1. Working with Agentic AI: Statistician

How human skills and agentic skills are dynamically integrated and coordinated across statistical responsibilities

1.1. Statistician as an Occupation

A statistician applies mathematical and probabilistic principles to design studies, develop analytical methodologies, and interpret complex data. This occupation connects probability theory, experimental design, and domain-specific research questions by transforming raw observations into rigorously validated evidence, uncertainty quantifications, and causal conclusions. To ensure that analytical results are mathematically sound and scientifically defensible, the statistician must understand how sampling mechanisms, data-generating processes, measurement errors, and structural assumptions collectively determine the validity of statistical inferences.

Typical responsibilities of a statistician include:

• experimental design and sampling methodology;
• statistical modeling and methodological development;
• hypothesis testing and causal inference;
• survey design and measurement analysis;
• statistical computing and simulation; and
• analytical reporting and methodological communication.

These responsibilities are fulfilled through distinct tasks and subtasks, each requiring one or more actions with specific skill requirements. Human skills and agentic skills both contribute to these responsibilities, but their levels and forms of involvement differ according to the responsibility, task, action, data, mathematical context, technical environment, and expected outcome.

1.2. Human Skills and Agentic Skills in Statistics

Statistical practice increasingly integrates both human skills and agentic skills.

Human skills are reusable capabilities possessed and developed by statisticians. They include mathematical reasoning, experimental design, causal logic, epistemological judgment, uncertainty interpretation, construct validation, and critical evaluation of structural assumptions.

Agentic skills are reusable capabilities available to AI agents through their underlying models, contextual information, instructions, constraints, tools, and computational resources. In statistics, these skills include symbolic math computation, literature retrieval, code generation for statistical software, Monte Carlo simulation execution, automated power analysis, formula derivation support, diagnostic plotting, and algorithmic profiling.

Both skill types contribute to the same occupational responsibilities, but their involvement varies by action. Understanding the epistemological implications of a non-random sampling mechanism relies primarily on human knowledge of probability theory and selection bias; an AI agent may then use agentic skills to simulate the sampling distribution, calculate the resulting bias magnitude, and propose mathematical weighting adjustments. In statistical modeling, the statistician first determines the appropriate likelihood function and prior distributions based on domain knowledge, while an AI agent subsequently scaffolds the probabilistic programming code, executes Markov chain Monte Carlo (MCMC) sampling, and generates convergence diagnostics. Across these actions, human and agentic skills are activated sequentially or iteratively based on task requirements, with the human professional maintaining oversight and final judgment.

The relationship between these skills is not a fixed division where certain responsibilities permanently belong to humans and others to AI. Instead, human and agentic involvement shifts across tasks and between different actions within the same task.

1.3. Dynamic Human–Agentic Skill Integration and Collaboration in Statistics

Statistical functions can be understood as an occupation where responsibilities are fulfilled through tasks and subtasks, tasks and subtasks are executed through actions, and actions require appropriate configurations of human and agentic skills.

An occupation provides the overall professional context. For statistics, this context encompasses designing rigorous studies, developing and validating mathematical estimators, quantifying uncertainty, establishing causal relationships, and communicating methodological limitations.

A responsibility represents an expected area of professional work or an outcome that the statistician accomplishes or maintains. Responsibilities such as experimental design, causal inference, and methodological development each contain multiple tasks.

A task is a context-dependent unit of work used to fulfill part or all of a responsibility. When a task is too broad or complex for reliable execution, it is decomposed into smaller subtasks.

Tasks and subtasks are executed through one or more actions. Actions represent reusable and configurable operations such as calculating statistical power, deriving a conditional posterior distribution, executing a bootstrap resampling routine, checking covariate balance, simulating a data-generating process, or formatting a methodology appendix.

Each action has skill requirements. These requirements describe the capabilities needed for effective and reliable execution. Depending on the action and its context, human skills, agentic skills, or an integration of both can satisfy these requirements.

