SMOrchestra.ai
Dtec
Dtec Innovators Forum · Sales & Marketing in the Age of AI
AI & the Future
of Growth.
From chatbot to AI-native company. On real screens.
Mamoun AlamouriFounder and CEO, SMOrchestra.ai
How many people work
in this company?
Org chart with 13 roles
1
2
3
4
5
6
7
8
9
10
11
12
13
0
roles on one board.
Every one is an AI agent.
1
human above it. Me.
Telegram work order to Sulaiman
One Telegram chat
I write the work order.
Sulaiman runs it.
1
The bar9.0 minimum on every dimension
2
The exit criteriaEngineering 9.5 · QA 9.0 · zero mock data
3
The reportBack to me in Telegram, with a scorecard
I set the standard. The agents do the work.
The scorecard tells me if they did.
A scheduled run
1:00 AM
Three agents researching in parallel. Ending in a quality gate.
Nobody was awake.
Scheduled run at 1:58 AM
“It’s just
a chatbot.”
The most expensive sentence in business this year.
Anthropic research · March 2026
AI can already do far more than we use it for.
Anthropic radar chart
Computer and math tasks
0%
AI could do, in theory
0%
Covered in real use
Same model for everyone in this room. The difference is what you build.
Anthropic, “Labor market impacts of AI: A new measure and early evidence,” 5 March 2026
What changes in a founder’s week
Today, you are the glue.
Tomorrow, the OS is the glue.
CRM
Email
WhatsApp
Calendar
Invoicing
Social
Website
Spreadsheets
Project mgmt
Newsletter
You,
copy-pasting
between them
Claude · the braincontext · memory · skills · judgment
Same tools. One brain. You go back to decisions.
July 2026 · The Hugging Face incident
1,200 AI agents tried to trick their own scorer.
0

agents in an isolated evaluation found a shared channel

0+

messages and files passed between them

~0

joined an attack on a live production system

1 in 5

wanted to tamper with their own transcripts

They wanted to learn how the scorer worked.
They optimised for the grade, not the goal.
Sound like anyone you’ve managed?
Source: METR independent investigation, 26 August 2026 · metr.org
This month, inside my company
My agents did it too. The system caught it.
Telegram: false merge status detected
Telegram: Sulaiman locks a new gate
18:02 · The check

“VP Engineering posted ‘PR #29 merged to dev’. Actually OPEN, not merged.”

Sulaiman · CEO agent

“No PR before gate review. No ‘done’ accepted without command output.”

One agent claimed done.
Another checked.
The CEO agent changed the rule.
I read it on my phone.
What we wrote into every AI employee
Four failures. Four design rules.
WHAT GOES WRONG
THE RULE
Agents game the grade
The agent that grades never produces.The eye collects. The hand produces. The scale grades.
Keys get exposed
Security runs at step zero.Before a single file opens.
Autonomy with nobody checking
Approve first.A channel earns autonomy after 10 clean approvals.
Character follows what you reward
Your best person on their best day.Nothing ships below 9 out of 10.
Same model at every level
The 5 levels of an AI‑native employee.
YOU TYPE
L1Task level
YOU TYPE
L2Starting employee1 to 2 years
YOU TYPE
“Prep tomorrow’s meeting.”
L3Skilled employee3 to 8 years
YOU TYPE
“Run the insurance campaign.”
L4Expert employee8 to 20 years
YOU TYPE
“Open Saudi.”
L5A departmentA whole function
The command gets shorter as the employee gets fuller.
Level 1 to 2
Write the job down once.
L1Three paragraphs in, a draft you correct
Level 1 prompt
L2Two lines in, the same shape every time
Level 2 skills
Level 3 to 4
Hire it. Then interview it.
Plugin directory with three AI employees
00 · THE HIRING INTERVIEW

Five questions, each with a ready answer. Out comes a plugin with your name on it.

