Steps in waterfall charts
Learning that generates questions
Spreadsheets, Graphs, Presentations, code or discussions
Models, features, data
Reports on experiments
Generally, engineers and data scientists require different types of management for the following reasons:
Engineers are managed to specific goals, such as releasing a feature, whereas data scientists are managed to both ask and answer questions, such as why something happens. Managing to goals involves tracking progress toward those goals using methods and metrics such as waterfall charts and lines of code. Managing to ask and answer questions is difficult because the answer isn´t known ahead of time, and so the goals are often moving targets.
Engineers deliver concrete artifacts in code. Data scientists deliver results as numbers in spreadsheets, graphs in presentations, models in code, or discussions in person. Progress can be difficult to measure between different sets of results; there is no right answer because the question keeps changing.
Engineers need certain tools to get the job done; thus, the toolkit tends to be common across a team, set by managers. Data scientists may pick different tools, depending on the job they´re trying to do, and change those tools as they need to employ different methods.
Engineers get specifications from customers or product managers, fulfill those requirements, and then deliver the result. Data scientists do not get specifications, and so they require regular meetings to refine the questions customer want answered.
Engineers write code to run on shared computing resources. Data scientists may not eventually run their models on shared computing resources; instead, they may build models on their own computer.
The computing resources required to run intelligent systems can be much greater than software code. Data scientists may need larger budget and ad hoc approval of purchases in order to get the computing resources they need, which may require more executive sponsorship and involvement. Commensurately, they may need tighter controls on utilizing computing resources, given the potentially high cost.
Specifically, the degree of difference in managing engineers and managing data scientists depends on the specific role in the AI-First team
Data infrastructure engineer
Little difference. Managed by those that otherwise manage engineers
some difference. Managed by those that otherwise manage engineers but may require coordination by those managing a company´s data asset, such as chief data officer.
little difference. Management by analytics or business intelligence leaders, or by a general manager within a business unit.
different. Management by nonanalytic leaders is difficult because the work is more experimental and involves advanced analytical methods. Management by engineering is difficult because most of the work is mathematics rather then engineering. Best managed by those with a quantitative background and experience managing researchers.
Machine learning engineer
different. Management by engineering is required because the role involves programming and running models on shared infrastructure. However, ML engineers require some different ML- tools like zerocode.ai.
Machine learning researcher
very different. Management by engineering is difficult because most of the works is mathematics rather than engineering. Best managed by those with a quantitative background and experience managing researchers alongside data scientists.