---
id: make-ai-stick
type: deck
title: What it takes to make AI stick
summary: >-
  Most companies stall at a few licences and one enthusiast. This full-day
  workshop covers the five things that make AI part of everyday work, from
  picking two or three workflows to tracking hours saved, with exercises and a
  30-day plan.
language: en
date: '2026-10-03'
speakers:
  - Nik Redl
  - Thomas Schlossmacher
publisher: Specialty Tokens
audience: >-
  Leadership teams and AI champions at companies that have bought AI licences
  but see little change in daily work.
topics:
  - ai adoption
  - ai rollout
  - ai champions
  - ai policy
  - change management
  - ai workflows
page: 'https://specialtytokens.com/make-ai-stick'
markdown: 'https://specialtytokens.com/resources/make-ai-stick.md'
slides: 'https://specialtytokens.com/slides/s/make-ai-stick?p=1'
slide_count: 37
access: The slides open after a short form on the page.
---

# What it takes to make AI stick

> Most companies stall at a few licences and one enthusiast. This full-day workshop covers the five things that make AI part of everyday work, from picking two or three workflows to tracking hours saved, with exercises and a 30-day plan.

## Key points

- Most companies rolling out AI have a few licences, one enthusiast and no system, according to Specialty Tokens' sales and discovery calls with those companies. (slide 5)
- Five things make AI stick in a company: pick two or three workflows, make it more than one person's job, fix the data AI reads, agree the rules with IT, and get from prototype to daily use. (slide 7)
- A good first AI workflow happens every week, is dreaded by someone on the team, and has a result you can check in a minute; boring and frequent beats impressive. (slide 10)
- AI champions should be curious people who already use AI on their own, one from each team that owns a chosen workflow, with two hours a week protected in the calendar; pick for curiosity, not seniority. (slide 15)
- One client's AI ambassador built 23 shared AI skills that the whole team now uses. (slide 18)
- To fix the data AI reads, companies should clean only the system their first workflow reads, give every important field one owner, and connect AI to the source rather than to copies on laptops. (slide 22)
- A one-page AI policy written with IT should state which tools are approved (for example ChatGPT Business or Claude Team), what data may go in (public, internal or never), and where builds live and who sees them. (slide 26)
- An AI build is done only when it passes three tests: it runs without the person who built it, it lives in a shared place rather than on a laptop, and someone owns fixing it when it breaks. (slide 30)
- Companies making AI stick should track three numbers every month: hours saved per week, workflows in daily use, and people using them weekly. (slide 33)
- The first 30 days of an AI rollout run in four weeks: pick three workflows and name three champions, clean the data source and sign off the policy with IT, let champions build in pairs in a shared space, then demo to leadership with the three numbers. (slide 34)

## What's inside

1. Pick two or three workflows
2. Make it more than one person's job
3. Fix the data it reads
4. Agree the rules with IT
5. Get from prototype to daily use
6. Your first 30 days

## The written edition

Most companies we talk to are in the same place: a few licences, one enthusiast, and no system. This workshop is built from our sales and discovery calls this year with companies rolling out AI. The same five needs came up in almost every call, and every quote below is real; we leave out names and companies.

### 1. Pick two or three workflows

> "These are general ideas, but it's not like we focused all the resources on this. It's more like here a side project, a side project."

Choose a few pieces of real work that are worth doing properly, instead of a dozen side projects. A good first workflow happens every week, not once a year; someone on the team dreads doing it; and you can check the result in a minute. Boring and frequent beats impressive.

**Exercise: list ten, keep three.** Write down ten tasks your team repeats every week. Score each on time spent, pain, and how easy the result is to check. Keep the top three and park the rest.

### 2. Make it more than one person's job

> "I'm somehow like a one-man army to enable this and to plan all of this."

Give a team a tool and it works for a week. Teach a champion to build, and the team keeps building. Champions are curious people who already use AI on their own, one from each team that owns one of your three workflows, with two hours a week protected in the calendar. Pick for curiosity, not seniority.

> "In the training they are all very, yeah, that's a cool thing. And then day-to-day business takes over again."

A training is a day; a habit is a calendar. Three things keep it going: a weekly build hour (same time, same room), one shared library everyone can use, and a monthly show and tell for leadership. One client's AI ambassador built 23 shared skills that the whole team now uses.

