Every dollar you
spend on AI, and
who spent it.

Your provider bill proves the spend exists.
It will not tell you whose it is.

Three answers to “who is spending this”,
because it is three different questions.

A person owns a key. A machine runs a process. And some of the cost belongs to neither — it belongs to a query in a loop that nobody has looked at since it was written. Shield 360 ranks all three, over the same window, from the same events.

Rank by code before you rank by person. A report that opens on a leaderboard of developers gets read as an accusation and stops being used; the same report opened on a list of expensive queries gets fixed.

Four states, and one of them is
an admission.

“Not covered” and “charged” are different questions, and most tools answer the first while printing the second. Shield 360 keeps them apart, and keeps a fourth state for the usage where nothing and nobody has said either way.

Find the AI nobody filed a ticket for.

Endpoint collection reads what AI tools write, so it finds the tools people install and nothing else. An application calling a provider directly writes no transcript and is invisible — while still spending. Shield 360 reads the socket tables the operating system already keeps, and reports which process, on which machine, as which user, talked to which provider.

Reading a socket table and reading a process’s memory are different sizes of thing, so they are two separate switches, both off until an administrator turns them on. The second is labelled reads process memory wherever it appears.

Some of your bill is a query
inside a loop.

Point the scanner at a repository and it reports the code that costs money to run: queries in loops, indexes made unusable by a function call, joins with no predicate, a connection opened per request. Each finding says what it costs and what to do instead.

for row in rows:
    cur.execute("SELECT * FROM accounts WHERE id = %s", (row.id,))
    enrich(row, cur.fetchone())

Nothing sits in the path
of a request.

Most tools that can tell you this much about AI traffic get there by terminating it: a proxy in the middle, a root certificate on every machine, decrypted traffic in somebody’s log. That buys visibility by weakening the thing it is watching. Shield 360 reads only what the operating system already records.

One paste per machine.
No administrator rights.

The collector downloads from your own Shield, authenticated by the enrolment key that is already in the command — so getting the agent costs no second credential, and the same paste works on an air-gapped network where your appliance is the only thing reachable.

Downloads live inside the app. Sign in, choose a plan, and Settings → Enrolment hands you a key and the exact command for each platform, already filled in. What is below is that command, so you can read the whole install before you commit to anything.

KEY='shield_ek_…'                       # Settings → Enrolment
ARCH=$(uname -m)
case "$ARCH" in aarch64|arm64) ARCH=arm64 ;; *) ARCH=amd64 ;; esac
curl -fsSL -H "Authorization: Bearer $KEY" \
  https://app.shield360.ai/api/v1/shield/agent/download/darwin-$ARCH -o shield-collector
chmod +x shield-collector

printf '%s' "$KEY" > /tmp/shield-key
./shield-collector enroll --endpoint https://app.shield360.ai --key-file /tmp/shield-key
rm /tmp/shield-key
./shield-collector run --since 2160h   # send the last 90 days, once
./shield-collector install             # then report hourly

$KEY = 'shield_ek_…'                   # Settings → Enrolment
curl.exe -fsSL -H "Authorization: Bearer $KEY" `
  "https://app.shield360.ai/api/v1/shield/agent/download/windows-amd64" -o shield-collector.exe

Set-Content -NoNewline "$env:TEMP\shield-key" $KEY
.\shield-collector.exe enroll --endpoint https://app.shield360.ai --key-file "$env:TEMP\shield-key"
Remove-Item "$env:TEMP\shield-key"
.\shield-collector.exe run --since 2160h   # send the last 90 days, once
.\shield-collector.exe install             # then report hourly

KEY='shield_ek_…'                       # Settings → Enrolment
ARCH=$(uname -m)
case "$ARCH" in aarch64|arm64) ARCH=arm64 ;; *) ARCH=amd64 ;; esac
curl -fsSL -H "Authorization: Bearer $KEY" \
  https://app.shield360.ai/api/v1/shield/agent/download/linux-$ARCH -o shield-collector
chmod +x shield-collector

printf '%s' "$KEY" > /tmp/shield-key
./shield-collector enroll --endpoint https://app.shield360.ai --key-file /tmp/shield-key
rm /tmp/shield-key
./shield-collector run --since 2160h   # send the last 90 days, once
./shield-collector install             # then report hourly

Rolling out across a fleet? The same binary installs through MDM, Intune or a configuration-management run — talk to us about the deployment path for your estate.

Connected providers say what
they can and cannot report.

Providers differ in what their billing APIs expose, and pretending otherwise produces confident wrong numbers. Every connection in Shield 360 declares its own limits, and a provider with no usage API is shown as unanswerable rather than as zero.

Cost is one risk. It is not
the one that ends careers.

Shield 360 produces the evidence an auditor asks for. It is not itself a certification, and nothing here should be read as one.

Priced on what you enrol,
not on what you look at.

Reports, findings and screens are never metered. Charging people to look at their own data teaches them not to look, which is the opposite of the point.