From first step to scale
A roadmap that fits your organization, a decision framework, a coherence diagnostic and a value simulator for your own figures.
Nine steps from value to scale
Start with the outcome, not the technology. Transform one process that matters end-to-end, prove the value, then industrialize what you built.
01Define valueIdentify the strategic business outcomes.+
Which outcomes would change our competitive position?- Value hypotheses
- Executive sponsor
- Outcome metrics
02Select processesFind high-value processes suitable for reinvention.+
Where do handoffs, waiting and coordination cost us most?- Process heatmap
- Value and feasibility scoring
- Shortlist
03Redesign workMap today's process and design the future human-agent workflow.+
If we designed this process today, what would it look like?- Current-state map
- Future human-agent workflow
- Role changes
04Define autonomyDecide what agents may recommend, decide and execute.+
What is the right autonomy for the risk of each decision?- Decision rights
- Escalation rules
- Risk classification
05Build the foundationPrepare data, architecture, governance and workforce capabilities.+
What must be in place before agents can act safely?- Data access
- Orchestration platform
- Control layer
- Skills
06Sequence for absorptionPlan what the organization can absorb, not what is fastest to build.+
What can teams absorb now without losing coherence?- Change load per team
- Friction removed first
- Phased plan with momentum
- Daily operations protected
07Launch a lighthouse transformationTransform one strategically important process end-to-end.+
Which process will prove the case to the whole organization?- Cross-functional squad
- Live operation
- Weekly value review
08Measure valueTrack outcomes, not usage.+
Did the process get faster, cheaper, better and safer?- Cycle time
- Cost
- Quality
- Revenue
- Customer outcomes
- Employee outcomes
- Risk
09ScaleIndustrialize reusable capabilities and expand to more processes.+
What can the next process reuse from the first?- Reusable agents
- Shared data products
- Playbooks
- Portfolio of processes
Where Should You Use Agentic AI?
Score a process on eleven criteria, or start from an example. The framework places it in one of four approaches.
- Recommended autonomy
- Readiness
Simplified heuristic for discussion. Horizontal axis averages complexity, coordination, handoffs, decision intensity and personalization. Vertical axis weighs business value twice and scalability once. High risk or limited readiness can change the result.
Where is your coherence bottleneck?
Rate eight dimensions from 1 to 5. Levels 1, 3 and 5 are described; 2 and 4 lie in between. Score how work actually runs today, not how it is designed on paper. The weakest dimension limits what AI can deliver, and your score feeds directly into the value simulator below.
Based on the Coherence Diagnostic Model from The Coherent Enterprise. Simplified self-assessment for discussion.
Compare three ways to apply AI to one process
Enter your own process figures. The defaults are example values, not benchmarks.
Flow · today and after
Friction per case. This is where agents create most of the speed.
Business case · conservative scenario
A CFO reads the conservative scenario first. Everything above it is upside.
How the estimate works
- Baseline hours = volume × processing time.
- GenAI assistance saves the assumed share of handling time on every case. Waiting time is unchanged.
- Agentic automation handles routine × automation potential of cases. These still need the human oversight share of today’s handling time, and finish in the set share of today’s cycle time. Remaining cases get the GenAI saving.
- Process reinvention adds the redesign improvement to both handling time and cycle time of the remaining cases.
- Coherence realization: the result above is multiplied by the share of the potential you capture. Your coherence score is half your weakest dimension and half your average. At full coherence you capture 100%. At the lowest score, GenAI assistance keeps 60%, agentic automation 30% and process reinvention 15%, because the more a process changes, the more it depends on coherence.
- Capacity in FTE assumes 1,600 productive hours per person per year.
- Handoffs: automated cases keep one handoff, to a person or a system. Redesign reduces the remaining handoffs by the redesign improvement. Flow improvements are also scaled by coherence.
- AI run cost per automated case is added to the yearly cost. Include change management and training in the one-off costs.
- Benefit = cost capacity released × conservative factor. Released capacity only becomes cash when it is redeployed, absorbs growth or avoids hiring.
- Year 1 realizes only the realization rate; the full benefit starts in year 2. Break-even is the first month in which cumulative net value turns positive.