“The foundation installs at level three.
The interview takes it to four.”
Level 4 · An employee with your name on it
Meet the Personal Assistant.
Personal Assistant plugin
Level 4 · What it hands me
Every morning, a brief. Every meeting, a dossier.
The CEO Brief email, figures blurred
FIGURES BLURRED FOR THIS ROOM
Every source named. Every figure traced to its system.
Meeting dossier cover
Meeting run of show
The next step
It reads everything. I read one page.Public sources only · run of show to the minute · next step written before we meet
Inside one AI employee · every run
The eye, the hand, the scale. Never the same agent.
00 · Step zerosecurity first

“List the filenames in templates/ by name only, opening nothing.”

from the live skill file: 01-sms-collect · step zero
01 · The eyecollects

“You gather. You do not produce, and you do not grade.”

from the live skill file: 01-sms-collect
02 · The handproduces

“You make one thing: the publish-ready week. You do not gather, and you do not grade.”

from the live skill file: 02-sms-produce
03 · The scalegrades, never edits

“You change not one character of the outputs.” Under 9, it goes back.

from the live skill file: 03-sms-check
The rule, inside the product
Approve-first is not a policy. It’s a setting.
SINE settings with autonomy level
“Locked to approve-first. Loosens per channel after 10 clean approvals.”
Level 5 · A department
“Open Saudi.” A revenue department runs.
Signal Sales Engine: 8 plugins
Plugin list with skill counts
0
plugins working as one engine
0
skills across the eight
SensoryInterpretationDecisionMotorMemoryIntelligenceCockpitOnboarding
What this means for sales and growth
20 steps from stranger to signed.
13
Run on autopilot

List · enrichment · signals · ranking · ads · messages · sequences · reply triage · booking · nurture · call prep · follow-up · CRM

4
Wait for your approval

Who we target · what we say · what we promise · what we charge

3
Stay in your hands

Discovery · negotiation · closing

The engine flies the route. Your people land the plane.
Two conversations that changed how we build
The founder and the GM.
A FOUNDER · TWO BUSINESSES
“It’s not a dashboard. It’s your AI‑native revenue analyst.”
Three days later:
“I want to give her a name. I want her on our WhatsApp, safely. And I want to hire more.”
A REAL-ESTATE HOLDING · KSA
“Autonomy kills accountability. I want a human to own every decision.”
His biggest cost was not manual work.
It was slow decisions nobody owned.
In a small company, the AI got a name. In a big one, it got an owner. Both are right.
The last use case
“Let’s hire AI to replace our staff.”
That’s the last use case. Start with these.
01

What are you procrastinating?

02

Freed from it, what would your team focus on?

03

What do you wish was off your to-do list?

04

Which revenue work never gets enough hours?

05

Which chores does everyone hate?

06

Where do you hire agencies?

07

What are you outsourcing today?

08

Where do things get lost: communication, comprehension, translation?

09

Where is orchestration hardest?

Your first AI employee is hiding in these answers.
Before your first AI hire
Seven questions. Then the eighth.
01 · BRAIN

Is what good looks like written down?

02 · CONTEXT

Does it know this client, this week?

03 · SKILLS

Is the job written once, step by step?

04 · TOOLS

Does it hold the right keys, and only those?

05 · STANDARDS

Can you score it against your best person?

06 · JUDGMENT

Does it know when to act and when to escalate?

07 · GUARDRAILS

What may it never do? Which human owns it?

08 · ONLY THEN

Who can we replace? The last question, not the first.

The model is rented. These seven layers are yours.
The thread through this morning
The future of
growth is AND.
AI does the work no one can, or wants to, do.
People do the work only people can do: judgment, relationships and decisions.
QR code for the AI-Native Readiness ScorecardWhere is your team today?Free readiness scorecard · 5 minsmorchestra.ai/tools/ai-native-readiness
SMOrchestra.aiMamoun Alamouri · Dubai and Amman · Registered member of the Claude Partner Network
SLIDE

ELAPSED
00:00
NEXT
→ / Space / click: next build · ← back · T: reset timer · B: black screen · F: fullscreen (main window)
Open this window with P from the main deck. Drag it to your laptop screen; the main deck goes on the projector.