Blog
TL;DR: We used to burn entire afternoons debating product decisions. Now we spend 20 minutes aligning on context, then let everyone explore solutions individually (both with and without AI). The result? Decisions in hours instead of weeks, engineers who understand product strategy and product managers who grasp technical constraints. Most importantly: we ship faster.
Product decisions have the capacity to consume meeting after meeting. Different disciplines defending their perspectives, talking past each other and eventually reaching compromise solutions that satisfied no one. Engineers focused on technical constraints, designers on user experience, product managers on business impact - all valid concerns, but hard to reconcile in real-time group discussion.
We've replaced this with a simple pattern: brief context alignment meetings followed by individual analysis and research. Teams independently converge on similar solutions through completely different reasoning paths. The unexpected outcome: this hasn't just accelerated our decision-making, it has also made us better at working together and increased our product velocity.
Consensus Convergence describes the shift from group debates to structured async decision-making, anchored by shared context and individual exploration. Instead of trying to reach consensus through discussion, we align on the problem and constraints, then pursue independent research and AI conversations before converging on compatible solutions.
The counterintuitive insight: teams agree more easily when they explore solutions separately (with AI) than when they debate face-to-face. The shared context ensures everyone's solving the same problem, while individual AI conversations allow each discipline to understand the decision through their own analytical lens.
This was the decision that started our entire approach. We were stuck in a meeting loop — 3+ meetings, 6+ hours total, circular discussions about which of our 12 crypto index methodologies to maintain vs retire. Engineers arguing about infrastructure costs, product pushing for client flexibility, UX citing industry standards.
Looking for a solution, we tried something different: a 25-minute alignment meeting where we focused on crafting an unbiased prompt. This was critical — with prompt engineering, you can make AI say whatever you want it to. The most important part wasn't agreeing on the context, but ensuring our prompt wouldn't lead the AI toward any predetermined conclusion.
Given these 5 specific methodologies (we shared the actual methodology documents), our current implementation timeline constraints, thousands of dollars in support costs, and our user segment analysis showing who actually cares about methodological differences — the core question became: "What can we do to increase our velocity and lower our cost without impacting users?" Crucially, we spent time ensuring this question didn't bias toward consolidation or any specific solution.
Individual AI Conversations (2 hours async on Slack):
Result: Saved thousands of dollars a month, freed engineering capacity, decision made in 2.5 hours total instead of weeks of meetings. Most importantly, this success convinced us to apply the same pattern to future decisions.
Six months and many successful decisions later, this process has become invaluable. Just last week, we needed to decide on the optimal UI pattern for our React Native CoinDesk mobile app. The process that once took us weeks now flows naturally.
The Old Way: 2+ meetings debating endless scroll vs pagination. Mobile team pushing for modern UX patterns, backend arguing for query efficiency, product asking for user behavior data that didn't exist yet.
The New Way: 20-minute context alignment: Market cap rankings for 5000+ assets:
We are shipping a React-Native CoinDesk mobile app targeting both crypto natives and traditional traders/investors migrating from equities, forex, and commodities. The app lists 50-4,000 crypto assets ranked by live market-cap—similar to how Bloomberg Terminal, TradingView, or Interactive Brokers display large asset universes. Prices and 24h % changes tick every second, but vertical rank order updates only on explicit user refresh (mimicking traditional market data terminals where rank stability is crucial for professional decision-making).
Individual AI Explorations (90 minutes async):
Result: Still pending as this was a meeting we had a few days ago. (We'll share updates in our next blog!)
Beyond faster decisions, this approach transformed how we work together:
What's remarkable is how different AI models approach the same context through completely different analytical frameworks, yet consistently reach compatible conclusions.
Claude tends to approach this with a systematic trade-off analysis, mathematical correlation, and structured decision frameworks — it excels at breaking down complex problems into logical components and weighing options methodically.
GPT-4 dives deep into implementation details, edge case exploration, and user experience mapping, often providing granular scenarios and detailed workflow considerations that others might miss.
Gemini focuses on technical feasibility assessment, performance optimization, and scalability concerns, bringing a practical engineering perspective to solution evaluation.
The shared context acts as a constraint that guides all models toward the same solution space, while allowing each team member to explore through their preferred reasoning approach. It's like having multiple expert advisors who reach the same recommendation through different paths — the systematic analyst, the detail-oriented implementer, and the pragmatic engineer all arriving at compatible conclusions because they're working from the same foundational understanding of the problem.
This convergence isn't coincidental; it suggests that when context is properly defined and unbiased, different analytical approaches naturally identify similar optimal solutions within the established constraints.
The transformation in our development speed has been dramatic, but it didn't happen overnight. Six months ago, identifying a product problem meant scheduling a discovery meeting for the following week. By the time we started building, the original problem had sometimes evolved or been deprioritized entirely.
Now, when we decided to add real-time price streaming to our asset rankings, we had context alignment and three AI-explored solutions posted on Slack within 4 hours. The engineer understood the UX implications of different scrolling patterns with live data updates, the UX designer grasped the database query implications of streaming data and the product manager researched how our competitors handled similar features.
Crucially, the competitive analysis revealed that most crypto apps suffered from crashes and reliability issues around real-time features, which reinforced our core USP: stability, accuracy, and timeliness. Since we were already touching the toplist infrastructure for streaming, we decided to improve the UX simultaneously while keeping things simple and efficient. We started implementation the next morning with complete confidence in our approach and a clear strategy to differentiate on reliability.
