Practical AI Training
Hands-on workshops where your team learns to give an AI tool the context it needs, on tasks they actually do, and to verify the result before anyone relies on it.
What a better request changes.
Add each part of a request and watch the answer change.
Your request: Summarise this document.
This document gives an overview of several important policy areas, including information security, suppliers and remote working. It makes a number of recommendations the organisation may wish to consider going forward.
Most breaches start with stolen passwords.Not from your document
The changes should cost about £40,000.Invented
Add a part to see what it changes.
Built around each role’s real work.
The same job by hand, and with AI plus human verification.
Check whether anyone signed in with a legacy protocol after it was blocked.
Typical time
We practiseProving a query on a known case before trusting its results.
SigninLogs
| where TimeGenerated > ago(7d)
| where ClientAppUsed in ("Exchange ActiveSync", "IMAP4", "POP3", "Authenticated SMTP", "Other clients")
| where ResultType == "0"
| summarize Attempts = count() by UserPrincipalName, ClientAppUsed, IPAddress
| order by Attempts descThe usual way
- Search the internal wiki for which workspace holds sign-in logs20 min
- Dig table and field names out of old runbooks25 min
- Read the KQL documentation for the operators you need30 min
- Write the query, fix the syntax errors, run it again35 min
- Check the results and write them up10 min
With AI + human verification
- AI Ask the team assistant, which already holds your log sources and runbooks1 min
- AI It checks Microsoft’s KQL reference and drafts the query first time2 min
- You Read the query, run it against a known test account, confirm the results12 min
Fictional examples. Typical times, including human verification; in a workshop we time your own tasks.
A sample half-day.
Mostly hands-on, with your own tasks, on site or remote.
- Also as
- Leadership briefing90 min
- Drop-in clinicsFollow-ups
10:00 to 10:45Audience, source, task, format and limits, practised live.
- Hands-on 2 h
- Talk and demo 1 h 15 min
Human verification, in practice.
This AI draft reads well. Check each claim against the source.
AI draft
Policy review: summary for the board
0 of 4 checked
Source
Pick a claim to see what the source says.
What your team keeps.
A workshop only matters if people still do it the week after.
- A half-day workshop on your own tasks
- A one-page guide for your tools and rules
- A shortlist of tasks to hand to AI next
Your team’s AI guide
Every request covers
- Audience
- Source
- Task
- Format
- Limits
Before you use the result
- Does every claim trace to a source?
- Did you recalculate the numbers?
- Would you sign your name to it?
What goes where
| Public | Any approved tool |
| Internal | Company tool only |
| Personal data | Only where policy allows |
Why have a security team do this?
“Can we let staff use these tools without leaking anything?”
Every session covers what can go into which tool and what a connected tool can see, taught by people who handle data protection for a living.
- Data questions firstWhat could leak is settled before switch-on.
- Access set up properlyWe configure and test it ourselves.
- Verified with your peopleSessions in plain English, not slide decks.