Updated April 2026

Agile in the Age of AI

The rules have changed. In an AI-augmented world, our traditional agile practices need to evolve. This is a guide to navigating that transformation—from hypothesis-driven development to measuring what actually matters.

Kasie Lombardi

Principal Agile Coach

The Vero Flow System

The Vero Flow System is a lightweight, high-velocity agile framework specifically engineered for SaaS teams building AI-native solutions. It replaces traditional process "toil" with a lean, discovery-to-delivery loop that prioritizes Time to Truth over traditional velocity or throughput.

The Core Philosophy

We stop measuring "how much" we ship and start measuring "how quickly" we validate a hypothesis. The goal is to reach Minimum Viable Intelligence (MVI) with the least amount of capital possible.

Why Change?

The catalyst for Vero Flow comes from three fundamental shifts happening in how we build products:

Shift in Roles: From Producers to Orchestrators

Traditional

Experts (PMs, Designers, Engineers) are the primary source of manual labor, executing tasks themselves.

Vero Flow

The Human-in-the-loop becomes an orchestrator of intelligence, directing virtual AI agents rather than doing all the work manually.

Shift in Collaboration: From Hand-offs to Co-Piloting

Traditional

Research, Design, and Engineering work in silos with sequential hand-offs—a relay race approach.

Vero Flow

These roles co-pilot the Signal and Vetting phases simultaneously. Engineering capital is only deployed once we have directional clarity on the core hypothesis.

Shift in Focus: From Output Velocity to Time to Truth

Traditional

Success measured by how much we ship—story points completed, features deployed.

Vero Flow

Success measured by how quickly we validate a hypothesis—reaching Minimum Viable Intelligence with minimal capital.

The New Bottleneck

Engineering is no longer the blocker. By leveraging an AI-agent workforce and thin-slicing work into 2-4 day cycles, we've effectively removed the traditional development bottleneck. The new friction points are downstream dependencies: Security, Legal, Marketing, and Sales. Flow isn't just an engineering change—it's an organizational mandate.

MVI: Minimum Viable Intelligence

The MVI is the smallest possible experiment designed to gather holistic intelligenceon a hypothesis. It is not a "lite" version of a product; it is a "high-density" search for truth.

MVP vs MVI

Traditional

MVP asks: 'Can we build a functional version of this feature?' Focus on technical implementation, basic usability, and shipping something. Risk: High capital expenditure before understanding the problem.

Vero Flow

MVI asks: 'Do we have enough information to justify further capital investment?' Focus on holistic signal gathering across User, Logic, and Economic intelligence.

We reach MVI when we have sufficient evidence to confirm that:

1
User Intelligence
Does this solve the documented friction point in a way the user actually values?
2
Logic Intelligence
Does the AI reasoning or system logic consistently produce the intended outcome?
3
Economic Intelligence
Is the cost to provide this intelligence (tokens/compute) lower than the value it generates?

From 'Can we build it?' to 'What do we know?'

•Truth over Features: We stop measuring success by deployment and start measuring by Confidence Score.
•Capital Protection: The MVI allows us to reach a "Fast No" in 3 days, saving the months of salary and opportunity cost typically wasted on unvetted MVPs.
•Directional Clarity: An MVI doesn't need to be bulletproof—it just needs to provide enough holistic signal to move from Proving to Scaling with confidence.

The Fundamentals

The Vero Flow System is built on four core principles:

Hypothesis over Task

Work is framed as experiments: "If we do X, then Y will change by B%."

Outcome over Output

Success is measured by customer success metrics, not the volume of code commits.

Tiered Intelligence

Prioritize cost-efficiency by starting with basic models. Reserve high-powered AI for tasks where advanced logic is essential.

Automation by Default

Any repeatable management task (Jira updates, test data generation, etc.) is handled by AI agents.

Desired Outcomes

Minimized Time to Truth

Shorten the duration between a research-backed hypothesis and a validated market result.

High-Confidence Prioritization

Ensure engineering efforts solve documented customer friction points validated by UX and market research.