**Exercise: name your champions.** One champion for each of your three workflows, their two hours a week blocked from next week, and the first show and tell in the calendar.

### 3. Fix the data it reads

AI works with what your systems hold. One client migrated 70,000 contacts into their CRM, most of them now dead. You don't need to clean everything: clean only the system your first workflow reads, give every important field one owner, and connect AI to the source, not to copies on laptops. Start small, but start at the source.

**Exercise: map the data.** For each workflow: where does its data live, who owns it and keeps it clean, and can AI reach it today? If not, what is missing?

### 4. Agree the rules with IT

> "We have two plans, but the people started using shadow AI."
>
> "IT have tried to kill the project, obviously because of data concerns."

Decide which tools, which data and which places are fine, so people can build without asking every time. Write the rules on one page: which tools are approved (for example ChatGPT Business or Claude Team); what data may go in (public, internal, never); and where builds live and who sees them (one shared space, not laptops). Write it with IT, not around IT.

**Exercise: draft your one-page policy.** List the tools you already pay for or allow, sort your data into public, internal and never, and agree where builds live. Take the draft to IT this week.

### 5. Get from prototype to daily use

> "Everything they are building is local files on their laptops. When they are on vacation, nothing is running because nobody has access to it."
>
> "Prototyping into production, that's the most difficult thing."

A demo impresses once; a tool saves time every week. Before you call something done, it passes three tests: it runs without the person who built it, it lives in a shared place rather than on a laptop, and someone owns fixing it when it breaks.

> "This whole measurement and outcome thing is also really important in the end for our management. They want to see results."

Track three numbers every month: hours saved per week (ask the people doing the work), workflows in daily use (actually used, not just built), and people using them weekly (across teams, not just champions). Demos get applause; numbers get budget.

### Your first 30 days

- **Week 1, pick and name:** three workflows, three champions, time blocked.
- **Week 2, data and rules:** clean the source; policy signed off with IT.
- **Week 3, build:** champions build in pairs, in the shared space.
- **Week 4, show:** demo to leadership with the three numbers.

Every plan ends in front of leadership. Two or three things, rolled out properly.

## The slides

### Slide 1: What it takes to make AI stick

FULL-DAY WORKSHOP
What it takes to make AI stick
Five things the companies we work with say they need to do, in their own words.
SPECIALTY TOKENS

### Slide 2: Your workshop leaders

Your workshop leaders
Nik, left · Thomas, right
Nik Redl
Director · Engineer, operator and AI advisor
Built Mokker past one million users before its acquisition by Soona
Former Soona Chief of Staff; now advises companies across AI
Thomas Schlossmacher
Director · Product-focused founder and educator
Two-time founder; built Resonant past 25 million dictations
Teaches AI and tech online to more than 75,000 subscribers
YOUR WORKSHOP LEADERS

### Slide 3: Where today comes from

Where today comes from
Our sales and discovery calls this year, with companies rolling out AI
Every quote you will see is real; we leave out names and companies
The same five needs came up in almost every call
Their words, not ours.

### Slide 4: Pick two or three workflows

Today
01
09.30
Pick two or three workflows
Exercise: list ten, keep three
02
10.45
Make it more than one person's job
Exercise: name your champions
03
13.00
Fix the data it reads
Exercise: map the data
04
14.15
Agree the rules with IT
Exercise: a one-page policy
05
15.30
From prototype to daily use
Exercise: your 30-day plan
LUNCH 12.15 · CLOSE 17.00
AGENDA

### Slide 5: A few licences, one enthusiast, and no system.

WHERE MOST COMPANIES ARE
A few licences, one enthusiast, and no system.

### Slide 6: Does this sound familiar?

Does this sound familiar?
“We just want to start.”
“You can't be the only owner of this.”
“Nothing is shareable in the end.”
“The main issue is also the data protection.”