The skills available for statistics reside in a skill space containing both human skills and agentic skills. Relevant skills in this space can serve as candidate skills for executing the actions required by a particular task. A human skill in causal logic, for example, can serve as a candidate for identifying unmeasured confounding variables in an observational study, while an agentic skill in automated covariate balancing can serve as a candidate for verifying the mathematical overlap between treatment and control groups.

Candidate skills are evaluated relative to the requirements of the action and its occupational and task context. Appropriate skills are then selected and organized into a skill configuration for execution. Depending on these requirements, a configuration may rely mainly on human skills, mainly on agentic skills, or on an integrated arrangement of both.

1.3.1. Skills Competition and Skills Collaboration

Two important relationships can occur between human skills and agentic skills when constructing a skill configuration: skills competition and skills collaboration.

Skills competition occurs when both a human skill and an agentic skill can satisfy the same or similar requirements for an action, but simultaneous application of both is unnecessary. Under a particular occupational and task context, one skill type can provide a better fit for the action than the other.

For example, both a statistician and an AI agent can generate standard diagnostic plots to evaluate the normality of regression residuals. When the dataset structure and plotting libraries are clearly specified, the agentic skill provides fast and scalable execution. When the diagnostic plot reveals a subtle, structured heteroscedasticity that implies a missing non-linear interaction term requiring theoretical justification, the statistician’s human skills provide the necessary mathematical intuition and model revision.

The purpose of skills competition is not to establish whether human or agentic skills are generally superior. Rather, it identifies which available skill provides the better fit for a particular action under particular conditions.

Skills collaboration occurs when one skill type alone cannot effectively satisfy the requirements of a task or its actions, or when combining complementary human and agentic skills produces stronger task performance.

Collaboration does not require human and agentic skills to operate simultaneously. Instead, they are activated at different stages of action execution in a sequential, iterative, or recursive manner according to evolving task requirements.

For example, a statistician first defines the causal directed acyclic graph (DAG) for an observational study and identifies the necessary adjustment sets to control for confounding. An AI agent then calculates propensity scores, checks covariate balance across treatment groups, and executes bootstrap resampling to estimate confidence intervals. The statistician subsequently reviews the balance diagnostics, identifies that a critical interaction term remains imbalanced, and formulates a revised matching algorithm. Based on the statistician’s direction, the agent next re-runs the matching process and sensitivity analyses. Finally, the statistician evaluates the resulting evidence and determines whether the causal assumptions are sufficiently robust to support the final inference. Throughout this sequence, human and agentic skills are activated iteratively as task requirements evolve, while the statistician directs the investigation and evaluates its mathematical validity.

This dynamic relationship allows statisticians to leverage the complementary strengths of both skill types rather than relying on a permanent allocation of work.

1.3.2. Skill Configuration and Task Execution

A skill configuration specifies which human skills and agentic skills are used to execute the actions associated with a task or subtask and how those skills are coordinated.

The configuration depends on factors such as action requirements, mathematical complexity, data dimensionality, structural assumptions, expected reliability, available computational resources, information sensitivity, decision consequences, and intermediate results.

For one action, an agentic skill may independently satisfy the requirements. For another, a human skill may provide the better fit. A complex task may require several configurations across multiple actions, with human and agentic skills repeatedly interacting as the task progresses.

The configuration is therefore dynamic rather than permanent. It may change when the action changes, new theoretical constraints are identified, an agentic result fails to satisfy expected mathematical properties, or human interpretation identifies a violation of model assumptions.

Throughout this process, the statistician remains the professional responsible for directing and controlling the work associated with the occupation and its responsibilities. Agentic AI provides powerful complementary capabilities that expand what the professional can simulate, derive, calculate, sample, optimize, visualize, and document.

1.3.3. Task Performance Evaluation and Skill Space Adaptation

The performance produced through a selected skill configuration is evaluated against predefined evaluation metrics. These metrics determine whether the human skills, agentic skills, or their integrated configuration are sufficiently capable of performing the relevant task and its required actions.

