>gabes/the letter
the letter
01 · 00:00:00
// the weekly letter · issue 07 · aug 10 to 14

five signals.
one week.
the word arrived.
the capability did not._

about 130 vendors out of thousands actually have the thing they are selling. the builders are spending four dollars for every one the market pays back. a policy written before the tool existed taught people to hide. and seventy three percent adoption changed how ten percent of them operate. every number this week is measuring a label.
// today, live on the page

the newest signal, pulled the moment you land.

loading the latest signal.
01proofmonday

of the thousands of vendors
selling agentic ai, about
130 actually have it. the
rest renamed the chatbot.

the same research desk expects more than 40% of agentic projects canceled by the end of 2027. cost is only the second reason.

two numbers from the same desk, and they explain each other. gartner puts real agentic capability at roughly 130 vendors out of the thousands claiming it, and expects more than 40 percent of agentic ai projects to be scrapped by the end of 2027, on cost, unclear value, or risk controls nobody built.

a canceled project does not just burn the budget. it burns a year of calendar and the company's appetite for the next attempt, which is the expensive part.

the fake move is picking a bigger platform and granting the agent more autonomy, which schedules the production incident sooner. do the version that holds: ask the vendor to name one decision the system makes without a human and what happens when that decision is wrong. then scope the smallest agent that touches nothing irreversible, and earn the next one.

domake the vendor name one decision the system makes without a human, and what happens when it gets that decision wrong.
not thisdo not answer a weak system with more autonomy. that schedules the incident, it does not prevent it.
a real answer to that question is specific and slightly uncomfortable. a rehearsed one is neither.
receipt gartner, june 2025: roughly 130 vendors with real agentic capability out of thousands claiming it · gartner: more than 40% of agentic ai projects expected to be canceled by the end of 2027, on cost, unclear value or absent risk controls
shipped monday ·xin
02shifttuesday

the companies building
the ai are spending about
four dollars for every one
the market pays them back.

their capex problem is not your problem, unless you bought the same way they did.

ai infrastructure spend is running near $400 billion a year against roughly $100 billion in enterprise ai revenue. chip stocks slid into a bear market this summer, and fund managers just logged a record reading on overspend worry.

the gap belongs to the hyperscalers. the buying habit it normalized landed on you: seats, tiers and capacity bought ahead of any workload that asked for them, on the theory that the capability would find a use later. it shows up in your p&l as licenses nobody activated and a committed tier you keep renewing because unwinding it now looks worse than paying for it.

the move going around is to wait out the shakeout and buy cheaper in a year, which is the same unattached purchase with a later date on it. do the version that holds: name the one workflow costing you the most hours this quarter, buy only what that workflow needs, and write the number it has to move before the renewal lands.

doname the one workflow costing the most hours, buy only what it needs, and write the number it must move before renewal.
not thisdo not wait out the shakeout to buy cheaper. that is the same unattached purchase with a later date on it.
a vendor who cannot price against a single workflow is asking you to fund their capacity, not your throughput.
receipt roughly $400bn annual ai infrastructure spend against about $100bn in enterprise ai revenue · chip stocks entered a bear market this summer · record fund manager reading on ai overspend concern
shipped tuesday ·xin
03trapwednesday

the ban did not stop
the pasting. it moved it
to accounts you cannot see.

shadow ai is not a discipline problem. it is what happens when the sanctioned option is worse than the one already on their phone.

77 percent of employees now paste content into ai chatbots, and roughly 82 percent of the risky pastes go through personal, unmanaged accounts. that is client data, pricing, contract language and code sitting in a consumer account tied to someone's personal email: outside your retention policy, outside legal hold, and out the door with them when they leave.

68 percent of security leaders report an ai-linked data leak while only 23 percent have any policy written down, which tells you the leaks are running well ahead of the paperwork.

the move everyone reaches for is the blocklist and the annual training module, and it produces exactly one measurable outcome: the same work, done on a phone, in a tab you have no visibility into. do the version that holds: stand up one sanctioned account with retention and admin controls on, tell people plainly it is there and why, then write the policy in three lines. use the company account, never paste client or employee data, one named person owns exceptions.

dostand up the sanctioned account with controls on first, tell people it exists, then write the policy in three lines.
not thisdo not lead with the blocklist and the training module. the work moves to a phone and you lose the last of your visibility.
a rule written before the sanctioned tool exists is a rule that teaches people to hide.
receipt usecure / layerx: 77% of employees paste content into genai tools, roughly 82% of risky pastes through unmanaged personal accounts · metomic: 68% of security leaders report an ai-linked leak against 23% with a written policy
shipped wednesday ·xin
// the week's argument, running

list what you bought.
write what it does.
read the gap.

every signal this week was a word standing in for a capability. put in the systems you pay for, write the one job each actually performs without a human, and tap whether you have watched it do that. nothing is saved or sent.
these rows are the week's five, seeded blank on purpose: edit every field, add your own. it updates as you type.
what you boughtthe job it does alonetap: have you seen it?
04movethursday

73% of companies say ai
runs through most of their
work. 10% say it changed
how the business runs.

the tool is in the building. the operating model never got the memo.

publicis sapient surveyed 1,550 ai decision-makers at companies with at least 500 people and $100m in revenue, and the two numbers that matter sit right next to each other. 73 percent say ai is used regularly or across most business processes. 10 percent say it is core to how the business operates.