Baked-in SaaS Profitability

Integrate real-time ROI and inference cost tracking, ensuring every AI feature contributes to healthy gross margin.

Process Toil Elimination

Leverage AI agents to automate administrative management, freeing human experts for high-value decisions.

Measurements of Success

Discovery Rate

Capital protection: Filters bad ideas early

How many bad ideas did we kill before they cost us engineering hours?

Time to Truth

Market agility: Measures discovery speed

Days from a research signal to a validated MVI (Minimum Viable Intelligence).

Token Efficiency

Gross margin health: Tracks SaaS profitability

Ensuring feature revenue exceeds our AI inference costs.

Reliability

High-level creative talent is consumed by manual UI production and time-intensive high-fidelity iterations, often downstream and isolated from the initial problem-discovery loop.

Vero Flow

The Designer moves upstream as a strategic co-pilot. They partner with the PM at the first 'Signal' to vet the core user logic and journey architecture before any production work begins.

The Change: They direct an automated creative engine to handle fidelity, focusing their expertise on the Human-AI relationship and significantly accelerating the Time to MVI.

The Framework in Practice

Beyond the philosophy, Vero Flow provides concrete ceremonies and practices. Here's how the day-to-day work flows:

Work Breakdown / Backlog Refinement

Purpose: Ensure the backlog is clear, well-ordered, and ready for work to be pulled, minimizing delays and rework later.

Cadence: Just in time / As needed—should happen at least weekly. Triggers: when an Epic is ready, when the Ready queue is low.

•Decompose work into small, flowing pieces; large items get stuck.
•Clarify requirements and proactively address dependencies to prevent blockers.
•Remove ambiguity to prevent rework or delays in "In Progress" items.
•Maintain a "Ready" state with a buffer of items for continuous team pull.
•Align with WIP limits by sizing items appropriately for system capacity.

Queue Replenishment

Purpose: Select and pull new work into the system, ensuring the right work is prioritized and WIP limits are respected.

Cadence: Just in time / As needed—should happen at least weekly. Triggers: when new work has been refined, when the Ready queue is low.

•Prioritize work for value, goals, and improved overall flow.
•Strictly respect WIP limits; pull new work only with available capacity.
•Analyze upstream flow to ensure item readiness for smooth pull.
•Balance demand and capacity to prevent bottlenecks.

Daily Stand-ups

Goal: Synchronize activities, identify impediments, and update the Kanban board.

•Walk the board right-to-left, prioritizing work closest to completion.
•Identify blockers immediately and address them urgently (use "Days in column" in Jira).
•Ask "What's preventing this from moving?" to focus on flow, not just status.
•Limit discussions to flow, deferring detailed problem-solving.
•Visualize flow daily and discuss any stagnation.
•Look for triggers, scheduling work breakdown or queue replenishment as needed.

Domain Review and Feedback Session

Goal: Solicit feedback and ideas that help the team maximize the value of the features they are building.

Cadence: On a regular cadence that enables the team to continue to move forward quickly. Common approach is weekly.

•Clarify purpose: Ensure attendees understand objectives.
•Support PMs: Ensure correct stakeholder invitation and attendance.
•Coach the "Why": Emphasize feedback and validation over mere reporting.
•Curate the "What": Review in-flight work for feedback, learnings from experiments, and demonstrate done work.
•Craft the Narrative: Help teams tell the story of what was learned, not just value delivered.

Retrospective

Goal: Inspect and adapt the Kanban system and the team's way of working to improve flow.

Cadence: Weekly or bi-weekly for 30-60 minutes.

•Review flow metrics (Cycle Time, Throughput, WIP, Flow Distribution) for trends and concerns.
•Look for specific concerns like cycle time outliers, high RTB/Risk/Defects.
•Discuss stalled/blocked work to identify root causes and prevention strategies.
•Identify bottlenecks by pinpointing work accumulation and WIP limit adherence.
•Propose flow improvement experiments for smoother workflow and reduced cycle time.
•Focus on process refinement to enhance speed, predictability, and quality of flow.
•Share learnings focusing on what you learned from the most recent experiments.

