What You Will Be Able To Do

• Map a process from trigger to result
• Spot steps that suit AI patterns
• Score value and feasibility
• Place ideas on a matrix
• Write a one-page use case brief

Theme Versus Use Case

• Theme: 'use AI in customer service'
• Use case: draft first replies to billing emails
• A use case has a team, a task and a goal
• Without that, results never show in the numbers

Step 1: Map The Process

• Start with the trigger, end with the result
• List steps, owners, systems, time and errors
• Talk to people who do the work
• Real process differs from written procedure

Trigger: supplier invoice arrives
1. Clerk opens email (2 min)
2. Clerk keys fields (8 min)
3. Manager approves (waits 2 days)
Result: invoice paid and recorded

Four Signals Of A Promising Step

• Repetition: same work many times a week
• Text or documents: reading, writing, searching
• Waiting: queue until someone reads or sorts
• Experts decide fast, newcomers struggle

Seven AI Patterns

• Draft, summarize, classify, extract
• Search, forecast, flag
• AI prepares and suggests, a person decides
• No pattern fits: probably not a first candidate

Case: Contract Review At A Large Bank

• Reported in 2017: loan agreements read by software
• Work said to take about 360,000 hours a year
• Many similar documents, clear questions, expert handles exceptions
• Target was first reading, not final judgment

Step 2: Score Value

• Frequency times time per task times share saved
• 400 items at 10 minutes is about 67 hours a month
• Halving the time saves about 33 hours
• Also consider revenue, risk and quality

Converting Value To A Score

• Agree ranges in advance, then score 1 to 5
• Write the assumption behind each score
• Do not claim precision you do not have
• Let others challenge the numbers

Step 3: Score Feasibility

• Data: exists, usable, allowed?
• Technology: connects to where work happens?
• Risk: what if output is wrong, who checks?
• Adoption: will the team use it, is there an owner?

The Value And Feasibility Matrix

• High value, high feasibility: quick win, do first
• High value, low feasibility: bet, plan with checkpoints
• Low value, high feasibility: fill-in, only if cheap
• Low value, low feasibility: drop

Value
  ^   BETS        | QUICK WINS
  |  (invest)     | (do first)
  |---------------+-------------
  |   DROP        | FILL-INS
  +-----------------------------> Feasibility

Build A Portfolio

• Two or three quick wins show results in weeks
• One or two bets change the business over a year
• Quick wins build credibility and skill
• Only bets lose patience, only quick wins stay small

The One-Page Brief

• Process, team and problem in one sentence
• AI pattern, baseline and value estimate
• Feasibility scores with doubts noted
• Risk, who checks output, business owner

Recap

• Map the work before choosing tools
• Match steps to AI patterns
• Score value and feasibility with written assumptions
• Choose a portfolio of quick wins and bets

Next Week

• Data readiness and governance
• Where company data lives
• Quality, access and ownership
• Why feasibility scores often fall here

Questions And Exercise

• Exercise: map one process and rank three to five ideas
• Write a one-page brief for the top quick win
• Share with a business owner for feedback
• Questions from the room