Every slide changes as you pick. To keep the colors, copy the values into colors at the top of the file.
BackgroundBehind every slide
Darker to lighter
MainHeadings, buttons, charts
Darker to lighter
AccentHighlights and contrast
Darker to lighter
Title and text
[Client]
× COPR
Body text in the ink color, with dimmed text beside it.
Buttons and tags
Accent text, like “Typical reply”
Chat
Where is [product] made?
[Product] is made in [country].
Charts
× COPR
Less than half of today is slides. The rest is for trying it and talking.
Built together, from the kickoff to today.
Round the table: your name, your role, and your experience with AI so far, the good and the bad.
At the kickoff we picked three areas.
Details for every product in the range
“
Product information
Where is product made, and how much does it weigh?
Answered by Product information 1A
Customers, products, quarterly trends
“
Sales statistics
What was customer’s revenue last quarter?
Answered by Sales statistics 2A
Return policies in customer agreements
“
Terms & conditions
How often is customer allowed to return products?
Answered by Terms & conditions 3A
At the kickoff, we agreed on two must-haves.
You’re the judges. Test both in the demo, and think about what it would take for you to rely on it every day.
The same question, asked two ways at the same moment.
Pick a question
Meanwhile, by email
The office answers one email at a time. We’ll leave them running in the background and see when the replies come in.
Ask about products, sales or terms.
A customer asks something you can’t answer from memory. Here’s how long it takes to find out.
Today
Email someone at the office, then wait until they have time to look it up and reply.
With the assistant
Ask on your phone and read the answer, with the customer still there.
That’s 40 to 250 times faster, and the conversation never has to stop.
Every employee can ask at the same time, and every answer still arrives in seconds.
Illustration. Each card is one employee asking a question.
Follow one question from the phone to [Client]’s data, and back.
Overview
We don’t let the AI search around on its own. We give it exact tools, and it answers only from what they return.
Letting it look around
“[Product] is made in [country A] and weighs [weight A].”Sounds certain, but it’s wrong
Giving it tools
“[Product] is made in [country B] and weighs [weight B].”Source Product information 1A
That’s what grounding means: the AI only puts into words the facts a tool hands it.
Enough slides.
After the break, it’s your turn.
XX questions out of the box. Outside them, it says so instead of guessing.
Follow-ups build on earlier answers and combine tools.
Example conversation, with the figures hidden.
Three parts make up the monthly cost.
The same three layers, month after month.
for 20 employees, 5 questions a day
The AI model is 22% of the bill, and the only part that grows with more users.
Before anyone asks, the data is collected, cleaned and combined.
Overview
Four layers around [Client]’s data.
Why we’re doing this
A customer asks two employees the same thing: “What does our agreement say about returning products that don’t sell?”
Another company’s employee
“Good question. Let me check with the office and get back to you.”
Answer later today, or tomorrow
[Client]’s employee
“[The answer, straight from the agreement]. Want me to check how [a product group] sold last quarter?”
Answer in about 7 seconds
“[Client] answered on the spot. Why do I always have to wait for the others?”
Same on the phone: no more “I’ll call you back.”For the customer
An answer on the spot, in person or on the phone.
For [Client]’s own team
Fewer lookup emails to the office, and more time for the work that matters.
A first draft of the road ahead, to shape together.
The fastest way to a great tool is real use and quick improvements.
What we improve, release by release
More tools
The experience
Tone and style
Small changes that make a big difference once more people use it.
Feedback on every answer
Workshops and guides
Personal to each employee
Three candidates from the kickoff, and room for new ideas.
[Name A]
[Why this name fits]
[Name B]
[Why this name fits]
[Name C]
[Why this name fits]
Room for a suggestion.
Thank you! That’s the presentation.
Back at --:--
Next up: the demo, and a look behind the curtain.
That’s all
This was an example presentation.