Next live sessionThu, Nov 5 · 11 am ET · 8 am PT · 16:00 UKStarts soon
See what every AI call costs, and who made it, before finance asks.
Know what every AI call costs, and who made it is a free, one-hour live workshop. The Cost Control Tower meters every model call in your own AWS account by user, app and team. In the hour it runs live: a budget hit, a request rerouted to a cheaper model, and the audit evidence it writes.
What you'll get
- Every AI call tied to a user, an app and a team
- Budgets enforced before the spend
- Evidence mapped to the EU AI Act, ISO 42001 and NIST AI RMF
11 am ET · 8 am PT · 16:00 UK · online, one hour, free
Teams we've built for
- Comic Relief
AWS Premier Tier Services Partner, one of about 300 among more than 130,000 AWS partners.
You can't steer a cost you can't see, call by call.
What works is a gateway in your own AWS account that every model call passes through, tied to a user, an app and a team as it happens, which a spreadsheet pulled together at month end can never be.
With an owner on every call, budgets can be enforced before the money is spent. The same gateway routes requests to a cheaper model wherever quality holds up and serves repeated questions from its cache. It also keeps the auditor's evidence as it runs.
What you'll take away from the hour
- 01
Every call has an owner
The Cost Control Tower meters every model call in your own AWS account, by user, app and team. Attributing cost is about a two-week build once the definitions exist.
- 02
Spend you can steer
Each user gets a budget at the gateway, which can also move a request to a cheaper model where quality holds and answer repeats from its cache.
- 03
Evidence for the audit
Requests screened for personal data, denied topics and prompt injection, with evidence mapped to the EU AI Act, ISO 42001 and NIST AI RMF.
How the Cost Control Tower works
It's an AWS-native, single-tenant gateway in your own account that covers your Amazon Bedrock traffic. At first it only watches, and it starts enforcing once you've seen what it would have done.
- 1
Attribution
Every AI call tied to a user, an app and a team. Attributing cost is about a two-week build once the definitions exist.
- 2
Budgets
Set per user and per team and enforced at the gateway as each request arrives, so an overspend is stopped before it happens rather than found on the bill.
- 3
Routing and caching
Requests sent to cheaper models where quality holds, and repeated questions answered from a cache.
- 4
Screening
Prompts checked in your account for personal data, denied topics and prompt injection, so off-policy requests stop at the gateway.
- 5
Evidence
Compliance evidence packs mapped to the EU AI Act, ISO 42001 and NIST AI RMF, generated as the traffic runs.
It's delivered in phases, starting with a read-only discovery, so nothing about how your teams work changes until you choose to enforce.
What we bring
Built, running in your account, and run by our engineers.
AI Cost Control Tower
A gateway in your own AWS account that meters every model call by user, app and team, enforces budgets and writes the audit evidence as the traffic runs.
Single-tenant, in your account; it starts read-only.
AK Run's Agent ControlPlane
Every AI agent's model calls go through one gate: only the models you approve, each call tied to the person who asked, cost tagged by team, and limits when you want them.
Runs in your own AWS account, on Amazon Bedrock AgentCore.
One hour, run by the engineers who do the work
See it
One month of AI calls attributed by team and by outcome, then a budget hit and a request rerouted to a cheaper model, live.
How it's built
Monitor first, then budgets and policies; what's read-only and what changes.
Your questions
Bring last month's AI bill and the question finance asked about it.
The private workshop: a read-only look at your own AI traffic
With your engineering and finance leads, on your real usage. Who spends what, on which feature, and what to control first.
Who comes
The engineering lead for the AI features, whoever owns the AI bill, and someone from security or compliance.
What to bring
Last month's usage export and the bill.
Your data
Usage metadata only: counts, tokens, models and tags. No prompts or outputs are needed.
What you leave with
- Your AI spend broken down by team and feature
- The definitions attribution needs, agreed
- What to budget, route or cache first
- The phases of a rollout, starting read-only
What clients say about working with us
Armakuni assisted in helping us to upskill our teams in modern devops practices, ways of working and product management.
Armakuni's availability, ease of communication, cost-effectiveness, and technical skills were outstanding.
They worked closely with our AWS team to optimize costs and minimize the financial impact of our migration.
- Under 0.25%of AWS partners are Premier Tier. We're one of about 300.
- 500+AWS customer launches
- 1,300+engineers across the UK, US, UAE, India and Canada
- 70%of the proofs of concept we scope reach production
Before you sign up
Is it free?
Yes, the live hour and the read-only look at your traffic. The gateway itself is built in phases only if you go ahead.
Do we need a gateway at all?
Not always. If your AI spend is small and owned by one team, a monthly report may be enough, and we'll say so.
Will it slow our AI calls?
The gateway runs in your own AWS account, in the path of the call; we measure its overhead in the read-only phase before anything is enforced.
Do you see our prompts?
No. It runs single-tenant in your account, and the read-only discovery needs usage metadata only.
What happens after?
A read-only discovery first, then budgets and policies in phases, each switched on by you.
Know what every AI call costs, and who made it
Thursday, November 5, 11 am ET · 8 am PT · 16:00 UK. Starts soon.
Save my seat