New: a readout deck for your customer meeting, and the Lens workshop. See it

Agentic AI readiness, assessed in an afternoon.

Try the demo Get notified at launch

For AWS and Azure consulting partners
Scored across seven pillars
Delivered under your brand

Amazon Web Services
Microsoft Azure
Read-only collector
Hosted in the UK
Open-source collector

Score any AWS or Azure estate for agent readiness

Customers Northgate Financial Q3 re-scan

Sample data

Ready to pilot

100% coverage · 38 of 38 checks assessed

P1Platform & Model Access
67
P2Identity & Access
66
P3Data Foundations
94
P4Security & Network
66
P5Observability & Operations
65
P6Cost & Capacity
73
P7Governance & Compliance
79

PartialHighModel invocation logging is enabledbedrock.get_model_invocation_logging_configuration · loggingConfig: CloudWatch Logs (/bedrock/invocations)
PartialHighGuardrails block prompt attacksbedrock.get_guardrail · PROMPT_ATTACK.inputStrength: LOW
PartialHighLong-lived IAM user access keys are minimisediam.get_credential_report · access_key_1_last_rotated: 2024-12-20T14:40:00Z
FailHighOver-privileged roles are limitediam.get_account_authorization_details · policy: AdministratorAccess

38 AWS and 36 Azure checks across seven pillars, aligned to the AWS Agentic AI Lens and Microsoft's AI guidance.

Features

AI and agent controls first
Evidence on every finding
Root-cause grouping
Coverage and confidence
Run-to-run comparison

Know what's already running

Customers Northgate Financial Q3 re-scan

Sample data

Current AI usage

2 models in use: about 61k calls and 80M tokens in the last 30 days, and about US$265 a month on Amazon Bedrock and SageMaker. Totals only, never prompts or outputs. Not scored.

Models, last 30 days

Model	Region	Calls	Tokens in / out
amazon.nova-pro-v1:0	Europe (London)	48k	61.4M / 9.8M
amazon.titan-embed-text-v2:0	Europe (London)	13k	8.7M / 0

AI spend by month

Jun 2026US$218
Jul 2026US$270
Aug 2026US$307

Average US$265 a month, mostly Amazon Bedrock.

AI outside Amazon Bedrock

Signs of AI used outside Bedrock (13 secrets and 40 Lambda functions checked by name). Worth asking who owns it and whether it follows the same guardrails and logging. Counts only: names are matched on the customer's machine and never leave it. Not scored.

OpenAI: 2 secrets and 3 Lambda functions named for it
Anthropic: 1 secret named for it

AI usage, spend and AI outside the platform, read as totals by the collector: never prompts or outputs.

Features

Models, calls and tokens
Monthly AI spend
Signs of AI outside Bedrock or Foundry
Actual spend beside the pilot's cost

Turn findings into a 90-day plan

Customers BrightBasket Retail Readiness assessment

Sample data

Today20/100Not ready
After Wave 154/100Foundational
After Wave 279/100Ready to pilot
After Wave 3100/100Ready to scale

Wave 1Days 0–30

S3 public access is blocked at account levelCritical
Root user has MFA enabledCritical
Multi-region CloudTrail logging is onCritical
Model invocation logging is enabledHigh

Wave 2Days 31–60

Long-lived IAM user access keys are minimisedHigh
Cost allocation tags are activeMedium
A data catalogue is in useMedium
Sensitive data discovery is enabledMedium

Wave 3Days 61–90

AWS Organizations is in useHigh
Centralised workforce identity (IAM Identity Center)High
A landing zone is in placeMedium
Regions are restricted for data residencyMedium

Every fix shows the score it earns, wave by wave, so your customer sees what the work buys.

Features

Three waves, ordered by dependency
Blockers first
Projected score per wave
Ready-to-run fixes: CLI, Terraform, CloudFormation, Bicep

Choose the first agent pilot with your customer

Customers Northgate Financial Q3 re-scan

Sample data

1Policy FAQ assistant · recommended first pilot · value 4, feasibility 5, low risk
2Complaint triage · value 5, feasibility 3, medium risk
3KYC document checks · value 5, feasibility 2, high risk
4Meeting summaries · value 2, feasibility 5, low risk
5Payment exception handling · value 4, feasibility 2, high risk
6Internal IT helpdesk · value 2, feasibility 2, medium risk

Low riskMedium riskHigh riskRecommended first pilot

Agent blueprint

Policy FAQ assistant

ModelAmazon Bedrock
Agent runtimeAmazon Bedrock AgentCore Runtime
KnowledgeAmazon Bedrock Knowledge Bases
Tool accessAmazon Bedrock AgentCore Gateway
GuardrailsAmazon Bedrock Guardrails

Model cost about $46.50–$93.00 a month3,000 runs a month, at list prices

Value, feasibility and risk scored together, with a recommended first pilot and its blueprint.

