Workshops on Eriks training corner

Workshop 1 – Use-case hunt

Mon, 01 Jan 0001 00:00:00 +0000

Workshop 1 – Use-case hunt

Time: 20 minutes Where: this workbook + Copilot Chat Output: 5–8 AI use-case candidates from your department

Work on your real department. If you prefer not to, use TRAXIA’s Road Systems and Smart City division. Put nothing confidential in the prompt – the activity type and volume are enough.

Step 1 – Alone: activities that eat time (8 min)

Look for activities that repeat, take a lot of time, involve lots of text, data or searching, or frustrate you. Be specific: which report, how often, how many hours, for whom.

Workshop 2 – Priorities and business case

Mon, 01 Jan 0001 00:00:00 +0000

Workshop 2 – Priorities and business case

Time: 25 minutes Where: Excel AI_ROI_calculator.xlsx + this workbook Output: prioritised use cases, a business case for the top 1, a 90-day plan

Step 1 – Fit test (5 min)

In Excel, sheet Use cases, enter 5 use cases from W1 and score six questions 0 = no, 1 = partly, 2 = yes:

#	Question
Q1	Volume: does it repeat often and take a lot of time?
Q2	Text and data: is it work with text, documents, data, searching?
Q3	Data: is the data available, current and trustworthy?
Q4	Error tolerance: is an error easy to spot and not fatal?
Q5	Baseline: can we measure today’s state?
Q6	Owner: does the process have an owner who wants change?

Score 9–12 = candidate · 6–8 = rework or wait · 0–5 = not now.

Traffic-light exercise

Mon, 01 Jan 0001 00:00:00 +0000

Traffic-light exercise

Time: 5 minutes, together Output: green, amber or red – and why

For each situation decide: green = fine · amber = only with conditions · red = no. The review ladder and the “which data in which tool” table help.

	Situation	Colour	Why / under what conditions
a	A salesperson pastes a competitor’s price list into their personal ChatGPT to prepare a comparison.
b	HR lets Copilot rank 60 technician CVs and invites the top 10 to interview.
c	A field technician asks the internal agent over the manuals what error E-217 on a level-crossing controller means.
d	Marketing generates an AI image of a “happy mayor” at a new traffic-light junction for a campaign.
e	A project manager gives Copilot Cowork the task of sending the weekly project summary to the customer automatically, unchecked.
f	A controller has Analyst recalculate project variances and checks the results in Excel before the report.

Review ladder – reminder

Level	Output	Check
1	Internal drafts	author checks it (verify, fix, approve)
2	Decision support	verify numbers and sources, disclose AI use
3	External outputs	second person + approval
4	Legal, financial, safety	AI only prepares, a human decides, four eyes
STOP	Automated decisions about people, emotion recognition	prohibited

TRAXIA case study

Mon, 01 Jan 0001 00:00:00 +0000

Case study – rolling out AI at TRAXIA

Time: 60 minutes Teams: 3 teams of 3–4 Output: a completed AI rollout canvas + a 4-minute pitch to the board

You are a management working group at TRAXIA s.r.o. (intelligent transport systems, ~220 employees, HQ in Banská Bystrica, branches in Bratislava, Žilina, Košice and Kraków). CEO Martina Kováčová wants a proposal by the end of the month on how to introduce AI in one process. The company runs Microsoft 365, and managers have Microsoft 365 Copilot licences. All names, companies and figures are fictional.