Sofpact field notes
AI strategy and operations, without the theatre.
Analysis for leaders working through use-case choice, governance, operating models, data readiness and adoption.
Written with AI assistance. Facts link to the primary sources named in each note, and Radko Diev decides what is published.
Cutover Is a Business Decision, Not an IT Date
The data can migrate perfectly and the business still stops. Five cutover decisions the operation must own, from reconciliation to hypercare.
Read field note
AI in Service Operations: Where the Measured Gains Land
A 14% average gain and about 34% for new starters: what the largest published field study of AI in customer support means for rollout, measurement and training.
Read field note
Running an AI Feature in Production: Plan for the Model to Change
Third-party dependencies and embedded models are among the fastest-growing risks named by supervisors. Six practices that keep an AI feature reliable over time.
Read field note
The AI Inventory: You Cannot Govern What You Have Not Listed
Most organisations run more AI than they think, because it arrives inside software already bought. What each entry needs, where to look, and how to sort by consequence.
Read field note
AI Literacy: The AI Act Obligation That Is Already Live
Article 4 has bound providers and deployers since February 2025 and is not tied to the high-risk timetable. What the Digital Omnibus changed and what measures look like.
Read field note
One Control Set, Not Two: AI Governance for UK Firms Serving Europe
There is no single UK AI statute, but existing law already applies and the EU regime may reach you. Why one control set beats two programmes, and where to start.
Read field note
AI in Insurance: Control the Decision, Not Just the Model
Govern insurance AI across underwriting, pricing and claims: high-risk classification, fairness testing, human oversight and explainable outcomes.
Read field note
AI in Regulated Life Sciences: Every Claim Needs a Traceable Record
AI in regulated life-sciences documentation: defensible attribution, retained records, pinned model versions and a narrow intended use.
Read field note
EU AI Act High-Risk Delay: What Changed, What Did Not
High-risk dates moved to 2 December 2027 and 2 August 2028. The obligations did not: the inventory, roles and evidence to prepare in the meantime.
Read field note
AI in the CEE Mid-Market: Sequence Adoption to the Team You Have
EU AI obligations apply at every size. Use regulatory sandboxes and SME measures, then sequence one workflow, evidence in context and an operating model.
Read field note