Sheldon Smith

SHELDON
SMITH

I build AI systems
and put them to work.

View the Simone trailer

58 seconds, with sound.

Four live
AI systems

Hover a row for its trailer.Tap a row for its trailer.

Watch them work

Every name, number and message here is made up.

Books the table

Hi, can I get a table for four this Friday?

Confirm booking

Done. Four at 9 pm on Friday. Your confirmation is on WhatsApp.

Confirmation sent
Simone, live every Friday in Dubai

Turns a brief
into a course

Brief

Closing the till, for new cashiers

Audience, goals, policy and three cases. One page.

ENAR

18 screens, storyboard approved

One brief in, one human review, one finished module out

Writes the
7:30 brief

Agent Smith7:30 am

Good morning. Three things today:

1. Design review at 10:00. Your notes are in Notion.

2. Two invoices are due Friday.

3. Gym at 6 pm. You skipped Monday.

Agent Smith, briefs me at 7:30

Sorts what
went wrong

Handoffs

Nobody owned the handoff
Specs changed after sign-off

Tools

Two trackers, two truths
Access came a week late

Timing

Test data arrived late
Review squeezed into Friday

6 cards, 3 themes, no names

Debrief, built in a day and used the next

How I work

  1. 1I write the spec.
  2. 2A model builds it.
  3. 3A second model reviews it.
  4. 4I decide what ships.

Credits

I also lead training and change for a national healthcare program across 16 hospitals and 62 clinics. Fourteen years of getting new technology adopted taught me what a system has to do before people will use it.

PMPProsciSeven Epic modulesAI-103, in progress14 years in the field

Also built

Needs discovery

A live workshop tool. Participants answer from their phones and the facilitator watches the pattern form.

Course recovery

A tool that unwraps legacy SCORM courses and rebuilds them as editable, validated modules.

Games for friends

Browser games built for friends, including a spy lane-runner. Range matters too.

Simone: the trailer

Simone’s two lines in the phone call were recorded from her live line. Her narration is scripted, in the same voice. The WhatsApp exchange is re-enacted, and the caller, the guest and the seat counts are made up.

The E-⁠Learning Pipeline: the trailer

Every course screen in this film is a capture of one real build, a fictional retailer’s till-closing course, and the first voice you hear is that course’s own narration. The lines after the title were written for this film and read in the same voice. The clock pictures an estimate; it does not time this build.

Agent Smith: the trailer

Her lines were written for this film and are read by a voice chosen for it, not the one she speaks to me in. Her schedule is real. The brief on screen is a stand-in, with nothing under its bars, and the two figures are ink silhouettes, not pictures of anyone.

Debrief: the trailer

The meeting at the start is staged. The app on screen is the real one, recorded while it ran a session for a fictional project; the counter of people joining is sped up, and the join code’s QR and address are covered. The cards, the votes and the action are made up, and all four voices are synthetic.