Qodo is introducing Approval Workflows in beta and Agent Session Review in research preview, operationalizing an 'earned autonomy' framework that gives engineering teams explicit, policy-driven controls over AI coding agents [1]. The releases address a critical governance gap: 75% of organizations have experienced a production incident with AI as a contributing factor, yet only 43% mandate human review of AI-generated code [2]. With 55.4% of AI decision-makers citing agent reliability and hallucination management as a top adoption challenge [3], Qodo's governance-first approach targets the oversight deficit that threatens to stall agentic SDLC adoption at scale.
What is Covered in this Article
- The human review bottleneck in agentic software development [1]
- Qodo's L1-L4 earned autonomy maturity model [1]
- Approval Workflows beta: policy-driven automated PR approvals [1]
- Agent Session Review research preview: cross-PR inspectability [1]
- The enterprise governance gap and market imperative [2][3]
The News: Qodo has announced two new capabilities targeting the oversight gap in agentic software development [1]. Approval Workflows, now in beta, lets teams encode recurring risk criteria, including blast radius, security findings, and rule violations, so low-risk pull requests proceed to merge automatically while higher-stakes changes stay in the human queue [1]. Agent Session Review, in research preview, consolidates an agent's work across PRs and repositories into a single assessment, connecting requirements, coding-session context, implementation, and Qodo's review findings, while surfacing behavioral decisions, departures from plan, and execution paths [1]. Both features advance Qodo's four-level L1-L4 autonomy framework, which defines how much responsibility a system can carry without human intervention at each maturity stage [1].
Qodo Builds the Governance Layer Agentic SDLC Has Been Missing
Analysis: Qodo's announcement addresses a structural problem that has quietly undermined agentic AI's productivity promise: the human review queue treats routine approvals and high-judgment architectural decisions as equivalent work [1]. As Qodo's own author notes, for several features built with coding agents, understanding the output and getting it into production took longer than developing the feature itself [1]. Approval Workflows and Agent Session Review represent a deliberate effort to separate those two categories of attention and govern each appropriately [1].
The Bottleneck Agentic Development Creates
Coding agents accelerate multi-repo, multi-component work, but they push the resulting complexity downstream into the human review queue. A reviewed, low-risk change waits alongside a feature spanning several PRs that requires someone to reconstruct behavior, dependencies, and risks from scattered diffs. Both demand human time, but only one demands human judgment. This conflation erodes the productivity gains agents are meant to deliver. Futurum research confirms the stakes: productivity improvement is the top metric organizations use to measure AI success, cited by 55.1% of respondents [3], and software engineering code generation and development assistance ranks as a top-five GenAI use case at 46.8% [3]. When the review queue becomes the constraint, the agent's speed advantage disappears at the handoff.
Earned Autonomy as Structured Policy
Qodo's L1-L4 maturity model gives teams a structured vocabulary for delegation decisions [1]. L1 covers AI-assisted individual tasks; L2 describes human-agent pairing; L3 represents agentic SDLC workflows where humans remain essential to completion; L4 describes a lights-on software factory that completes a defined class of work through verification, approval, and release within delegated authority, containing problems autonomously before escalating with evidence [1]. Critically, the framework is scoped: a team can operate at L4 for some workflows while keeping others at L3. This granularity matters because it grounds delegation in explicit policy rather than implicit trust, giving engineers a defensible basis for expanding or contracting the system's authority as evidence accumulates [1].
Approval Workflows: Making Judgment Repeatable
Approval Workflows makes the L3-to-L4 transition actionable for pull request approvals [1]. Teams define which findings, risks, and rule violations require human attention. A team might permit minor maintainability findings but require human review for any security finding, or restrict automatic approval to changes with a small blast radius. Changes meeting those criteria proceed without waiting for a person to repeat the same assessment. The authority structure is preserved: approval remains conditional on the code meeting defined standards, and the reasoning behind each automated decision stays available for inspection. This design aligns with Futurum's own framework for agentic trust, which requires transparency about when and how AI is being used, explainability into how conclusions are reached, and a strong audit trail with human-in-the-loop accountability [4].