Evaluation criteria depend on the work being performed. In statistics, they include mathematical correctness, estimator unbiasedness, asymptotic consistency, computational efficiency, model identifiability, coverage probability of confidence intervals, reproducibility, and adherence to experimental protocols.

When a selected skill or skill configuration produces performance above the defined threshold, it passes the performance evaluation and is retained for future applications under appropriate conditions. Repeated successful use provides additional evidence about where that skill or configuration performs effectively.

When performance falls below the required threshold, the corresponding skill or configuration is not reused without modification. The relevant human or agentic skills may be refined or reconfigured, existing skills can be combined differently, or new human and agentic skills may be developed or constructed from scratch when the existing skill space lacks adequate capability.

For example, an AI agent first generates code to fit a complex hierarchical Bayesian model, but the resulting MCMC chains repeatedly fail to converge due to highly correlated posterior geometries. Task performance evaluation identifies this computational failure as falling below the required threshold. The agentic skill is then refined through the integration of advanced Hamiltonian Monte Carlo algorithms, automated reparameterization instructions, or stricter convergence diagnostics. The revised skill is evaluated again, and once it satisfies the required performance criteria, it is retained for appropriate future modeling tasks. Human skills follow a corresponding development process: when a statistician lacks the theoretical background required to evaluate a newly introduced topological data analysis method, the relevant capability is developed through studying the mathematical proofs and experimenting with synthetic datasets before its performance is evaluated through subsequent work.

Refined and newly constructed skills that satisfy the required performance criteria are added to the skill space. In this way, task performance evaluation provides feedback not only for completing current work but also for continuously optimizing and expanding the available skill space.

The skill space is not constructed as an isolated collection of occupation-specific representations. Many human skills and agentic skills used in statistics are reusable across occupations. Mathematical reasoning, probability theory, code generation, simulation, literature retrieval, and documentation contribute to work in data science, econometrics, epidemiology, psychometrics, operations research, and many other fields.

Reusable skills can therefore be represented once and configured differently according to the responsibilities, tasks, actions, and contexts of different occupations. This allows the skill space to serve a growing range of occupations without unnecessarily duplicating or exploding the number of skill representations. Occupational specialization is introduced through task context, action requirements, parameters, data sources, tools, constraints, and skill configurations rather than by creating a completely separate skill representation whenever the same underlying capability appears in another occupation.

The resulting process is adaptive: occupational responsibility → task and subtask → actions → skill requirements → candidate human and agentic skills → skill configuration → task execution → task performance evaluation → skill refinement, reconfiguration, or construction → optimized and expanded skill space

This cycle allows human and agentic capabilities to evolve alongside the requirements of statistical work.

1.4. Essential Responsibilities of Statistician

1.4.1. Experimental Design and Sampling Methodology

Experimental design and sampling methodology establishes the mathematical and logistical framework required to collect data that supports valid statistical inference. This responsibility includes defining the primary estimand, determining sample sizes through power analysis, designing randomized controlled trials or stratified sampling schemes, and mitigating selection bias.

Agentic skills contribute extensively to automated power calculations, simulation of sampling distributions, literature review on design protocols, and generation of randomization sequences. Human skills remain essential for defining the practical constraints of the study, ensuring the ethical and logical validity of the randomization mechanism, and determining whether the proposed design adequately controls for expected attrition and intra-cluster correlations.

For example, a research team plans to evaluate a new organizational intervention. The statistician defines the primary estimand and determines the necessary sample size to achieve adequate statistical power. An AI agent generates the mathematical formulas for complex stratified sampling designs, simulates various allocation strategies, and drafts the randomization code. The statistician reviews the simulations to ensure they account for expected attrition rates and structural dependencies, revising the allocation ratios to maintain mathematical validity. Finally, the statistician authorizes the finalized experimental protocol.

1.4.2. Statistical Modeling and Methodological Development

Statistical modeling and methodological development involves specifying mathematical frameworks to represent data-generating processes, fitting complex models, and developing novel estimators when standard methods are inadequate. This responsibility includes selecting likelihood functions, defining prior distributions, evaluating model identifiability, and verifying asymptotic properties.