47 percent think the technology is already good enough for what they need today, while 42 percent say their own company is not built to capture the value. that is a room full of buyers quietly telling you the bottleneck is not on the vendor's side of the table. the cost is a year of licenses, training hours and pilot calendar spent buying a faster version of the same process, with the same handoff, the same approver, and the same undocumented definition of good sitting underneath it.

the fake move is the next platform, or the agent layer on top of the platform, which adds capability to work nobody has written down. do the version that holds: take the one process that touches revenue most often, write its real steps as they happen now and not as the org chart claims, name who owns each handoff, define a good output in one paragraph, then let the tool into that.

dowrite the revenue process as it actually runs, name the owner of every handoff, define a good output in one paragraph, then let the tool in.
not thisdo not buy the next platform or the agent layer above it. that adds capability to work nobody has written down.
22% already name their own operating model as the barrier. the diagnosis is not the expensive part anymore. the writing down is.
receipt publicis sapient 2026 global enterprise ai report, june 2026, 1,550 ai decision-makers at organisations with 500+ employees and $100m+ revenue: 73% use ai across most business processes against 10% who call it core to operations, 47% say the technology already meets today's needs, 42% say their organisation is not set up to capture the value, 22% name the way their organisation operates as the primary barrier
shipped thursday ·xin
05warningfriday

the ftc has filed thirteen
cases over what companies
called ai. the enforcement
lands about two years
after your money did.

a regulator can eventually price the lie. it cannot give you back the year you built around it.

ai-washing is now an enforcement category, with thirteen federal actions since 2024 over claims about what a product could actually do. the one worth reading is air ai, banned this march from marketing business opportunities after selling small businesses an ai phone agent at $25,000 to $100,000 upfront, with the complaint describing software that was glitchy or not consistently available and buyer losses reaching $250,000.

notice the structure, because it repeats far below the level that gets a press release. the capability lives in the demo and the deck, the money is due before anything runs, and the contract you sign describes a license to software rather than a job the software does. the cost is not only the check. it is the quarter you spent staffing around a capability that never arrived, and the credibility you spent internally telling everyone it would.

the fake protection is due diligence theater: asking the vendor whether the ai is real, sitting through a longer demo, reading a security whitepaper written by the same marketing team. do the version that holds: make them write into the contract the specific task the system performs without a human, pay monthly against it rather than upfront, and put in a thirty day exit if the task is not being performed.

dowrite the specific task into the contract, pay monthly against it, and keep a thirty day exit if the task is not being performed.
not thisdo not run diligence theater. a longer demo and a vendor-written whitepaper are the same sales process in a second costume.
a vendor selling capability will sign that. one selling a category name will explain why their pricing model does not allow it, at length.
receipt dla piper, may 2026: 13 ftc ai-washing enforcement actions since 2024 · ftc press release, march 2026: air ai and its owners banned from marketing business opportunities, $25k to $100k upfront license fees, buyer losses up to $250k, software described as glitchy or not consistently available
shipped friday ·xin
// take it with you

five labels.
here is how you
read the load.

every signal this week came down to one question you have to put in writing. these five prompts each produce the page that asks it. paste into any assistant (chatgpt, claude, gemini) and use today. each card says what it does and what you get back.
no email. no signup. no follow-up sequence waiting for you. take it and go.
01
proof
writes the one question that separates a real system from a renamed chatbot.
you get the decision it makes alone, what happens when it is wrong, and the smallest safe scope.
02
shift
attaches every purchase to a single workflow and a single number.
you get the workflow, what it actually needs, and the number that has to move before renewal.
03
trap
writes the sanctioned account plan and the three-line policy, in that order.
you get the account with controls on, the three rules, and the named owner of exceptions.
04
move
writes your revenue process as it actually runs, not as the org chart claims.
you get the real steps, the owner of every handoff, and one paragraph defining a good output.
05
warning
turns a vendor pitch into three contract terms.
you get the task written into the contract, monthly payment against it, and the thirty day exit.
one file. opens in any notes app. drop it next to the work. no gate, no catch, nothing waiting for you after.
saved. that is the point of us: the value shows up before the invoice does.
// the signal daily, the letter weekly

five labels.
five loads.
one invoice each.

one signal a day, monday to friday, public. the letter every friday, the same hour the fifth signal lands: the week compiled, connected, and pointed at what to do. no quizzes, no funnels, no webinar.
good. fridays, then. bring us the part that keeps breaking.
>gabes · operating systems for companies done improvising · issue 07 · past issues → · past signals →
> scroll. watch the labels get measured.