Representative engagements, anonymised by default

On arrival

Design licences had been bought ad hoc over several years across multiple vendors. Nobody held a complete list, some seats were assigned to former staff, and two renewal dates had already been missed.

What we did

• Full entitlement review across every design and productivity vendor in use
• Reassignment of orphaned named-user seats and retirement of duplicate subscriptions
• Consolidation onto collection-level licensing where the usage pattern justified it
• Licence register established with entitlement counts and renewal dates
• Licence server configured for the seats that needed to float between workstations

What changed

The practice now works from a single licence register with a known renewal calendar, and reassigning a seat when someone joins or leaves is an administrative task rather than a procurement exercise.

• Software licensing
• Autodesk
• Adobe
• Asset management

On arrival

Mail and documents were split between an ageing on-premise server and personal cloud accounts staff had set up themselves. There was a backup job, but no record of it ever having been restored.

What we did

• Workload assessment and migration plan with a documented rollback path
• Tenant and identity configuration, including conditional access policy
• Staged mailbox and document migration with reconciliation at each stage
• Consolidation of shadow cloud accounts into managed identities
• Recovery procedure written and then actually exercised against live data

What changed

Mail and documents run on managed identities with access policy applied centrally, and the firm has a recovery procedure that has been tested rather than assumed.

• Microsoft 365
• Migration
• Identity
• Backup & recovery

On arrival

Stock and accounting systems each held their own version of product and customer records. Month-end reporting was a manual reconciliation in a spreadsheet, and the two systems never quite agreed.

What we did

• Integration mapping across both systems, with field-level contracts documented
• Middleware built to synchronise product, customer and transaction records
• Data cleansing and de-duplication before the first synchronisation run
• Reconciliation report generated for every run, not just at month-end
• Reporting rebuilt against agreed definitions of the shared measures

What changed

The two systems now reconcile continuously with a report on every run, so a discrepancy is visible when it happens rather than at month-end.

• Integration
• Data migration
• Reporting
• Middleware

On arrival

Staff and student devices spanned several Windows versions with inconsistent patching and no central policy. Academic licensing entitlements were not being applied to eligible purchases.

What we did

• Device inventory and patch-state assessment across the estate
• Academic entitlement review, with eligible purchases re-sourced correctly
• Central device management and update policy configured
• Access scoped separately for staff and students against real requirements
• Endpoint protection deployed with a consistent policy baseline

What changed

Devices are patched under a central policy, staff and student access are scoped separately, and licensing is applied against the correct academic entitlement.

• Device management
• Academic licensing
• Endpoint security
• Identity

On arrival

Several services had been built separately over time. Each had its own login and its own copy of the same user records, so a change of address had to be made in three places and often was not.

What we did

• Service and data-model review across the existing applications
• A shared identity and permission layer introduced across the estate
• Each service rebuilt against that layer with its own permission scope
• User records unified, with document handling and notifications centralised
• Administrative roles and audit logging configured per service

What changed

Users hold one account across every service, administrators work from a single record, and adding the next service extends the existing layer rather than starting another standalone application.

• Platform engineering
• Identity
• Data consolidation
• Application development

On arrival

A team was manually transcribing fields from a high volume of incoming documents into a line-of-business system — slow, and error-prone in a way that was hard to detect after the fact.

What we did

• Feasibility assessment on a sample of real documents before any build
• Extraction and classification scoped to the specific document types in scope
• Accuracy measured per field against a manually verified baseline
• Review interface built so every extracted record is approved by a person
• Known failure modes documented, with an escalation path for low-confidence output

What changed

Extraction is proposed by the system and confirmed by an operator, so the transcription effort drops while the record of who approved each entry is stronger than before.

• Applied AI
• Document processing
• Workflow automation
• Evaluation