Activate and share
People learn to delegate, describe, verify and use AI responsibly on their own business cases. Useful practices stop staying isolated.
Individual productivity
and quality of use
and quality of use
GenAI vision · from technology to the ground Understand before deploying
Hand more of the work to agents, without losing the state or the control of it.
This conviction drives my practice: start from real processes and real work, build agents that are actually useful, measure their effects, and decide with clear eyes how far execution can be handed over.
UnderstandStructureBuildMeasure
01
The bottleneck and the value are moving
When producing becomes abundant, value moves to whatever stays scarce.
Models are improving fast. They write, analyse, search, handle files and now act inside tools. But raw capability does not know the right process, the authorised source, the business exception, the accountable person or the acceptable level of risk.
So the question is no longer only what AI can do. It is deciding what it should do, giving it the right context, fitting its output into a real process, checking its results and measuring whether anything actually improved.
And as agents absorb more execution and more coordination, another need becomes critical: keeping an understandable state of the work, of its decisions, of its evidence and of who answers for it.
I call value shear the twin movement that compresses the value of tasks which have become abundant and pushes it up towards the new bottlenecks.
The exact pace remains uncertain. But as soon as a once-scarce capability becomes widely available, an organisation has to move its attention to what becomes limiting instead: context, orchestration, checkpoints, accountability and real learning.
What becomes abundant What becomes decisive
Produce
Transform
What becomes abundant loses relative value. Value migrates to the new bottleneck.
Deploying a tool gives access to a capability. Transforming the work creates durable value.
02
Value is built at three levels
From scattered individual gains to a collective capability that can be integrated and steered.
The decisive step is not only going from “no AI” to “with AI”. It is anchoring the value beyond a handful of well-equipped individuals.
Structure
AI is embedded in a real activity: steps, data, responsibilities, checkpoints and fallbacks. Gains become visible at department level.
Lead time · quality · volume
cost per case
cost per case
Arbitrate and govern
Initiatives, costs, risks and results are consolidated. Leadership can decide where to invest, what to improve and which uses to stop.
Overall performance
and better-informed decisions
and better-informed decisions
03
A complete transformation framework
Four pillars make an organisation able to deploy, supervise and prove the value of its AI uses.
I structured a GenAI Operating Model to avoid fragmented efforts: training without process, use cases without data, a platform without adoption, or an ROI announced without defensible measurement.
Skills, culture and supervision
The critical skill is not prompting. Teams need to delegate, describe, discern and act with diligence. With agents, they also need to test behavioural stability, spot the limits and take back control.
Evolving data and knowledge
Do not wait for perfect data. Start by structuring the documents, data, rules, decisions and knowledge that the priority use cases and their measurement actually need, then grow that asset step by step.
Agent platform and control plane
The right technical level is the simplest one that meets the need without hiding what will break at scale. The platform also decides what can be traced, governed and proven.
Innovation and value creation
Use cases are prioritised with four questions: value, feasibility, measurability and risk. Every pilot starts from a hypothesis, a comparable baseline and thresholds agreed before the test.
Cross-cutting governance
It sets who decides, which uses are launched or stopped, what can be proven, and how agents stay visible and under control over time.
The pillars execute. Governance decides. Security informs. Measurement disciplines the value story.
04
I do not stop at the recommendation
The framework becomes a short intervention on a real episode of work.
At IoD Solutions, this vision takes the shape of the GenAI Commando Sprint. The client is not buying a number of assistants. They are buying the measured transformation of one targeted activity.
- 01
Qualify
Secure the sponsor, the scope, the data, the test cases and the initial measurement.
- 02
Observe
Follow one real business episode: its files, its decisions, its friction and its workarounds.
- 03
Eliminate
Remove, reduce or standardise what should be, before introducing AI.
- 04
Augment
Build the minimal solution with the users: assistant, skill, script, plugin or workflow.
- 05
Measure
Compare before and after, document the limits, then decide: deploy, improve, instrument further or stop.
You are not only designing an assistant. You are designing a new way of working.
The level of evidence depends on the ground available: declarative, light observation or instrumentation. I would rather be explicit about what can genuinely be measured than turn a technical demo into a promise of transformation.
05
Four ways in, at IoD Solutions
Choosing the level of intervention that fits the problem and the company’s situation.
Build literacy and train
Make teams autonomous on their own business cases, and grow their ability to judge what AI produces.
Diagnose
Start from processes, friction and real opportunities before choosing a tool.
Create assistants and workflows
Test a focused case quickly inside the existing environment, and add complexity only when it becomes necessary.
Govern
Anchor uses in the routines, track their value, cost and risk, then improve or retire what no longer serves.
Every company starts from a different position. A scan across the framework’s four pillars identifies its strengths, weak points and priorities, then guides the right level of intervention: building literacy, transforming a process or structuring governance.
06
The same logic across my projects
Understand the system, build the mechanism, confront the result.
The GenAI framework and the Commando Sprint support the integration of reactive agents into a defined process. Moving to more autonomous agents changes the problem. They need a reliable representation of the situation: what has been done, what is still expected and how far they may act. Private Twin explores that missing piece, bringing expected work and observed work together to maintain a current state that is sourced, correctable, governable and shared between people and agents.
From scattered AI usage to an activity that is genuinely transformed.
If you want to identify a relevant process, build a first use case with your teams, or structure a fuller trajectory, let’s discuss the level of intervention that fits your situation and priorities.