Skip to main content

Dealing with Ambiguity

Practical approaches for navigating unclear goals, evolving customer needs, and uncertain situations. This is a core skill for senior+ roles, especially in Applied AI where requirements shift fast and the tech landscape evolves rapidly.

Understanding Ambiguity​

Ambiguity usually comes from:

  • Unclear goals - Unclear objectives or success criteria
  • Evolving customer needs - Changing requirements and expectations
  • Immature technology - New technologies with uncertain capabilities

The key is to not freeze in ambiguous situations—instead, clarify what is known and start from there.

Core Approach​

1. Recognize the Ambiguity Early​

Identify ambiguity quickly and acknowledge what you don't know.

Example: "We weren't sure whether customers wanted natural language prompts or pre-built templates, so I scoped a lightweight experiment to test both."

2. Anchor to Outcomes, Not Solutions​

Focus on business/user value rather than technical purity.

Key Questions:

  • What problem are we solving?
  • What metric matters here?

Example: "Rather than debate LLM architectures, we focused on reducing the time it took a user to create their first automation from 20 minutes to under 5."

3. Use Iteration to Reduce Uncertainty​

Build small, testable prototypes and use data to guide decisions.

Approach:

  • Run experiments as soon as possible
  • Use metrics, customer feedback, or A/B testing
  • Let data inform next steps
  • Dive deep into results and iterate

Example: "I shipped a rough prototype to 50 beta users in under 2 weeks, then pivoted based on their usage patterns."

4. Collaborate & Re-align Constantly​

Partner with PMs, designers, and other engineers to reframe scope as needs shift.

Communication:

  • Communicate tradeoffs clearly
  • Show humility and flexibility
  • Don't be married to a single solution

Example: "If we go with option A, we deliver faster but risk scaling issues. If B, we're slower but future-proof."

5. Adapt Through Continuous Learning​

Be explicit about lessons learned from pivots and changes.

Example: "Initially we over-invested in fine-tuning, but customer adoption lagged. I learned to validate product-market fit before scaling model work."

Real-World Examples​

Search Customer Experience​

Problem: People were abandoning their search requests on amazon.com for fashion products. How to solve it was not clear—we didn't have clear goals.

Approach: Ran multiple experiments that taught us:

  • How people perceive ads
  • The impact of placing a search widget at the top/middle/bottom of a page
  • The impact of getting the images right on a widget

Outcome: Data-driven insights informed the solution approach.

Profile of 1​

Another example of navigating ambiguity through experimentation and iteration.

Interview Framework​

Sample Answer Structure:

"In applied AI, ambiguity is the default. One example was when we were building an LLM-powered feature, but we weren't sure which workflow would resonate with customers. Instead of debating internally, I proposed a lightweight prototype to validate assumptions quickly.

I worked with PM to define success as improving task completion time. We launched two small variants to a pilot group, measured usage, and collected direct customer feedback. When results showed customers wanted structured templates more than free-form prompting, we pivoted.

The outcome was a production-ready feature that reduced onboarding friction by 40%. More importantly, the process taught me how to reduce risk by running small experiments early, and how to stay aligned with product even as requirements changed."

Key Principles​

  • Ground things in the customer - Let customer data and behavior influence decisions
  • Use POCs/experiments - Have them inform next steps
  • Be outcome-focused - Focus on what matters to users and business
  • Do the best with what you have and evolve - Start with what's known and iterate