This velocity change cascades through everything we build. For our upcoming Q2 roadmap, instead of the usual three full-day planning sessions that still leave everyone uncertain about technical feasibility, we're planning to apply the same context alignment approach. Each feature will be pre-explored by relevant team members with AI, so we'll arrive at planning with solutions rather than just problems.
We expect engineering estimates to become more accurate because product requirements will be clearer from the start. UX designs should require fewer iterations because they'll be grounded in technical constraints upfront.
The psychological shift has been just as important as the time savings. There's genuine excitement about alignment sessions because everyone knows we'll walk away with clarity and direction. Engineers report higher job satisfaction from spending 70% of their time coding instead of 40% - 50%. Product managers can focus on strategy and user research rather than facilitating endless requirement debates. The UX team has more time for user testing and design iteration because they're not constantly reworking solutions based on last-minute constraint discoveries.
Most importantly, this approach scales with team growth. When we hired two new engineers last week, they could understand our past decisions by reading the AI conversation threads rather than requiring lengthy knowledge transfer sessions. When we expanded into options trading exchange metadata features, the same context-alignment pattern worked perfectly for a domain none of us had deep expertise in. The AI conversations had simultaneously became our institutional memory and onboarding documentation.
The process is straightforward but requires discipline:
Context Template:
Context: [Problem background, current state, and technical environment]
Options to Evaluate: [Specific alternatives being considered with detailed descriptions]
Analysis Framework: [What perspective to take, what factors to consider]
Key Evaluation Areas: [Specific dimensions for assessment]
Research Required: [External benchmarking, competitive analysis, industry standards needed]
Deliverables: [Expected outputs including analysis format, research components, and final recommendation structure]
Personal Tips:
The traditional tech industry mantra of "disagree and commit" assumes that teams will inevitably reach impasses requiring hierarchical resolution. Someone makes the call, others commit despite reservations, and execution suffers from lukewarm buy-in. But most product debates aren't about fundamental disagreements over priorities - they're about implementation details where everyone wants the same outcome but sees different paths to get there.
Consensus Convergence works best for this specific class of decisions: when teams agree on the problem and success criteria but disagree on approach.
The mobile UI case wasn't about whether to prioritize user experience - everyone wanted that. The disagreement was over which UX pattern would actually deliver it.
The index methodology decision wasn't about whether to reduce costs - that was universally desired. The debate was over which consolidation approach would achieve savings without user impact.
The critical insight is that the shared context and prompt become collaborative artifacts, not imposed constraints. When the team collectively crafts the problem definition, success criteria, and evaluation framework, they're not just aligning on what to ask the AI — they're explicitly codifying their shared priorities. The engineer who helps write "performance requirements <50ms load" can't later claim the solution unfairly prioritizes speed over functionality. The designer who contributes "visual stability critical for traders" has already acknowledged this constraint as legitimate.
This collaborative prompt engineering serves as a forcing function for surfacing hidden disagreements early. If someone objects to including "maintain current user workflows" in the context, that reveals a fundamental priority conflict that needs resolution before any solution exploration. By the time individual AI conversations begin, the team has already worked through their core disagreements to create shared evaluation criteria.
This approach has clear limitations. It doesn't resolve conflicts over resource allocation, strategic priorities, or fundamental technical philosophies. When engineers and product managers disagree about whether to prioritize new features or technical debt, no amount of AI conversation will bridge that gap. But for the vast majority of day-to-day product decisions — where teams share goals but debate methods — it transforms resistant compliance into genuine advocacy.
The ownership that emerges is real because the constraints are shared, the analysis is independent, and the conclusions are self-reached. That's the foundation for execution velocity that matches decision velocity.
Consensus Convergence isn't about replacing human judgment with AI recommendations — it's about using AI to enable better collaboration within shared constraints. We've discovered that consensus emerges more naturally when people can explore solutions individually with AI partners, rather than defending positions in group settings.
This approach fundamentally shifts how we use our expertise within the team. Instead of burning expert colleagues' time during the learning and exploration phase, we use AI as individual research assistants to rapidly understand domains outside our core competencies. The engineer uses Claude to grasp UX implications, the designer uses GPT-4 to understand technical constraints, the product manager uses Gemini to research competitive landscapes. Only after this individual knowledge-building phase do we engage each other's expertise — but now for validation, refinement, and edge case identification rather than basic education.
The AI isn't running our company or making our decisions. It's accelerating each team member's ability to understand the full problem space before we collaborate. When the engineer posts their AI conversation about UX patterns, they're not saying "Claude decided this" — they're saying "I used Claude to research industry standards and user behavior patterns, here's what I learned, and here's my informed recommendation." The difference is crucial: expert human judgment backed by AI-accelerated research, not AI judgment validated by human approval.
The most valuable outcome isn't just faster decisions — it's the improved team dynamics and higher product velocity. Engineers understand product strategy better, product managers grasp technical constraints, and everyone spends more time building instead of debating.We've moved from "let's schedule a meeting to discuss this" to "let's align on context and explore with AI." The result is decisions in hours instead of weeks, better cross-team understanding, and significantly higher product velocity. That's the kind of convergence that scales teams and products.
Consensus Convergence (n.) — The phenomenon where teams replace lengthy product debates by agreeing on shared context and prompts in a single meeting, then pursuing individual AI conversations that independently arrive at similar conclusions — leading to faster decisions, better cross-team understanding, and higher product velocity.
Get our latest research, reports and event news delivered straight to your inbox.