Agentic skills assist with symbolic derivation, code scaffolding for probabilistic programming languages, MCMC chain diagnostics, and automated formula manipulation. Human skills are especially important for selecting the appropriate mathematical structure based on domain knowledge, evaluating the epistemological implications of prior assumptions, and diagnosing structural flaws in the model specification.

For example, a statistician needs to model hierarchical data with missing observations. An AI agent scaffolds the probabilistic programming code, derives the conditional posterior distributions, and runs initial MCMC diagnostics. The statistician evaluates the chain convergence, identifies that the chosen prior distributions are overly informative for a specific variance component, and revises the mathematical specification. The agent then re-executes the sampling algorithm and generates updated diagnostic plots, which the statistician verifies for mathematical consistency.

1.4.3. Hypothesis Testing and Causal Inference

Hypothesis testing and causal inference evaluates whether observed data provides sufficient evidence to support specific scientific claims or estimate causal effects, particularly in the presence of confounding variables. This responsibility includes formulating causal graphs, identifying adjustment sets, executing propensity score matching, and conducting sensitivity analyses for unmeasured confounding.

Agentic skills retrieve methodological references, calculate propensity scores, check covariate balance, execute bootstrap resampling, and automate sensitivity analysis routines. Human skills evaluate the plausibility of causal assumptions (such as exchangeability and positivity), identify unmeasured confounders based on domain theory, and interpret the practical significance of the estimated effects.

For example, an organization wants to estimate the causal effect of a policy change using observational data. The statistician formulates the causal directed acyclic graph (DAG) and identifies the necessary adjustment sets to control for confounding. An AI agent then calculates propensity scores, checks covariate balance across treatment groups, and executes bootstrap resampling to estimate confidence intervals. The statistician reviews the balance diagnostics, identifies a remaining structural imbalance, and directs the agent to apply a doubly robust estimation method. Finally, the statistician evaluates the final causal estimate and its sensitivity to unmeasured variables.

1.4.4. Survey Design and Measurement Analysis

Survey design and measurement analysis ensures that data collected through questionnaires or observational instruments accurately reflects the underlying theoretical constructs while minimizing measurement error and non-response bias. This responsibility includes cognitive pre-testing, evaluating construct validity, designing missing data imputation strategies, and calculating survey weights.

Agentic skills analyze historical response patterns, simulate potential non-response biases across demographic groups, propose automated weighting adjustments, and execute multiple imputation algorithms. Human skills define the theoretical constructs to be measured, design question structures to minimize cognitive bias, and evaluate whether the mathematical adjustments preserve the substantive meaning of the survey responses.

For example, a consortium is designing a large-scale population survey. The statistician defines the theoretical constructs to be measured and designs question structures to minimize cognitive bias. An AI agent analyzes historical response patterns, simulates potential non-response biases across demographic groups, and proposes automated weighting adjustments. The statistician reviews the proposed weights and identifies that adjusting for a specific geographic variable inadvertently masks a crucial socioeconomic disparity. The statistician revises the weighting constraints, and the agent recalculates the final survey weights to ensure both mathematical representativeness and sociological validity.

1.4.5. Statistical Computing and Simulation

Statistical computing and simulation involves writing efficient algorithms, verifying the mathematical properties of novel estimators through Monte Carlo trials, and optimizing computational performance for high-dimensional data. This responsibility includes defining data-generating processes, profiling algorithmic complexity, and managing distributed computing environments.

Agentic skills write simulation loops, execute Monte Carlo trials across parallel computing environments, profile code performance, and identify memory bottlenecks. Human skills define the theoretical data-generating processes, interpret edge-case failures in the estimator, and determine whether computational shortcuts violate fundamental statistical properties.

For example, a statistician is developing a novel estimator for high-dimensional, sparse data. The statistician defines the asymptotic properties the estimator must satisfy and designs the theoretical data-generating process. An AI agent writes the simulation loops, executes the Monte Carlo trials across distributed computing environments, and profiles the algorithmic complexity. The agent flags that the estimator fails to converge in a specific sparsity regime. The statistician investigates the mathematical proof, identifies a violated regularity condition in that regime, and revises the estimator’s penalty term. The agent then re-runs the simulations to verify the mathematical correction.

