CTRS Framework • Pillar 1
Decision Velocity
The metric that measures how quickly intelligence becomes action in your enterprise. When AI optimizes for accuracy but ignores decision latency, you've built the wrong system.
The Decision Velocity Formula
Decisions Made
Time to Execute
=
Decision
Velocity
Speed Index
Actual vs. target decision latency. How quickly can you act on intelligence? Measures time from signal to action.
Accuracy Index
Net decision quality (benefit minus cost) vs. target. Are decisions generating value? Captures economic impact.
Resilience Index
KPI stability under drift, outages, and shocks vs. SLO. Can you maintain performance when systems degrade?
Coverage
Share of eligible volume under governed automation. How much of the opportunity are you capturing with safe guardrails?
The $50 Billion Blind Spot
A major bank’s fraud detection model reaches 97.3% accuracy. The bank still loses $2 million every month. Why? Because no one has clear authority to block transactions when the model flags risk.
This is the paradox. Enterprises have mastered building predictive models but remain incapable of making decisions based on those predictions. The result: sophisticated, expensive shelfware with no impact on the P&L.
90%
of AI initiatives fail to meet objectives
70-90%
stall in "pilot purgatory"
$0
P&L impact from most AI investments
How to Measure Decision Velocity
Every decision emits telemetry: timestamps, inputs, model versions, policy versions, decision path, reason codes, outcomes, and SLA compliance. This powers real-time Decision Velocity dashboards and audit-ready replay.
Example: Fraud Detection Stream
Speed: Target 250ms (p50), actual 300ms → Speed Index = 250/300 = 0.83
Accuracy: Target 95%, actual 93% → Accuracy Index = 0.98
Resilience: Target <10% degradation under shock, observed 8% → Resilience Index = 1.00
Coverage: 65% of eligible transactions under governed automation → Coverage = 0.65
DVI = 0.83 × 0.98 × 1.00 × 0.65 = 0.53
Actions: Improve feature freshness to hit 250ms target. Lift straight-through processing in safe segments to 75%. Maintain resilience guardrails.
Board Dashboard Per Decision Stream
For each stream, show current vs. target for Speed, Accuracy, Resilience, and Coverage. Include supporting metrics on adoption (usage rates, bypass rates, override rates) and ROI (net dollar impact, payback period). Display 6-12 month trendlines and a DVI gauge.
Publish Decision Velocity monthly in the board pack alongside ROE and EBITDA. Track trendlines. Only promote autonomy when Speed, Accuracy, and Resilience meet thresholds. Coverage follows, not leads.
4 minutes
Average decision latency
$3.2M
Monthly P&L impact
99.7%
Uptime under production load
65%
Coverage with governance
The 4D Decision Intelligence Cycle
Decision Intelligence is the discipline of turning information into better actions. Rather than starting with models or data, start with the decision: its objectives, owner, constraints, and acceptable risk.
1
Define the Decision
Name the decision explicitly. Identify the accountable owner. Define objectives, risk tolerance, and decision SLAs.
- Name decision & owner
- Set objectives & SLAs
- Document economic value
2
Design the Decision Flow
Map inputs, decision logic, and human-in-the-loop touchpoints. Build in governance hooks from the start.
- Map inputs & logic
- Design HIL touchpoints
- Build governance hooks
3
Deploy the Action
Trigger interventions in operational systems. Ensure traceability. Capture overrides for learning.
- Trigger interventions
- Link actions to evidence
- Capture overrides
4
Defend & Learn
Track decision KPIs. Compare outcomes to objectives. Feed insights back into the cycle to improve continuously.
- Track decision KPIs
- Compare actual vs. target
- Institutionalize learning
Continuous
Decision Intelligence
Start With One Money Decision
Don't boil the ocean. Select one high-value decision with clear P&L or risk impact, latency sensitivity, an actionable endpoint, and sufficiently reliable inputs. If you can't summarize the decision owner, SLA, and evidence plan on one page, it's not the right starting point.
90-Day Implementation Roadmap
1
Days 1-30
Discover & Align
- Appoint executive sponsor
- Identify one money decision
- Form cross-functional DecisionOps team
- Conduct decision audit: map current flow, latency, cost, error rates
2
Days 31-60
Design & Validate
- Codify future state with Decision Canvas
- Document ownership, inputs, governance, KPIs
- Validate new logic in digital twin
- Deploy to shadow mode for testing
3
Days 61-90
Deploy & Defend
- Go live with limited traffic
- Establish Decision Review Board
- Track overrides as learning source
- Measure Decision Velocity baseline
Expected Outcome
20-35% latency reduction • 12-18 month payback • Proven decision factory model
Once proven, expand to the next 3-5 money decisions using the same playbook. Institutionalize the Decision Value Chain: discover, design, deploy, defend.
Related Frameworks
Decision Velocity is Pillar 1 of the CTRS framework. Explore the other pillars that enable enterprise reasoning at scale.
Control Tower Reasoning System
The complete framework for enterprise reasoning systems with compliance built in. Integrates Decision Velocity, Trust Layer, and Agent Orchestration.
Trust Layer & Version Control
Prevent Version Drift. Ensure AI retrieves the correct policy version every time. Critical for regulated environments where citing wrong policy carries liability.
Agent Orchestration via MCP
Autonomous agents with governance. The Model Context Protocol provides standardized context management for multi-agent coordination with built-in accountability.
Calculate Your Decision Velocity
Most organizations discover they're optimizing the wrong metrics. Map your high-value decisions and measure what actually moves the P&L.