Features

Use-case workshop
Recommended first pilot
Agent blueprint
Running-cost estimate

Run the review workshop, pre-filled from evidence

Customers Northgate Financial Q3 re-scan

Sample data

Agentic AI Lens workshop14 of 40 questions done

18 of 146 best practices evidenced from configuration; 35 more answered in the workshop.

AGENTOPS06How do you implement testing, evaluation, and validation frameworks?

FailEvaluate and track ongoing agent performance

Tested by hand before release; no evaluation set yet.

Customer-reported, via workshop

AGENTOPS01How do you establish operational practices for agentic AI systems?

The policy FAQ assistant has a named owner (HR operations). Success criteria not yet written down.

Customer-reported, via workshop

Every best practice in the AWS Agentic AI Lens, evidenced from configuration where it can be, and answered with your customer where it can't.

Features

Lens evidence pack (PDF)
Answers labelled customer-reported
Progress across all 40 questions
Microsoft guidance mapped for Azure

Deliver it under your brand

Customers Northgate Financial Q3 re-scan

Sample data

Your logo

Agentic AI readinessNorthgate Financial

Prepared by your firm

Executive summary Approved
Identity & Access Approved
Data Foundations Approved
Root causesAI draft, not yet approved
RoadmapAI draft, not yet approved

AI drafts the narrative. Your consultant approves every word. Then resell it at your own price.

Features

Executive PDF
Readout deck (PowerPoint)
Technical workbook with fixes
Agent blueprint
Re-scan reminders every 90 days

From collector to report.

01
Your customer runs the collector
A read-only Python script in AWS CloudShell or Azure Cloud Shell. It produces one file per AWS account, or one for the whole Azure tenant.
02
The files come to you
Your customer sends them, or uploads them through a private link. Any missing permissions are flagged before you evaluate, so no run is wasted.
03
Review and run the workshop
Scores, evidence, root causes and a roadmap; then pick the first pilot with your customer.
04
Approve and export
Approve each AI-drafted section and export the report, readout deck, workbook, blueprint and Lens pack under your brand.

Built for your customer's security team.

curl -fsSLO https://flightglass.io/collector/aws/flightglass_collector.py
python3 flightglass_collector.py \
  --regions eu-west-2,us-east-1 \
  --output flightglass-bundle.json

Read-only by design

AWS: the SecurityAudit and ViewOnlyAccess policies plus eight read actions. Azure: Reader, Security Reader, Cost Management Reader and Entra Global Reader. Nothing can be changed.

Metadata, not data

Configuration and AI usage totals only, from an allow-list of fields: no secret values, no data contents, no prompts or model outputs. Spend on anything but AI services is dropped on your customer's machine.

Raw files deleted after 90 days

Collector files are deleted automatically; results stay until you delete the assessment, which you can do at any time.

Your customer stays in control

They run the collector in their own cloud shell and share the resulting files, by email or a private upload link. Flightglass never holds credentials for their accounts.

A person approves every word

Scores come from rules, never from AI. Claude drafts the narrative from the findings; nothing is exported until a consultant approves each section.

Audited and hosted in the UK

Every upload, export, approval and deletion is in your firm's audit log. Hosted in London; each firm's data is isolated from every other.

Start free. Pay per assessment or subscribe.

Prices exclude VAT · Resell assessments at your own price

Pay as you go

Buy assessments when you need them. Every assessment includes re-scans and a white-label report.

Free

£0

Your first assessment, up to 5 accounts or subscriptions, with 1 re-scan within 30 days. Marked as a free trial.

Assessment

£950each

Up to 25 AWS accounts or Azure subscriptions, with 2 re-scans within 12 months.

Assessment, large estate

£1,950each

Any size (26+ accounts or subscriptions), with 2 re-scans within 12 months.

Bundle of 5

£4,000for 5

Five assessments at £800 each, each with 2 re-scans within 12 months. Use them within 12 months of purchase.

Subscriptions

For firms running assessments every month. No minimum term; annual billing gets about two months free.

Partner

£2,500per month

or £25,000 a year

5 assessments a month (60 a year, pooled, on annual billing), 10 consultants, 4 re-scans each. Extra assessments £450.

Partner Pro

£4,500per month

or £45,000 a year

Unlimited assessments (fair use), unlimited consultants, large estates included, 12 re-scans each, priority support and annual invoicing.

Be first in when Flightglass opens.

Flightglass opens to partners soon. Leave your email and we'll tell you the day it does. Until then, the demo is open to everyone.

Or try the demo now.