Agent Session Review: Inspectability at Scale
Agent Session Review addresses the complementary problem of understanding what an agent actually decided [1]. An agent-built feature may span a backend service, an API, and a client, with changes distributed across several PRs. Reading each diff in isolation leaves engineers to reconstruct the agent's reasoning themselves. Agent Session Review surfaces behavioral and technical decisions, highlights departures from the original plan, and traces execution paths through the code in a unified view [1]. This directly supports the kind of human oversight Futurum has identified as essential: vendors must deliver explainability, giving users a clear view into how an AI reached its conclusion, alongside a human-in-the-loop approach that preserves final control and accountability [4]. The feature also creates a feedback loop: recurring architectural corrections surfaced in session reviews become candidates for explicit standards, tightening the factory's rules over time [1].
Market Timing and the Governance Gap
The governance gap Qodo is targeting is acute. Futurum research documents that 75% of organizations experienced a production incident in the last 12 months with AI as a contributing factor, yet only 43% mandate human review of AI-generated code, and agent governance ranks as the least-mature SLE practice [2]. Meanwhile, 55.4% of AI decision-makers cite agent reliability and hallucination management in production as a top adoption challenge [3], and autonomous coding and software engineering is a leading planned deployment area for agentic AI at 39.6% [3]. The AI platforms market is projected to reach $5.3 trillion by 2030 at a CAGR of 52.5% from 2025 [5]. As Futurum has noted, AI tracking and governance dashboards are becoming table stakes, with vendors announcing agent observability and policy-management consoles to give enterprises the confidence to scale autonomous AI [4]. Qodo's releases position it squarely in that emerging category.
What to Watch
- Approval Workflows adoption curve: which engineering team profiles, platform-scale versus startup, move from beta to production deployment first and at what velocity [1]
- Governance gap closure: whether the 43% human-review mandate rate rises materially as policy-driven tools such as Approval Workflows lower the implementation cost of oversight [2]
- Competitive response: how adjacent agentic SDLC vendors reprice, repackage, or accelerate their own observability and policy-management consoles in Q1 2027 [4]
- Agent Session Review progression: whether the research preview advances to general availability and which cross-repo complexity thresholds drive the strongest enterprise uptake [1]
- Feedback loop maturity: whether teams using approval rate tracking alongside defect and compliance metrics demonstrably expand their delegated autonomy scope in the first half of 2027 [1]
Sources
1. Earned Autonomy: Expanding What We Trust the Software Factory to Do, Qodo, November 2026
2. The AI Stack: How Vendors Are Composing AI Strategy, Futurum Research, July 2026
3. 2H 2026 AI Platforms Global Enterprise Decision Maker Survey Report, Futurum Research, October 2026
4. Agentic AI Expansion Across SDLC – Building Trust in AI, Futurum Research, October 2025
5. 2H 2026 AI Platforms Market Sizing & Five-Year Forecast, Futurum Research, November 2026
Disclosure: Futurum is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article.
Read the full Futurum Group Disclosure.
Other Insights from Futurum:
Qodo v3.1: Governing AI Code Review Before Noise Kills Adoption
Qodo 3.0 Puts Governance at the Center of Agentic Code
AI Code Generation Scaled. Verification Didn't.
Author Information
This content is written by a commercial general-purpose language model (LLM) along with the Futurum Intelligence Platform, and has not been curated or reviewed by editors. Due to the inherent limitations in using AI tools, please consider the probability of error. The accuracy, completeness, or timeliness of this content cannot be guaranteed. It is generated on the date indicated at the top of the page, based on the content available, and it may be automatically updated as new content becomes available. The content does not consider any other information or perform any independent analysis.

