Fee
Register to see indicative feesConfirmed in writing after scoping01 / Assessment · Flagship
AI Security Readiness Assessment
You have been asked where the organisation stands on AI risk, and the honest answer is that nobody has looked.
AssessAssessment
Typical duration
4 to 6 weeksDepends on AI systems and business units in scopePrimary artefact
Scored readiness positionPlus 4 moreWho leads it
Steve SharmaFounder & Principal04 / What this is
What AI Security Readiness Assessment is
A bounded, fixed-fee assessment of where your organisation actually stands on AI security and governance: across what AI you use, how it is governed, what it is exposed to, and what controls exist.
And what this service produces
It produces a scored position you can put in front of a board, and a remediation sequence that is prioritised, effort-estimated and costed, so the next phase can be scoped rather than guessed at.
05 / Who it is for
Built for a real assurance trigger, not a badge project.
01
This fits when
- AI is in use and nobody has assessed the position on it.
- Someone newly accountable for AI needs to know what they inherited.
- The board has asked for a position and received an opinion.
- You are about to fund AI security work without knowing where to start.
02
The usual trigger
- A board or audit committee has asked for an AI risk position.
- A budget cycle is approaching and you need a defensible number.
- A customer questionnaire arrived and you could not answer it.
- AI has been deployed and nobody has assessed it.
06 / The obligation
What an AI readiness assessment actually measures, and what most of them miss
Most AI assessments measure the model. That is the smaller risk, and the one an organisation has least control over.
The failures are organisational
The failures that actually happen are organisational: an AI tool procured without review, staff pasting client data into a public interface, an agent with broader system permissions than anyone intended, a vendor whose model changed under a contract nobody re-read. None of these are model failures.
Four things, measured
So this assessment measures four things: what AI is actually in use (rarely the list you were given), how decisions about it are made, what data reaches it, and what controls exist around it.
Inspected, not self-reported
It is evidence-based. We inspect configurations, review procurement records and interview the people using the tools, rather than accepting a questionnaire answer.
The word that matters is evidence-based. A defensible assessment separates what is claimed from what can be demonstrated, and the gap between the two is usually where the report earns its fee.
07 / What you receive
Five artefacts. Each one yours to keep.
- 01
Scored readiness position
A rating across five domains, with the evidence behind each score stated rather than asserted.
Report + scorecardBoard-ready - 02
Costed remediation sequence
What to do first, what it costs, how long it takes, and what can wait, so the next phase is scoped, not estimated.
Sequenced roadmapBudget-ready - 03
AI use register
Every AI system, tool and vendor in use. Usually the first complete list the organisation has had.
RegisterGovernance owner - 04
Gap analysis with severity
Each gap, why it matters, and what an adversary or an auditor would do with it.
Report sectionTechnical and executive versions - 05
Executive summary
Three pages a board can read without translation, with the position, the risk and the ask.
Summary documentBoard pack
08 / Scope
The whole system is visible before work begins.
AI inventory
- Systems, tools and vendors identified
- Shadow AI surfaced via procurement and expense data
- Data flows into and out of each
Governance
- Decision rights and approval gates
- Policy existence and actual application
- Accountability mapping
Data exposure
- What data reaches which systems
- Vendor contract and retention review
- Third-party model dependencies
Technical controls
- Access and permission boundaries
- Logging and monitoring coverage
- Agent and integration permissions
Assurance
- Evidence testing: claimed versus demonstrated
- Control sampling
- Vendor assurance review
09 / Included / not included
A useful boundary is part of the deliverable.
Included
- AI inventory, including shadow AI
- Governance and accountability review
- Data exposure and vendor review
- Technical control review
- Evidence testing and control sampling
Not included
- Remediation: This assessment finds and sequences gaps. Closing them is separate work, and we will say what it costs.
- Penetration testing or red teaming: Adversarial testing of AI systems is a different service with a different method.
- Model evaluation: We do not assess model performance, bias or accuracy. That is a data science engagement.
- Certification readiness: If you need ISO 42001, this assessment informs it but does not substitute for it.
- Continuous monitoring: This is a point-in-time position. We will tell you how quickly it decays.
10 / How we do it
A traceable path from scope to evidence.
Scoping and inventory
We establish what AI is actually in use, using procurement and expense data rather than a survey.
Evidence collection
Configurations inspected, contracts reviewed, controls sampled, users interviewed.
Assessment and scoring
Each domain scored against the framework, with evidence stated behind every score.
Sequencing and costing
Gaps prioritised by risk and effort, then costed so the next phase can be scoped.
Reporting and readout
Technical report, executive summary, and a session with whoever has to act on it.
Mapped againstISO/IEC 42001NIST AI RMFOWASP LLM Top 10MITRE ATLAS
11 / Timeline
4 to 6 weeks of work.
Elapsed time and customer effort are separated because they are not the same number.
