In the work of modern delivery, your first job is not building. It’s navigating the fog—the layered uncertainty, conflicting signals, and organizational complexity. Only then can we identify the confident first next step toward value.
This white paper explores a second order of Foundational Frameworks on how to scope confidently into delivery using three of the most powerful mental models available:
- Knightian Uncertainty (Four Types of Uncertainty) – Clarifies what kind of unknown you’re facing.
- Cynefin Framework – Positions the work within the right system type: simple, complicated, complex, or chaotic.
- Bayesian Updating – Offers a disciplined method to iteratively reduce uncertainty and gain confidence as we move forward.
Together, these form a decision-making compass for transformation: when to anchor, when to probe, when to pivot—and how to earn confidence through iteration.
Navigating Complexity is the Job
The modern technology consultant is not just a builder. They are a navigator of ambiguity, working through competing incentives, tangled legacy systems, and ever-changing constraints. Confidence doesn’t come from clarity upfront—it comes from a process that generates it.
This is where our three frameworks come in. Let’s define them first, then show how they operate together to scope toward delivery.
1. The Four Types of Uncertainty (Knightian Framework)
This framework classifies uncertainty into:
| Type | Description |
|---|---|
| Known Knowns | We know the variables and outcomes. |
| Known Unknowns | We know what we don’t know. We can ask the right questions. |
| Unknown Knowns | Institutional knowledge exists, but it’s hidden or siloed. |
| Unknown Unknowns | We don’t know what we don’t know—new risks, new dynamics. |
Use this early in discovery. Ask: What type of uncertainty are we facing?
This becomes your heatmap for what can be scoped confidently, what needs research, and where a probe is required instead of a plan.
2. The Cynefin Framework (Sensemaking in Systems)
Cynefin helps us understand the type of system we are working in:
- Obvious: Best practices apply. Deliver.
- Complicated: Experts can analyze. Plan and then deliver.
- Complex: Cause and effect only clear in hindsight. Probe > Sense > Respond.
- Chaotic: No relationship between cause and effect. Act to establish order.
- Apex: Confusion: When the system is misdiagnosed.
Use this to frame the delivery environment and select the appropriate next step.
Cynefin doesn’t tell you what to do. It tells you what kind of doing is required. Complex systems don’t need more analysis. They need action to generate learning.
3. Bayesian Updating (Iterative Confidence Building)
Bayesian Updating is the backbone of modern product iteration. You form a hypothesis about what will work, test it, and update your belief based on results.
- Start with a prior belief: “I think this will work.”
- Run a lightweight probe.
- Update your confidence: “How did reality respond?”
This is the method for scoping confidently, even in ambiguity. Not because we’re sure—but because we’re learning fast.
The mindset here is not “we need full certainty before we build,” but: “we can afford to learn our way into certainty.”
The Loop: Scoping into Delivery
Here’s how these models work together in a loop:
- Clarify the Uncertainty
→ Use Knightian to define knowns/unknowns and surface what’s hidden. - Classify the System
→ Use Cynefin to understand the domain: is this a probe, a plan, or a pivot? - Design the First Probe
→ Use Bayesian thinking to define a hypothesis and a small, fast way to test it. - Update Confidence
→ Use the results to shift the plan. Shrink the scope or move forward. - Repeat at Smaller Scope
→ Zoom in: run this again at the sub-component or team level.
This loop continues until you’ve carved down the problem to an MVP of the MVP—a first, most valuable deliverable.
Case Example: AI-Powered Legacy Automation
Let’s say a client wants to use GenAI to automate a manual claims process in their legacy finance system.
Knightian Scan:
- Known knowns: Manual steps, claim rules.
- Known unknowns: Volume and edge case frequency.
- Unknown knowns: Tribal knowledge in staff about exceptions.
- Unknown unknowns: Hallucination risks, new AI behaviors.
Cynefin Frame:
- The base process is Complicated (analyzable).
- The AI integration is Complex (emergent).
→ Plan the known; probe the AI behavior.
Bayesian Probe:
- Pilot AI on 100 claim samples.
- Hypothesis: “AI can process 80% of claims without escalation.”
- Result: 60% success. Edge cases more diverse than expected.
Update and Iterate:
- New scope: Focus AI on first 2 of 5 steps.
- Design UI for human escalation on the rest.
- Confidence increases. Time to build v1.
Final Thought: Complexity Rewards Process
We often ask “Can we scope this?” The better question is: Do we have a process for scoping it, repeatedly, with increasing confidence?
Mastering delivery is not about knowing the answer.
It’s about having a method for finding it fast, small, and safely.
Using Knightian uncertainty to understand what you’re facing, Cynefin to guide how you act, and Bayesian updating to improve as you go—you build the muscle not just to deliver, but to navigate complexity with confidence.