Self-assessment

Score yourself

Answer each question honestly. Yes = 2, Partly = 1, No = 0. Keep a running total.

Your score

0/ 24

0 of 12 answered

Not scored yet

The map

0 / 6

01 Is there a current inventory of every system that holds personal data or acts on consent, made before evaluating any platform?

02 Do you know which of those systems have APIs, and which need adapters, file bridges or staged replacement?

03 Is the integration order decided: which systems connect in phase one, and why those?

The architecture

0 / 6

04 Is the architecture expectation written: API-first and modular, so connections can be added, tested and replaced without rebuilding the whole?

05 Is the deployment posture decided: your data centres, your cloud tenancy, or vendor hosting, with security, sectoral and cost input?

06 Is infrastructure responsibility settled, and are production, disaster recovery, test and development sized and separated?

The duties

0 / 6

07 When consent changes, does the change reach every system that acts on it, with pass or fail test evidence?

08 When a right is exercised, does the request execute against the systems where the data actually lives?

09 When retention ends, does deletion reach archives and backups through the integration?

The exit

0 / 6

10 Was any platform tested on your systems and your volumes, not on the vendor's sample data?

11 If you changed vendors tomorrow, would your notices, consent events, artefacts and configurations export cleanly?

12 If AI is used inside privacy operations, are the guardrails written: human review of any decision about people, AI activity logged in the processing records, model documentation available to the DPO?

Readiness bands for this self-assessment
Score Where you stand What to do next
0 to 8ExposedThe duties live in documents, not in systems. Map the estate before anything else; buying before mapping buys shelfware.
9 to 16HalfwayConnections exist but the duties do not all execute. Test propagation, rights execution and deletion reach, with evidence.
17 to 21Strong baseRun the exit test: a full export of notices, events, artefacts and configurations.
22 to 24LeadingKeep the connection register current as systems are added, retired and replaced.

Any single "No" is a named gap, whatever the total says. Nothing leaves your browser.

What integration means here

The discovery and mapping work tells you which systems hold personal data and act on consent: core business systems, CRM, marketing platforms, data warehouses, HR systems, document management, partner connections. Integration makes the duties executable across them. In banking the connection list is consistent: the core banking system, internet and mobile banking, payment and digital lending platforms, payment rails, identity services on a strict consent basis, document management, HR systems and third-party providers.

The framework: four moves

Move 1: Map before you buy. Inventory the systems that must connect before evaluating any platform. The map comes first; the platform is chosen against it.

Move 2: Set the architecture and the posture. API-first and modular. Decide deployment early: your own data centres, your own cloud tenancy, or vendor hosting. Regulated enterprises increasingly insist on their own environment; some mandate deployment on their own public-cloud tenancy. The choice is hard to reverse, so make it with security, sectoral and cost input. Settle infrastructure responsibility and environment sizing. Plan the legacy path: systems without APIs need adapters, file-based bridges or staged replacement; ignoring them leaves the oldest data outside the programme.

Move 3: Connect and test the duties. Consent propagation, request execution, deletion reach: each tested like a duty, with pass or fail evidence. A demonstration on the vendor's sample data proves nothing about your estate.

Move 4: Plan the exit and the AI guardrails. The export test: notices, consent events, artefacts, configurations, all exportable in readable form, so consent evidence never dies inside a subscription. On AI inside privacy operations, careful practice admits it only with guardrails: no autonomous AI decisions about people without human review, AI activity logged in the processing records, model documentation available to the DPO for audit, and an AI bill of materials alongside the software and hardware ones. The AI tools page covers the wider picture.

Template: the integration connection register

One row per system. Copy this and fill it in.

SystemWhat it holdsConnection typeDuties it must executeTest statusLast testedOwner
Example: marketing platformCustomer profiles, consentsAPIConsent propagation, erasurePassingDigital
Example: legacy loan systemClosed loan filesFile bridgeRights retrieval, deletionPartialIT

The downloadable file adds the column guide and the duty-level test list.

What good looks like at the end

You should have: an integration architecture and API specifications, a decided deployment posture, environment sizing, and integration test results proving the duties execute across systems.

The live test: capture consent on one channel, withdraw it on another, and verify the stop in every downstream system and at one partner. Then ask the vendor to export everything: notices, events, artefacts, configurations. Both results tell you whether the integration is real.

Where enterprises go wrong

  • Buying the platform before mapping the systems it must touch.
  • Accepting a demonstration on the vendor's sample data as proof of fit.
  • No exit plan: consent evidence locked inside a subscription.
  • Core system replacement smuggled into a privacy project. Integration connects privacy capability to what exists; rebuilding what exists is a separate programme with a separate budget.

The decisions to make

  • Deployment model and who hosts.
  • Infrastructure responsibility and environment sizing.
  • Integration breadth: how many systems in phase one, and which.
  • How legacy systems without APIs are handled.
  • Whether and how AI is admitted into privacy operations, under which guardrails.

Next step

Training the people who touch personal data, because systems do not leak data on their own. If you need help with any of this, write to us and we will help.