1.4.6. Analytical Reporting and Methodological Communication

Analytical reporting and methodological communication translates complex mathematical concepts, uncertainty quantifications, and methodological limitations into accessible formats for researchers, policymakers, and peer reviewers. This responsibility includes drafting methodology appendices, preventing the misinterpretation of statistical significance, and ensuring the reproducibility of the analytical workflow.

Agentic skills draft methodology sections, format complex mathematical tables, generate posterior density plots, and produce plain-language summaries of credible intervals. Human skills ensure that the distinction between statistical significance and practical significance is clearly communicated, prevent the overstatement of causal claims, and guarantee that the limitations of the statistical model are transparently reported.

For example, a statistician must present the results of a complex Bayesian analysis to non-technical stakeholders. An AI agent drafts the methodology appendix, formats the posterior density plots, and generates plain-language summaries of the credible intervals. The statistician reviews the materials and identifies that the automated summary conflates a Bayesian credible interval with a frequentist confidence interval, a common misinterpretation that could mislead decision-makers. The statistician rewrites the narrative to accurately reflect the probabilistic interpretation of the parameters and ensures the final report transparently communicates the model’s structural assumptions.

1.5. Changing Human and Agentic Involvement Across Statistician Responsibilities

Human skills and agentic skills are involved throughout statistical practice, but their relative involvement differs across responsibilities and changes during task execution.

These involvement levels are not permanent allocations. A single responsibility may contain actions where human and agentic skills compete as alternative capabilities and other actions where they collaborate sequentially, iteratively, or recursively.

The practical objective is to construct skill configurations that use the most suitable available capabilities for each part of the work and revise those configurations when task performance evaluation shows that another arrangement performs more effectively or reliably.

1.6. Working as a Statistician with Agentic AI

Agentic AI is becoming an increasingly powerful component of statistical practice. Statisticians should therefore adopt and integrate AI agents as complementary participants in professional workflows rather than treating them only as occasional tools for generating code, formatting tables, or running simulations.

This shift does not reduce the importance of mathematical and epistemological expertise. Human professionals remain responsible for the occupation and its corresponding responsibilities. They define the estimands, formulate causal assumptions, evaluate model identifiability, interpret uncertainty, communicate methodological limitations, control consequential analytical decisions, evaluate task performance, and decide how human and agentic skills are configured and improved.

At the same time, agentic skills greatly expand the volume and complexity of mathematical work that can be executed. AI agents derive symbolic formulas, scaffold probabilistic code, execute millions of Monte Carlo simulations, calculate high-dimensional propensity scores, profile algorithmic complexity, and draft extensive methodology documentation at a scale difficult to achieve through direct human effort alone.

Strong statistical practice is therefore not based on maximizing human work or agentic work independently. It is based on finding appropriate configurations of human skills and agentic skills for the responsibilities, tasks, subtasks, and actions being performed.

Where a human skill and an agentic skill can independently perform the same action, skills competition identifies which is better suited to the current context. Where work requires complementary capabilities, skills collaboration results in human and agentic skills being activated sequentially, iteratively, or recursively to achieve stronger task performance than either skill type achieves alone.

Task performance evaluation then determines whether selected skills and configurations satisfy predefined expectations. Successful skills are retained and reused, while inadequate skills are refined, reconfigured, or newly developed. The resulting capabilities return to an increasingly useful skill space that supports statistics and, where capabilities are reusable, other occupations as well.

For a statistician, working effectively with agentic AI means treating AI as a powerful complementary capability integrated into everyday professional routines while maintaining human control of mathematical validity and occupational responsibilities. Through appropriate skill selection, competition, collaboration, evaluation, and continuous improvement, human professionals and agentic AI work together to achieve statistical practice that is more efficient, scalable, mathematically rigorous, and fundamentally reliable.