By Abdul Ghani

Artificial intelligence is increasingly automating repetitive tasks across data and machine-learning professions, but its adoption is shifting human workers towards strategic direction, interpretation, quality assurance and oversight rather than eliminating entire occupations, according to a new analysis.

DataCamp's The State of AI Careers 2026 finds a substantial gap between what AI could theoretically automate and what workers are actually using it for, suggesting that the transformation of professional roles is occurring more gradually than headline estimates of automation potential might imply.

The report, available with Wealth Pakistan, cites research comparing theoretical AI capabilities with observed use and finds that even occupations considered highly exposed to generative AI are far from fully automated.

In computer and mathematical occupations, early models projected that 94% of tasks could be automated, while observed exposure based on Claude usage covered only about 33% of those tasks.

The report says this gap reflects an important distinction between technical capability and practical use. AI systems tend to perform well on relatively easy-to-learn tasks with objective measures of success, such as drafting standard code or summarising documents, while much professional work continues to require situational awareness and complex judgment.

Data-related professions illustrate this transition particularly clearly. Rather than removing the need for data professionals, AI is automating specific parts of their workflows while shifting human responsibility towards areas requiring direction, review and judgment.

In data cleaning, AI systems can detect statistical outliers and structural inconsistencies, align differing field names across databases, fill missing values and mask sensitive personal information. Automated systems can also perform preliminary formatting and resolve conflicting entries.

Human involvement, however, remains important for setting parameters, ensuring the quality of imputed data and reviewing compliance requirements, particularly when datasets are complex.

A similar shift is taking place in data analysis. AI systems can perform standard statistical calculations, translate plain-English questions into structured database queries, detect trends, construct dashboards and generate textual summaries of data patterns.

This reduces the need for manual execution of routine analytical processes and shifts the data professional's role towards providing strategic direction, interpreting outputs, assuring quality and assessing the business implications of findings.

In data labelling, AI-powered systems can generate initial labels, extend tags across sequential data and classify text and images according to semantic patterns. High-confidence samples can pass through the pipeline without manual supervision, while more complex edge cases are routed to human reviewers.

The result is a shift in human effort away from repetitive manual annotation towards defining initial requirements and maintaining quality.

Machine-learning development is undergoing a comparable transformation. Automated platforms can perform feature engineering, test multiple algorithms, tune mathematical parameters and package successful models for production.

Rather than spending as much time on repetitive optimisation work, machine-learning professionals increasingly focus on defining strategic objectives, specifying what models should achieve, reviewing their performance and assessing their business impact.

AI is also changing machine-learning operations, or MLOps. Automated systems can detect changes in incoming data, initiate retraining when performance falls below defined targets, validate updated models and deploy them into production.

Human specialists remain important for setting performance thresholds, conducting quality assurance and making architectural decisions around production systems.

In data governance, AI can collect audit evidence, map internal information against regulatory frameworks, identify privacy and policy deviations and automatically generate technical documentation. Automated systems can also monitor data lineage and access records on a continuous basis.

Human responsibility nevertheless remains central in areas such as audit review, policy definition, incident resolution and quality assurance.

Taken together, the examples point to a broader restructuring of data careers. The report argues that a complex occupation consists of many individual tasks, and AI is more likely to automate particular repetitive and well-defined tasks than eliminate the occupation as a whole.

This distinction is important because theoretical estimates of AI exposure can overstate the immediate risk to workers. A profession may include tasks that AI can technically perform while still relying on human judgment for other parts of the job.

The report therefore argues that career resilience will increasingly depend on workers' ability to use and complement AI tools. It identifies programming, computer science, database management and communication as important foundations, while recommending additional AI capabilities including prompt engineering, agent management and domain-specific applications.

The emerging picture is consequently less one of AI replacing entire data careers and more one of those careers being redesigned. As automated systems absorb a larger share of repetitive execution, human responsibilities are increasingly shifting towards strategic objectives, interpretation, quality assurance, and the application of technical outputs to real-world problems.

Credit: INP-WealthPk