two billion-dollar companies flew me to florida

hey friends,

this week, two billion-dollar companies flew me to Florida (20 hour trip from Chiang Mai Thailand 😅)

all so my team & i could watch people work & list opportunities.

our goal at Cyndra is to find where talented people waste their time with manual tasks. then automate it so they can focus on their zone of genius.

don’t replace jobs(humans are important). free up time so companies can scale without hiring sprees.

here’s the biggest lesson i learned from 4 days on-site overlooking billion dollar copmanies

btw, if you want a free AI audit from the Cyndra team → book one here

what i saw

these are massive companies with smart people, experienced leaders & millions of dollars moving through their systems.

but behind the scenes, people waste hours copying information between spreadsheets, PDFs, emails & software that doesn't talk to each other.

they were used to broken processes. wasting time on useless tasks instead of their speciality

it reminded me how to think about AI

you can't automate what you don't know

don’t ask → “where can we use AI?”

that leads to stuff nobody uses.

before you automate anything, you need to understand how work really gets done.

what really happens when someone opens their laptop

good automation starts with good documentation.

start with one real workflow

pick one task someone on your team completes frequently

ask them to record their screen while completing a real example from start to finish.

watch what they actually click, search, copy, check, send & wait for.

then turn that recording into a simple SOP.

what a useful SOP needs

the trigger

what causes the work to begin?

an email arrives. a form is submitted. a payment clears. a document is uploaded. the month ends.

the inputs

what information, files, systems & approvals are required before someone can start?

the steps

what does the person actually do, in order?

include the small steps people normally leave out because they feel obvious.

the decisions

where do they follow a clear rule, & where do they use judgment?

the exceptions

what causes the normal process to break?

missing information. mismatched totals. unusual requests. approval delays.

the output

what gets created, sent, approved or updated?

the definition of done

how does everyone know the workflow is complete?

the failure path

what should happen when the AI doesn't know what to do?

this matters more than people think.

a safe AI employee shouldn't confidently guess.

it should stop, explain what's missing & ask the right person for help

decide what should stay human

documentation doesn't only show you what to automate.

it shows you what not to automate.

some tasks are repetitive, low stakes & easy to reverse.

those are great places to start.

other tasks can be prepared by AI, but should still require human approval.

AI should give your team more time for that work.

try this with your team

this week, choose one recurring task & do this:

• record one real example from start to finish

• write down the trigger, inputs, steps, decisions, exceptions & output

• highlight every place someone searches, copies, checks, waits or follows up

• separate clear rules from human judgment

• choose one low-stakes section to automate first

• keep a human approval step until the system proves it can be trusted

start small.

run it on real work.

watch where it fails.

improve the SOP.

then automate the next section.

most businesses don't need more AI tools.

they need to understand their own business.

because you can't automate what you don't know.

ways to get involved

want weekly calls with me where we break down real workflows & build AI employees together?

join AI Operators.

want my team at Cyndra to give your company a free, in-depth AI audit?

book a call here & we'll identify what can be safely automated, what should stay human & where you should start.

no pitch. just value.

i appreciate you reading this.

if something resonated or you have questions, just reply. i read & respond to every email.

i'm rooting for you & hope the best.

→ johann

forward this to someone who's trying to automate a process they haven't documented.

subscribe here if someone sent this to you.