Applied Digitalisation · Module 2

Analytics, AI, and control loops

Collecting data is the easy part.

36 min 4 outcomes Digitalisation Intermediate

Previously

Data pipelines and flows

A pipeline is only valuable when each step is owned and tested.

This module

Analytics, AI, and control loops

Collecting data is the easy part.

Next

APIs and system integration

APIs are the contracts that keep systems aligned.

Progress

Mark this module complete when you can explain it without rereading every paragraph.

Why this matters

A team builds an automated rule.

What you will be able to do

  • 1 Explain analytics, ai, and control loops in your own words and apply it to a realistic scenario.
  • 2 Analytics becomes powerful when it closes a loop: measure, decide, act, measure again.
  • 3 Check the assumption "Measures drive decisions" and explain what changes if it is false.
  • 4 Check the assumption "Loop owners exist" and explain what changes if it is false.

Before you begin

  • Foundations-level vocabulary and concepts
  • Confidence with basic diagrams and section terminology

Common ways people get this wrong

  • Metric gaming. When metrics become targets, behaviour changes. Design for that.
  • No response path. If nobody acts on signals, observability is decoration.

Main idea at a glance

Sense, interpret, act

A digitalised loop

Stage 1

Sense

Real-time data from sensors, meters, user events, and system logs. This is the input that triggers the rest of the loop. The quality and timeliness of sensing determines how well the loop performs.

Sensing is only as good as your instrumentation. If you are measuring the wrong thing, or measuring the right thing at the wrong frequency, the entire loop will make bad decisions confidently.

The loop must be stable. Without damping and safeguards, automated control can oscillate and cause more harm than the problem it was solving.

Collecting data is the easy part. The hard part is turning it into action safely. Analytics and AI help you detect patterns and predict outcomes. Control loops are how the system responds.

Worked example. A smart rule that caused oscillation

Worked example. A smart rule that caused oscillation

A team builds an automated rule. If demand is high, reduce load by sending a message to flexible devices. Many devices respond at the same time. Demand drops sharply. The rule then stops. Devices recover. Demand spikes again. The system starts oscillating.

My opinion is that automation without systems thinking is how you create elegant chaos. You need rate limits, damping, and measurement to prevent the control loop from fighting itself.

Common mistakes with analytics and automation

Automation risk anti-patterns

These mistakes cause fast-moving failure in production.

  1. Treating model output as a decision

    Model output should inform decisions, not replace governance logic.

  2. Ignoring drift monitoring

    Without drift checks, performance degrades quietly until harm is visible.

  3. Missing safety boundaries

    High-speed automation can amplify the wrong action at scale.

Verification. A safe automation checklist

Safe automation checklist

Use this before enabling live automated actions.

  1. Define objective and success metric

    State the outcome and the measurable signal proving it worked.

  2. Design for failure detection

    Identify likely failure modes and the earliest detection signal.

  3. Set human override authority

    Document who can halt automation and under what conditions.

  4. Rehearse rollback

    Validate rollback steps before production dependence.

Reflection prompt

Think of one automation you would not allow to run without a human approval step. Why that one.

Mental model

Control loops in practice

Analytics becomes powerful when it closes a loop: measure, decide, act, measure again.

  1. 1

    Measure

  2. 2

    Decide

  3. 3

    Act

  4. 4

    Review

Assumptions to keep in mind

  • Measures drive decisions. If measures do not drive decisions, they become noise.
  • Loop owners exist. If nobody owns the loop, it never closes.

Failure modes to notice

  • Metric gaming. When metrics become targets, behaviour changes. Design for that.
  • No response path. If nobody acts on signals, observability is decoration.

Check yourself

Quick check. Control loops and safe automation

0 of 6 opened

What is a control loop in a digital system

A cycle where the system senses what is happening, decides what to do, acts, then measures the result and adjusts.

Why can automation create oscillation

Because many components react at once, overshoot the target, and then react back again without damping or rate limits.

What is drift

When the real world changes so the patterns a model learned no longer match reality, which can quietly reduce performance.

Scenario. A model outputs a score of 0.91. Why is that not a decision

It is input to a decision. You still need thresholds, context, and safety rules for what happens on the bad day.

Name one guardrail that makes automation safer

Rate limits, human approval for high impact actions, anomaly detection, or a hard stop when confidence is low.

Why is a rollback plan part of automation design

Because automation can scale mistakes. You need a fast way to stop and revert when outcomes move in the wrong direction.

Artefact and reflection

Artefact

A one-page decision note with assumption, evidence, and chosen action

Reflection

Where in your work would explain analytics, ai, and control loops in your own words and apply it to a realistic scenario. change a decision, and what evidence would make you trust that change?

Optional practice

Work through one scenario and justify the decision with evidence

Source GOV.UK Service Standard points 13 and 14
Source ISO/IEC 38500:2024 governance of IT
Source Ofgem Data Best Practice Guidance
Source NESO Sector Digitalisation Plan