A Note on Estimating and Forecasting

We recommend moving away from story points in favor of thin-slicing and empirical forecasting:

•Work items should be thin-slices of iterative value that can flow through the system in 2-4 days.
•T-Shirt sizing is recommended for Epics to help inform prioritization, staffing, and sequencing.
•Burn up/Burn down charts using Monte Carlo simulations should be used for forecasting delivery timelines.

Why? AI behavior is unpredictable and much of what we are doing is greenfield. Time spent debating estimates can be better focused on Time to Truth. Points can create "commitment bias", encouraging teams to finish work rather than pivot when needed.

The Acceleration Gap

As we accelerate engineering velocity, we must address the bottleneck displacement effect that naturally occurs across the organization.

The Reality of High-Velocity Engineering

•The Displacement Effect: By leveraging an AI agent workforce and the MVI mindset, we have successfully cleared the traditional engineering bottleneck. Code and core logic are now proven in days.
•The Downstream Dam: As technical velocity increases, the friction points naturally slide to ancillary teams (Security, Compliance, Marketing, Brand, and Sales).
•The Organizational Risk: If it takes 3 days to find a "Truth" but 3 weeks to clear cross-functional alignment, the domain's ultimate Time to Value remains stalled. We must modernize the rest of the business to interface with a high-velocity product engine.

The Strategy: Shift from Permission to Guardrails

From Gates to Tracks:We move away from making ancillary teams act as downstream "traffic cops" who manually inspect every feature. Instead, they become upstream "highway architects" who build the safe tracks our teams sprint on.

Cross-Functional Enablement Strategies

Security & Compliance: Automated Guardrails over Rigid Reviews

•Pre-Approved Security "Wrappers": Define strict data and environment boundaries for the Proving (MVI) phase. If an experiment uses synthetic data, anonymized telemetry, or isolates itself from core PII, it is auto-cleared to bypass traditional review cycles.
•Compliance Linting via AI: Embed corporate compliance and data-privacy frameworks directly into our engineers' AI coding agents. Security policies become automated code guardrails rather than late-stage manual audits.

Marketing & Brand: Embedded Representation & Automated Toolkits

•Dedicated "Flow" Marketers: Embed specialized Marketing and Brand advocates directly into the Vetting phase. They don't wait for a release note; they deeply understand the product logic and user friction from Day One.
•Dynamic Brand Toolkits: Give the domain's AI creative engine pre-approved tone-of-voice, copy constraints, and UI asset kits. This allows teams to safely auto-generate micro-copy and feature release messages without violating brand integrity.

Sales: Driving "Time to Revenue"

•The Co-Innovation Client Vault: Establish a pre-vetted cohort of trusted enterprise customers who have signed broad master waivers to test unpolished, MVI-stage software. Sales can place new experiments in front of these design partners instantly.
•Evidence Telemetry for Reps: Feed real-time user-impact data from successful MVIs directly back into the Sales channel. Sales reps can sell immediate, evidence-backed resolutions to customer pain points rather than waiting for formal marketing case studies.

Leading the Change

Transformation requires leadership buy-in. Here's what we need from leaders to make Vero Flow succeed:

Outcome over Output

Change the narrative from "When will Feature X be done?" to "What did we learn about problem Y this week?"

Permission to Pivot

Support teams when they kill an idea in the prove phase. A "Fast No" saves capital.

Trust in Agency

Give your teams the destination, then step back and let them navigate.

Accountability Shift

PMs are accountable for the quality of the problem, EMs for the health of the system, and Engineers/Teams for time to truth and reaching MVI as fast as possible.

Closing Thoughts

The age of AI demands a fundamental rethinking of how we build products. The old metrics—story points, velocity, features shipped—optimized for a world where engineering was the bottleneck and building was the hard part.

Today, building is fast. Discovery is what matters. The teams that win will be the ones who can find truth faster, protect capital better, and pivot with confidence when the data says they should.

The Vero Flow System isn't just a new process—it's a new mindset. It asks us to value learning over launching, truth over throughput, and intelligence over output.