The Work Smarter Company

Leading in the Age of AI

This is the full reference version of the session. Every slide is here in order, from the opening hook through the adoption map to the close.

Switch back to the presentation view with the button in the corner, or press E. Use M for the menu at any time.

Colin Cox, The Work Smarter Company

Practical AI in Action for Small Businesses

What AI can do for a small business this year, shown live

The Work Smarter Company01
This week

"The people building AI earnestly believe that it could kill us all by the end of the decade."

Jacob Coxon, on resigning from Anthropic after roughly three years of pretraining research at OpenAI and then Anthropic. Posted on X, September 8, 2026

The Work Smarter Company02
The headline

Extinction estimates from researchers and the labs

0%10%50%70%100% A 2022 survey of AI researchers Median 5 percent. No timeframe asked Jacob Coxon, resigned from Anthropic About 10 percent within ten years Evan Hubinger, Anthropic Head of alignment stress testing More than 10 percent within ten years Marcus Williams, OpenAI safety team 70 percent within three years if nothing slows down International AI Safety Report 2026, 100 experts No number. Current systems lack the capability

These are personal estimates from serious people. They range from 5 to 70 percent on different timeframes

Coxon and Hubinger as quoted by TechCrunch and Time, September 2026. Williams on X, September 10, 2026. AI Impacts survey, 2022. International AI Safety Report, 2026

The headline

Recursive self-improvement

Every estimate on the last slide rests on one idea. A model good enough to build a better model, which builds a better one, until people are no longer in the loop

Model 1Trained by peopleModel 2Trained mostly by Model 1Model 3Trained by Model 2, faster BUILDS BUILDS and so on, with less human involvement each round
The headline

Past fears of new technology

about 370 BCWritingTrust in writing "will discourage theuse of their own memory."Plato, Phaedrus1895Bicycle faceDoctors warned cycling produced "awearied and exhausted bicycle face,usually flushed, but sometimes pale."The Literary Digest1906The phonographChildren who hear only phonographs will"become simply human phonographs,without soul or expression."John Philip Sousa, Appleton's Magazine1995The internet"The Internet will soon go spectacularlysupernova and in 1996 catastrophicallycollapse." In 1997 he blended the columnand ate it on stage.Robert Metcalfe, inventor of Ethernet, InfoWorld1998The internet again"By 2005 or so, it will become clearthat the Internet's impact on theeconomy has been no greater than the faxmachine's."Paul Krugman, Red Herring

None of this proves AI is safe. It shows the fear arrives before the benefit

The headline
Post by Naval on X. If AI is risky like fire is risky, then you want everyone to have it. If AI is risky like a nuclear weapon is risky, then you want no one to have it.
The Work Smarter Company06
Factfulness

Doom versus abundance

People are drawn to bad news. Hans Rosling called it the negativity instinct. The doom scenario gets the airtime and the abundance scenario gets almost none.

Doom
The machines take over

Every film, every headline, every dinner party. We have been rehearsing this one since 1984.

Abundance
Intelligence gets cheap

Cheaper intelligence, compute, energy and healthcare. That is what material abundance is made of.

Factfulness

Where abundance is already real

Breast cancer screening

Over 105,000 Swedish women in a randomised trial. AI-supported screening detected 29 percent more cancers, led to 27 percent fewer aggressive cancers later, and cut radiologists' reading work by 44 percent.

MASAI trial, Lund University, The Lancet journals 2023 to 2026

Intelligence is nearly free

Reading a year of your customer emails and telling you what they say now costs less than a coffee. Three years ago that was a week of someone's time.

When intelligence is nearly free, the question stops being what it costs and becomes what you would do with it.

Forty cents per million words read, OpenAI list price, 2025

Jevons paradox

William Stanley Jevons, 1865. Cheaper coal power did not reduce coal use. It multiplied it. The same economics now apply to your work

01Faster and cheaper02More work worth doing03More experiments04More demand
01

What small businesses are using AI for

Six things I have done or watched a client do this year. None of them needed a developer

What small businesses are using AI for

The leading AI tools

What small businesses are using AI for

Finishing the corporate year-end

The Work Smarter Company12
What small businesses are using AI for

Drafting proposals

The Work Smarter Company13
What small businesses are using AI for

Presentations. Websites. Apps

The Work Smarter Company14
What small businesses are using AI for

Acting as a pricing expert

The Work Smarter Company15
What small businesses are using AI for

Acting as a B2B sales coach

The Work Smarter Company16
What small businesses are using AI for

Preparing for a sales call

The Work Smarter Company17
02

How business leaders can use AI well

The habits behind every example you just saw

Just ask

Not sure? Ask the AI

Not sure AI can do it? Ask it.
Not sure how to do it? Ask it.

The Work Smarter Company19
How business leaders can use AI well

Treat AI like a smart new hire

Very good at some things and unreliable at others. Lean on it here and check it there.

Lean on it for

First drafts of anything. Emails, quotes, job ads, plans.

Options when you are stuck.

Summaries of the long thing you did not have time to read.

Keep your hands on

Facts, numbers and names.

Anything a customer receives.

The final decision.

How business leaders can use AI well

Your part and the AI's part

You set the task and you check the result. The AI does the middle.

