SIGNAL DESK / RESEARCH WORKSPACE Independent market, news & business research

LEARNING ACADEMY / COURSE 06

AI, data & online security

Use emerging technology responsibly and make better evidence-based decisions.

← All coursesCOURSE 06 OF 06 · 4 DETAILED LESSONS + VIDEO
Your progress0 of 4 lessons completed
LESSON 01

Find reliable information

Identify who published a claim, when it was collected and what may be missing. Prefer primary sources for financial and regulatory facts.

In depth

Reliability means knowing who produced information, how they measured it and when the observation was made. An official release is stronger evidence for an official statistic than a commentary repost, but primary sources can still need context.

Worked example / context

A rise in “mentions” does not prove a rise in sales. Check definitions, population, denominator and time interval. Distinguish correlation from an explanation supported by independent evidence.

Practical checklist: Find the first published source for one viral claim; note its date, method and limitations.

Put it into practice

Compare the same statistic from two sources and note any difference in definitions.

LESSON 02

Use AI as an assistant, not an oracle

AI can help draft, categorise and brainstorm, but may invent sources or miss context. Keep humans responsible for high-impact decisions.

In depth

Generative AI can accelerate drafting, classification and brainstorming, but can produce plausible false statements. Treat generated summaries as drafts and verify every material number, source or safety-critical instruction. Avoid exposing customer confidential information without appropriate safeguards.

Worked example / context

When using an AI-generated market summary, open the cited article and compare the exact claim. If the source is missing or unrelated, do not present the claim as verified. Maintain human approval before actions with financial consequences.

Practical checklist: Use a short verification checklist: source present, time current, numbers reproduced, uncertainty stated, human review completed.

Put it into practice

Ask an AI to summarise a public article, then verify every material claim against the original.

LESSON 03

Data privacy and cyber hygiene

Use strong unique passwords, multi-factor authentication, least-privilege access and backups. Avoid uploading customer secrets to tools unnecessarily.

In depth

Cyber hygiene starts with strong unique passwords in a reputable password manager, phishing-resistant MFA where available, prompt updates and offline or protected backups. Restrict staff access to only what they need.

Worked example / context

A shared administrator login makes auditing difficult and increases damage from compromise. Use individual accounts, recovery procedures and access reviews. Practice restoring a backup rather than assuming backups work.

Practical checklist: Enable MFA on key accounts, review administrator rights, check backup recovery and document an incident contact list.

Put it into practice

List your highest-risk accounts and check MFA and recovery methods.

LESSON 04

Build an evidence-based dashboard

Define metrics, refresh rates, source licensing and what happens when a feed fails. Transparent error states build more trust than invented results.

In depth

Every dashboard metric needs an owner, definition, provenance and refresh policy. Track the last successful data timestamp, not just when the page checked. Do not silently switch a failed provider to an unrelated estimate.

Worked example / context

If a currency feed publishes daily reference rates but your website calls it every minute, users are still seeing a daily observation. Say so. When a source fails, preserve the previous value only if it is clearly marked stale.

Practical checklist: Create a data dictionary for three metrics, including units, expected lag, licence and error state.

Put it into practice

Design three KPI cards and explain the data source, update frequency and limitation for each.

Knowledge check

What is the safest way to handle an AI-generated business claim?

Choose the most appropriate answer to review the explanation.