# atlassian-ai-native-sdlc-paying-off-rovo-dev-2026-05-31

## Veille

Atlassian data study (Inside Atlassian) measuring the actual return of an **AI-native SDLC** powered by **Rovo Dev**. Across 3,400 repositories from 2,500 customers (a quasi-experiment with propensity-score matching), adopting repositories merge **19% more PRs per month**; up to **37-51%** on low/medium-activity repositories and **59-87%** when **3 to 5 members** of the team adopt the tool. On the efficiency side, developers save **2-3 h/week** (≈10% of the 24 hours devoted to coding and review), i.e. 20-30 hours/week reinvested for a team of 10. The thesis: resolve Solow's (1987) "productivity paradox" by shifting from **usage metrics** (tokens) to **impact metrics** (throughput, time saved, failure rate, satisfaction). Recommendation: start with a **team** (not an individual) and measure 2-3 months later.

## Titre Article

The AI-native SDLC is paying off: 19% more PRs and 2–3 hours saved per developer per week

## Date

2026-05-31

## URL

https://www.atlassian.com/blog/ai-at-work/ai-native-sdlc-paying-off-per-developer-per-week

## Keywords

AI-native SDLC, Rovo Dev, coding agents, developer productivity, PR throughput, pull request throughput, hours saved, productivity paradox, Solow, quasi-experiment, propensity score, impact metrics, usage metrics, tokens, change failure rate, developer satisfaction, Teamwork Graph, context, team adoption, measurement, DevEx, DORA

## Authors

Robbie Geoghegan, Fan Jiang (Atlassian)

## Ton

Profile: corporate blog article (Inside Atlassian) co-authored by two data scientists, a vendor's "we" perspective measuring its own product, a data-driven and rigorous register (quasi-experimental methodology made explicit), moderate technical level, targeting engineering leaders, EMs, DevEx/platform leads and AI tooling purchase decision-makers. The tone combines **methodological authority** (sample size, propensity matching, conservative percentile for survey estimates) with **openly commercial argumentation**: the article demonstrates Rovo Dev's value while presenting itself as a generalizable measurement framework. The rhetoric relies on **calibrated, nuanced figures** (ranges, segmentation by repository activity and by number of adopters) rather than absolute promises, which reinforces its credibility. Intellectual framing via the Solow quote (1987, "you can see the computer age everywhere but in the productivity statistics") positions AI within the lineage of productivity-paradox debates, and resolves the paradox through a metrics shift (usage → impact). An implicit human-agent partnership metaphor runs through each stage of the cycle.

## Pense-betes

- **Central thesis**: the AI-native SDLC "pays off" — but only if measured by **impact** (PR throughput, time saved, quality, satisfaction), not by **usage** (tokens consumed).
- **Headline figure**: repositories adopting **Rovo Dev** merge **+19% more PRs per month** vs. non-adopters.
- **Activity-based amplification effect**: **+37-51%** throughput on low/medium-activity repositories.
- **Team effect (key)**: when **3 to 5 members** adopt the tool, the gain **doubles** to **+59-87%**. → collective adoption outperforms individual adoption.
- **Efficiency**: **2-3 h/week** saved per developer on coding and review, i.e. **~10%** of the ~24 h devoted to these tasks → **20-30 h/week** reinvested for a team of 10.
- **Usage context**: **93%** of developers use AI tools; **~30%** of code is written by AI; **96%** of surveyed Atlassian developers are Rovo Dev users.
- **5-stage AI-native SDLC** (agent + human): **Plan** (the agent proposes technical breakdowns and estimates for validation) → **Orchestrate** (coordinate humans and agents) → **Code** (autonomous agents on well-scoped work, PRs ready for review) → **Review** (the agent reviews against team standards before the human) → **Operate** (always-on incident copilots: alert triage, proposed fixes).
- **4-dimension measurement framework**: **Speed** = PR throughput; **Efficiency** = time saved; **Quality** = change failure rate; **Satisfaction** = developer satisfaction.
- **Productivity paradox (Solow, 1987)**: "you can see the computer age everywhere but in the productivity statistics" → AI massively adopted without clear measurement; resolution = shifting from **usage metrics** to **impact metrics**.
- **Role of context (Teamwork Graph)**: context-rich AI delivers results **+44% more accurate** while consuming **−48% fewer tokens** → context beats raw consumption.
- **Methodology**: 3,400 repositories / 2,500 customers; **quasi-experiment** with propensity-score matching; survey of **6,200+** Atlassian developers, estimates taken at the **20th percentile** (conservative); internal Atlassian teams = **1.35×** more Rovo Dev-assisted merged PRs than external customers.
- **Actionable recommendation**: "Start with a team, not an individual. Pick a repo with 3–5 engineers who will actively use the product. Measure throughput and time savings 2-3 months after implementation." → measure **after the novelty effect** wears off.
- **Related**: converges with DORA (delivery metrics), the FinOps token→outcome doctrine (Tokenomics Foundation, Gupta), *Failing Faster* (quality), and the idea that team/harness adoption > the individual tool (Eight Levels, Rafal).