First 10 to 20%Instruct
The middle 60 to 80%The work
Final 10 to 20%Check
Instruct. Say who it is, what it needs to know and what good looks like.
The work. Drafting, structuring, options, first passes.
Check. Judgment, taste, accountability. You send it.
How business leaders can use AI well

Role Context Task

A repeatable shape for a request. It makes you say up front what the AI needs, so it stops guessing.

R
Role
"You are an experienced restaurant operations manager."
C
Context
"We run two locations with 40 staff. Friday close-out takes ninety minutes and the shift leads all do it differently."
T
Task
"Draft a one-page close-out checklist my shift leads can use this Friday."
→
A request the AI can act on
Less back and forth. A usable answer on the first try.
How business leaders can use AI well

Ask AI to critique its own answer

The first draft is rarely the best one. Before you accept anything, say "What is weak here? What would a tough critic say? Now fix it."

First draftFine, and generic.
→
Critique it"What is weak here? What would a sharp critic say?"
→
Stronger draftIt rewrites with its own critique in hand.
Run the loop twice before you accept anything
Preventing AI slop

Slop grenades

Hand out AI work you have not checked and your team spends the afternoon sorting through the pieces.

SEND ???
How business leaders can use AI well

Let AI write your prompt

You do not have to write the perfect request. Describe what you want in plain words, then say this.

"Help me write a prompt that will get me the best result for this."

It will ask you the questions you forgot to answer, then write the request for you.

03

Connecting AI to other software

The answers stop being generic when the tool can see your week

Connecting AI to other software

Email, calendar, files and accounting

Once connected, the AI can read and act where you already work, with your permission. The answers stop being generic because it can see your week.

AI WITH YOUR PERMISSION EmailSort the inbox and draft replies in contextCalendarSee the week and prepare for the next meetingFilesPull facts from a real documentAccounting or CRMAsk questions of your own numbers
Connecting AI to other software

Make AI your first port of call

Before you ask a colleague, book a meeting or sit on a question, ask the AI first. This one habit separates the people who get value from the people who do not.

  • 01
    "I have a supplier meeting at two. Brief me and give me three questions to ask."
  • 02
    "Here is our new price list. What would a loyal customer push back on?"
  • 03
    "I need to tell a long-time client we are dropping a service they use. Draft the call notes."
It sounds too simple. Almost nobody does it. If you do not know whether the AI can help, that is the first question to ask it.
Two people working through notes at a desk with laptops open
LIVE DEMO

Live demo

Give me a question you doubt AI can answer. We will ask it together.

WHAT TO WATCH FOR
  • 01
    It interviews me before it answers
  • 02
    It uses my calendar and inbox because I let it
  • 03
    I ask it to argue against my plan
04

New capabilities in the past two weeks

Three things I built that I could not have built a month ago

Two people working through notes at a desk with laptops open
LIVE DEMO

Live demo of three things I built

  • 01
    A fishing game set in Nootka Sound

    With a salmon modelled in Blender from a prompt.

  • 02
    A speaker reel

    Cut from raw footage.

  • 03
    A junior hockey season site

    With highlight breakdowns of every game.

None of these started with a skill. Each started with a question.

New capabilities in the past two weeks

What clients have built

MOMENTS OF TRUTHGreeting at the doorOrder taken rightFood at the tableGoodbye and thanksLOCATION 3Greeting at the door9.4this week, up from 8.1

An app that tracks the customer moments of truth

Quick service restaurant group

Where are we exposed?Two of five marketsdepend on one buyerShow me the numbersSTRATEGY GRID

A strategic advisor for the planning cycle

Professional services firm

EVERY WEEK BY HAND30 hours a monthback in two calendars

Thirty hours a month of routine work automated

Payroll technology company, two leaders, one half-day session

05

Guardrails

Two rules and you are covered

Guardrails

Where your data goes

  • 01
    On a business plan your work does not train the model

    Claude Team, ChatGPT Business and Microsoft 365 Copilot on a work account do not train on your conversations. Personal accounts, including free Copilot, train unless you turn it off.

  • 02
    Still use judgment with sensitive data

    Think twice before pasting personal, customer or financial details when the task does not need them.

The simple rule. If you would not put it in an email to someone outside the company, pause before putting it into any tool.
Guardrails

A human approves anything that leaves the building or spends money

Start where mistakes are cheap and easy to undo

The Work Smarter Company35
Close

One thing to do tomorrow morning

Open the tool and say "Interview me, then set up my profile so you always know who I am and what I care about." Then ask it one real question from your week

Ten minutes. No committee, no platform decision, no strategy document.

Close

Who we work with

ABC RecyclingBackcountry BrewingBC Cancer FoundationBCAABrandliveCredit Counselling SocietyCross DevelopmentsFasken MartineauGreater Vancouver Board of TradeHall ConstructorsHeritage Trust CompanyHicks MorleyHigh Impact LabsLeavitt GroupMacKay CEO ForumsPHL FinancialPMACPowerexSee GroupThyssen MiningZayZoon

The free CEO guide to rolling out AI

Scan for the guide and its five templates

worksmartercompany.com/resources/ceos-ai-playbook

This presentation is in the resources section at worksmartercompany.com/resources

Colin Cox, The Work Smarter Company, colin@worksmartercompany.com

The Work Smarter Company38
1 / 42

Keyboard

Press Escape to close
Right, Space
Next build step, then next slide
Left
Back
Home, End
First slide, last slide
F
Full screen
M
Contents
E
Switch between presenting and reading
T
Start or pause the timer on a practice slide
R
Reset that timer
?
This list