Customer data separation
Can AI work for one customer without ever seeing another customer’s data?
- Each customer’s data walled off at every layer, including AI search
- Separate, revocable access for every AI integration
AI readiness assessment
Know exactly what is stopping your AI, and how to fix it.
In two to three weeks our experts review your data, systems, cloud and security. You get a readiness score, the risks ranked by business impact, and a costed 90-day roadmap to AI that works in production.
$ tke assess --scope platform --areas all
✓ scanned 14 repos · 3 pipelines · 2 subscriptions
data_separation
82
knowledge
46
safe_actions
58
testing
74
cost
39
security
61
readiness 60/1003 blockers before agent rollout
Is this you?
The assessment is built for leaders who need AI to deliver, not just demo.
- Your AI pilot worked, but nothing has made it into production.
- The board is asking for an AI plan and you need one grounded in reality.
- Security or compliance won’t approve AI until they understand the risks.
- AI costs are rising and nobody can show the return.
- You are about to invest heavily in AI and want to avoid expensive mistakes.
What we assess
Six questions every AI rollout has to answer.
Each area is scored against evidence in your systems, not opinions in a workshop. Together they cover the whole journey from idea to AI running safely in front of customers.
Data and knowledge quality
Is the information your AI relies on accurate, current and owned by someone?
- One approved source for every answer
- Content labelled with owner, status and last review
Safe system connections
Can AI take actions in your systems without being able to do damage?
- AI proposes, people approve the risky steps
- Every action limited, logged and reversible
Testing and release
Can you change an AI feature and prove it still works before customers see it?
- Automated quality checks on every AI change
- Repeatable environments defined in code
Cost and performance
Do you know what AI costs per customer and per feature, and whether it pays back?
- Spend tracked per customer and per feature
- Budgets and alerts before the bill surprises you
Security, risk and compliance
Would your AI pass a security review, an audit and a regulator’s questions?
- Defences against prompt injection and data leaks
- An audit trail a regulator would accept
What we usually find
Six blockers that hide behind a successful demo.
None of them show up in a pilot. All of them show up at scale, in an audit or on an invoice.
Customer data that is not truly separate
Fine when your own app is the only thing asking for data. Dangerous when an AI assistant writes the request.
Knowledge with no owner
Three versions of the refund policy across a wiki, a PDF and a help centre. The AI finds all three.
Systems AI can’t safely act on
AI needs clear, limited actions it can preview and undo, not unrestricted access to your database.
AI changes nobody tests
Prompts and model settings edited by hand in a console, with no testing, history or way back.
No idea what AI costs per customer
The first big invoice arrives and nobody can say which customer or feature generated it.
Untrusted content treated as instructions
Documents, emails and web pages fed to an AI with nothing stopping hidden instructions inside them.
What you receive
Clear answers your leadership team can act on.
- Week 1: discovery. Stakeholder interviews and a read-only review of your systems, data and cloud.
- Week 2: analysis. Scoring with evidence, a security and risk review of AI use, and a cost baseline.
- Week 3: roadmap. Target architecture and a costed 90-day plan, presented to your leadership team.
Deliverables
- Readiness score. Your organisation scored across six areas, with the evidence behind every score.
- Risks ranked by impact. Security, privacy, compliance, cost and delivery risks, in priority order.
- Target architecture. What your data, systems and cloud need to look like for AI to run safely.
- 90-day roadmap. What to fix first, in what order, with effort and cost estimates.
- Board-ready summary. A plain-language report your leadership team can decide on.
Formats
Start where you are.
Executive assessment
Leadership interviews and a high-level review of your data, systems and AI plans. A readiness score, top risks and priorities for the board.
Discuss this format →Full AI readiness assessment
Everything in the executive assessment plus a hands-on review of your code, data, cloud and delivery process. Target architecture and costed 90-day roadmap.
Discuss this format →Assessment and pilot
The full assessment, then we fix the critical blockers and take your first AI use case into production with you.
Discuss this format →Questions
Frequently asked.
Who should be involved from our side?
Usually a technology leader, a business owner for the AI use cases, and someone from security or compliance. We run four to six short interviews and need read-only access to the relevant systems, or exported configuration if access is restricted. No production customer data is required.
Do we need to be on Azure?
No. Azure is where we go deepest, but the six areas apply to any cloud. We assess AWS, Google Cloud and on-premises estates on the same framework and call out where your platform changes the recommendation.
What do we actually receive?
A readiness score for each of the six areas with the evidence behind it, a risk register ranked by business impact, a target architecture, a costed 90-day roadmap, and a plain-language summary for your leadership team. It is yours to act on with your own team, with us, or with another partner.
We already have AI in production. Is it too late?
No, that is a common starting point. The assessment then focuses on hardening what is live (data separation, quality checks, cost tracking and security) before the next AI feature adds more risk.
How is this different from an AI strategy workshop?
A workshop produces ideas. The assessment produces evidence. We look at your actual data, systems and processes, and every finding is tied to something real and a concrete change that fixes it.
Get a clear, evidence-based view of your AI readiness.
Tell us where you are with AI. A founder will reply within one working day with how we would scope your assessment.