### Slide 7: Five things that make AI stick

Five things that make AI stick
01
Pick two or three workflows
Not twenty side projects
02
Make it more than one person's job
Champions with real time
03
Fix the data it reads
AI is only as good as its inputs
04
Agree the rules with IT
Before shadow AI does it for you
05
Get from prototype to daily use
And show it to leadership

### Slide 8: Pick two or three workflows

PART 1 OF 5
09.30 TO 10.30
Pick two or three workflows
Choose a few pieces of real work that are worth doing properly, instead of a dozen side projects.

### Slide 9: “These are general ideas, but it's not like we focused all the resources on this. It's…

WHAT WE HEARD
“These are general ideas, but it's not like we focused all the resources on this. It's more like here a side project, a side project.”
A client, in a discovery call
PART 1 · PICK TWO OR THREE WORKFLOWS

### Slide 10: A good first workflow

A good first workflow
1
Happens every week, not once a year
2
Someone on the team dreads doing it
3
You can check the result in a minute
Boring and frequent beats impressive.
PART 1 · PICK TWO OR THREE WORKFLOWS

### Slide 11: List ten, keep three

EXERCISE 1
25 MIN
List ten, keep three
1
Write down ten tasks your team repeats every week
2
Score each on time spent, pain and how easy it is to check
3
Keep the top three; park the rest
YOU LEAVE WITH: YOUR THREE WORKFLOWS
PART 1 · PICK TWO OR THREE WORKFLOWS

### Slide 12: Make it more than one person's job

PART 2 OF 5
10.45 TO 12.15
Make it more than one person's job
Give named champions the time and the place to build, so the work survives holidays and busy weeks.

### Slide 13: “I'm somehow like a one-man army to enable this and to plan all of this.”

WHAT WE HEARD
“I'm somehow like a one-man army to enable this and to plan all of this.”
A client, in a discovery call
PART 2 · MAKE IT MORE THAN ONE PERSON'S JOB

### Slide 14: Give a team a tool,

Give a team a tool,
and it works for a week.
Teach a champion to build,
and the team keeps building.
PART 2 · MAKE IT MORE THAN ONE PERSON'S JOB

### Slide 15: Who should carry it

Who should carry it
Curious people who already use AI on their own
One from each team that owns one of your three workflows
Two hours a week, protected in the calendar
Pick for curiosity, not seniority.
PART 2 · MAKE IT MORE THAN ONE PERSON'S JOB

### Slide 16: “In the training they are all very, yeah, that's a cool thing. And then day-to-day…

WHAT WE HEARD
“In the training they are all very, yeah, that's a cool thing. And then day-to-day business takes over again.”
A client, in a discovery call
PART 2 · MAKE IT MORE THAN ONE PERSON'S JOB

### Slide 17: What keeps it going after today

What keeps it going after today
1
A weekly build hour, same time, same room
2
One shared library everyone can use
3
A monthly show and tell for leadership
A training is a day. A habit is a calendar.
PART 2 · MAKE IT MORE THAN ONE PERSON'S JOB

### Slide 18: shared skills built by one client's AI ambassador, used by the whole team.

23
shared skills built by one client's AI ambassador, used by the whole team.
PART 2 · MAKE IT MORE THAN ONE PERSON'S JOB

### Slide 19: Name your champions

EXERCISE 2
20 MIN
Name your champions
1
One champion for each of your three workflows
2
Block their two hours a week, starting next week
3
Put the first show and tell in the calendar
YOU LEAVE WITH: NAMES AND DATES
PART 2 · MAKE IT MORE THAN ONE PERSON'S JOB

### Slide 20: Fix the data it reads

PART 3 OF 5
13.00 TO 14.00
Fix the data it reads
AI works with what your systems hold. Clean the part your first workflow needs, and let AI reach it.

### Slide 21: contacts one client migrated into their CRM, most of them now dead.

70,000
contacts one client migrated into their CRM, most of them now dead.
PART 3 · FIX THE DATA IT READS

### Slide 22: AI reads what you already have

AI reads what you already have
Clean only the system your first workflow reads
Give every important field one owner
Connect AI to the source, not to copies on laptops
Start small, but start at the source.
PART 3 · FIX THE DATA IT READS

### Slide 23: Map the data

EXERCISE 3
20 MIN
Map the data
1
Where does each workflow's data live?
2
Who owns it and keeps it clean?
3
Can AI reach it today? If not, what is missing?
YOU LEAVE WITH: A DATA MAP FOR YOUR THREE WORKFLOWS
PART 3 · FIX THE DATA IT READS

### Slide 24: Agree the rules with IT

PART 4 OF 5
14.15 TO 15.15
Agree the rules with IT
Decide which tools, which data and which places are fine, so people can build without asking every time.