Week 1
Scoping, access and inventory
Scope confirmed, access arranged and the AI inventory built.
Customer effort
One 90-minute kickoff workshop, plus procurement and expense data access.
Weeks 2 to 3
Evidence collection
Configurations, contracts and interviews.
Customer effort
5-8 interviews at 45 minutes, plus system read access.
Week 4
Assessment, scoring and sequencing
Each domain scored with its evidence, then gaps sequenced and costed.
Customer effort
Availability to clarify findings, roughly 2 hours.
Weeks 5 to 6
Reporting, review and readout
Report, review and readout. Across the whole engagement, expect roughly 12-16 hours of your team's time. We state that figure because understating it is the fastest route to a difficult week three.
Customer effort
One review cycle and a 60-minute readout session.
Our work: 4 to 6 weeks
12 / Scope check
Test whether AI Security Readiness Assessment
is the right next move.
We will cover likely scope, cost and fit.
Book a confidential strategy sessionWith Steve Sharma, who would lead.13 / Who leads the engagement
Named lead. Direct assessment.
Your assessment lead
Steve Sharma
Founder & Principal
Conducts the assessment personally. The interviews, the evidence review and the scoring are not delegated.
ASD-endorsed IRAP AssessorCISSPCREST
Every engagement has a named lead. Where a second executive or specialist is needed, they are named in the proposal and subject to client approval.
Review leadership record14 / Why Cybersecurity Link
Three decisions.
Accountability.
Fixed fee, with the exclusions stated.
You can see what this covers and what it does not before you contact us. Anything outside the standard scope is quoted before it starts.
Evidence-based, not questionnaire-based.
We inspect configurations, sample controls and review contracts. A claimed control and a demonstrated one get reported differently.
The output is a budget artefact, not a PDF.
Every gap carries an effort estimate and a cost, so the next phase can be scoped rather than guessed. That is the difference between a report and a decision.
15 / The boundary
What this engagement does not claim.
Remediation
This is an assessment. It finds and sequences gaps; it does not close them.
Remediation is separate work, and we will tell you what it costs.
Operations
We are not a managed security service.
We do not run your SOC or monitor your environment.
Rating
A readiness rating is ours, not a certification or a government endorsement.
We prepare organisations for assessment and certification, but we do not certify our own work.
Frameworks
One assessment, mapped across four references.
The work maps to ISO/IEC 42001, the NIST AI RMF, the OWASP LLM Top 10 and MITRE ATLAS.
16 / Client evidence
What this work produces
Financial services400-800 staffAnonymised, permission on file
The challenge
A board had approved an AI programme before anyone had established who was accountable for its risk, and a customer's procurement questionnaire arrived asking for a position the organisation did not have.
The approach
More of the work The inventory phase surfaced 23 AI touchpoints against a provided list of nine. Scoring and sequencing were built around the six that carried client data, rather than across all 23.
5weeks
From engagement to a board-accepted position, with a costed 18-month sequence attached. The scope discipline in week one did most of that.
17 / Cost
Fixed fee · confirmed in writing after scoping
Register to see indicative fees
Price your scopeWhat moves the fee
Per AI system in scope beyond the first eight.
Per additional business unit in scope.
Where procurement and configuration data is already centralised.
A fixed fee for the standard scopeScope and exclusions confirmed in writingAnything outside scope quoted before it startsNever absorbed and invoiced later
18 / How this fits
Where this sits in the wider programme.
19 / When this is not right
Sometimes the best recommendation is a smaller service, or a later date.
01
You already have a recent, evidenced assessment.
Bring us the gap list instead, and we will tell you honestly whether it needs re-testing.
02
You have no AI in production and no immediate plans.
The free self-assessment is enough for now.
03
You need adversarial testing of a specific system.
Red teaming is the right service, and this would be a detour.
20 / FAQ
The questions that change the decision.
Then it was a waste of your money, and we would rather find that out in the scoping call than after the invoice. If you have a recent evidenced position and a clear gap list, skip this and bring us the gap list. We will tell you whether we think it needs re-testing, and we will say so if we do not.
Roughly 12-16 hours in total, mostly interviews and access provisioning. The timeline above states it per phase. It is the question that stalls most assessments and almost nobody publishes it.
It is ours, scored against a framework we publish. It is not a certification and not a government rating, and we would rather say that plainly than let a score do work it cannot support.
We tell you immediately rather than saving it for the report. That has happened, and a finding sitting in a draft for three weeks is a failure of the engagement, not a feature of it.
Yes, as separate scoped work, and we will also tell you where your own team or a cheaper provider is the better option. We do not assess our own remediation.
Faster than most assessments. AI estates change monthly. We state a review horizon in the report, and it is usually six to nine months rather than the annual cycle people expect.
21 / Next move
If an AI Security Readiness Assessment is what you need, the next step is a conversation about scope.
45 minutes, with the executive who would lead it: what your scope actually looks like, and whether this is the right service for you at all.