### Slide 25: “We have two plans, but the people started using shadow AI.”

SHADOW AI
“We have two plans, but the people started using shadow AI.”
A BLOCKED PROJECT
“IT have tried to kill the project, obviously because of data concerns.”
PART 4 · AGREE THE RULES WITH IT

### Slide 26: Write the rules on one page

Write the rules on one page
01
Which tools are approved
e.g. ChatGPT Business, Claude Team
02
What data may go in
public · internal · never
03
Where builds live, who sees them
one shared space, not laptops
Write it with IT, not around IT.
PART 4 · AGREE THE RULES WITH IT

### Slide 27: Draft your one-page policy

EXERCISE 4
25 MIN
Draft your one-page policy
1
List the tools you already pay for or allow
2
Sort your data into public, internal and never
3
Agree where builds live and who can see them
YOU LEAVE WITH: A DRAFT TO TAKE TO IT THIS WEEK
PART 4 · AGREE THE RULES WITH IT

### Slide 28: From prototype to daily use

PART 5 OF 5
15.30 TO 17.00
From prototype to daily use
Make what your champions build run without them, measure it, and show it to leadership.

### Slide 29: “Everything they are building is local files on their laptops. When they are on…

WHAT WE HEARD
“Everything they are building is local files on their laptops. When they are on vacation, nothing is running because nobody has access to it.”
A client, in a discovery call
PART 5 · FROM PROTOTYPE TO DAILY USE

### Slide 30: Three tests before you call it done

Three tests before you call it done
1
It runs without the person who built it
2
It lives in a shared place, not on a laptop
3
Someone owns fixing it when it breaks
A demo impresses once. A tool saves time every week.
PART 5 · FROM PROTOTYPE TO DAILY USE

### Slide 31: “Prototyping into production, that's the most difficult thing.”

WHAT WE HEARD
“Prototyping into production, that's the most difficult thing.”
A client, in a discovery call
PART 5 · FROM PROTOTYPE TO DAILY USE

### Slide 32: “This whole measurement and outcome thing is also really important in the end for our…

WHAT WE HEARD
“This whole measurement and outcome thing is also really important in the end for our management. They want to see results.”
A client, in a discovery call
PART 5 · FROM PROTOTYPE TO DAILY USE

### Slide 33: Track three numbers, every month

Track three numbers, every month
01
Hours saved per week
Ask the people doing the work
02
Workflows in daily use
Not built: actually used
03
People using them weekly
Across teams, not just champions
Demos get applause. Numbers get budget.
PART 5 · FROM PROTOTYPE TO DAILY USE

### Slide 34: Your first 30 days

Your first 30 days
WEEK 1
Pick and name
Three workflows, three champions, time blocked
WEEK 2
Data and rules
Clean the source; policy signed off with IT
WEEK 3
Build
Champions build in pairs, in the shared space
WEEK 4
Show
Demo to leadership with the three numbers
Every plan ends in front of leadership.
PART 5 · FROM PROTOTYPE TO DAILY USE

### Slide 35: Write your 30-day plan

EXERCISE 5
30 MIN
Write your 30-day plan
1
Put a name and a date on every week
2
Pick the day you show leadership
3
Present it to the room in two minutes
YOU LEAVE WITH: ONE PLAN PER COMPANY, SHARED TONIGHT
PART 5 · FROM PROTOTYPE TO DAILY USE

### Slide 36: Five things to take home

Five things to take home
01
Pick two or three workflows
Boring and frequent
02
Make it more than one person's job
Champions with real time
03
Fix the data it reads
Start at the source
04
Agree the rules with IT
One page, written together
05
Get from prototype to daily use
Measure it, show it

### Slide 37: Two or three things, rolled out properly.

Two or three things, rolled out properly.
Questions about your rollout? Talk to us at specialtytokens.com/contact
NIK REDL · THOMAS SCHLOSSMACHER
SPECIALTY TOKENS