## RésuméDe400mots

On its Inside Atlassian blog, Atlassian publishes a data study co-authored by two data scientists (Robbie Geoghegan, Fan Jiang) measuring the actual return of an **AI-native SDLC** powered by its **Rovo Dev** agent. The stakes are framed from the outset around the "productivity paradox" formulated by Robert Solow in 1987 ("you can see the computer age everywhere but in the productivity statistics"): AI is massively adopted — 93% of developers use AI tools, nearly 30% of code is written by AI — but its impact remains unclear as long as it is measured in **usage** (tokens) rather than **impact**.

The results, drawn from a quasi-experiment across 3,400 repositories from 2,500 customers (propensity-score matching), are quantified and segmented. Repositories adopting Rovo Dev merge **19% more pull requests per month** than non-adopters. The gain rises to **37-51%** on low- or medium-activity repositories, and **doubles to 59-87%** when **3 to 5 members** of the team adopt the tool: collective adoption clearly outperforms individual adoption. On the efficiency side, a survey of more than 6,200 developers (estimates taken at the 20th percentile, hence conservative) establishes a gain of **2-3 hours per week** on coding and review tasks, or about 10% of the 24 hours they involve — that is, 20-30 hours per week reinvested for a team of ten.

The article proposes a **five-stage AI-native SDLC** in which the agent supports the human: Plan (proposed breakdowns and estimates), Orchestrate (human/agent coordination), Code (autonomous agents on well-scoped work, PRs ready for review), Review (review against team standards before the human) and Operate (always-on incident copilots). It pairs this with a **four-dimension measurement framework**: Speed (PR throughput), Efficiency (time saved), Quality (change failure rate) and Satisfaction (developer satisfaction) — so as not to reduce value to velocity alone.

Two points reinforce the argument. First, the role of **context**: thanks to Atlassian's Teamwork Graph, context-rich AI delivers results that are 44% more accurate while consuming 48% fewer tokens. Second, the **operational recommendation**: start with a team (not an individual), choose a repository with 3-5 engineers who are actual users, and measure throughput and time savings 2-3 months after deployment, once the novelty effect has worn off. The underlying message: the value of AI is real but conditioned on rigorous impact measurement and team-level adoption.

## GrapheDeConnaissance

- Atlassian —publie→ The AI-native SDLC is paying off (DOCUMENT, 0.97)
- Robbie Geoghegan —travaille_chez→ Atlassian (ORGANISATION, 0.95)
- Fan Jiang —travaille_chez→ Atlassian (ORGANISATION, 0.95)
- Atlassian —mesure→ +19 % de PR mergées par mois pour les dépôts adoptant Rovo Dev (MESURE, 0.95)
- Atlassian —mesure→ 2-3 heures gagnées par développeur et par semaine (MESURE, 0.93)
- Rovo Dev —améliore→ débit de pull requests des équipes (CONCEPT, 0.92)
- adoption par 3 à 5 membres —améliore→ gain de débit de PR (+59-87 %) (MESURE, 0.9)
- SDLC AI-native —fait_partie_de→ partenariat humain-agent sur cinq étapes (CONCEPT, 0.9)
- SDLC AI-native —utilise→ Rovo Dev (TECHNOLOGIE, 0.88)
- Atlassian —recommande→ démarrer par une équipe (3-5 ingénieurs), pas un individu, et mesurer 2-3 mois après (AFFIRMATION, 0.92)
- Atlassian —recommande→ passer des métriques d'usage (tokens) aux métriques d'impact (AFFIRMATION, 0.9)
- article —référence→ paradoxe de la productivité de Solow (1987) (CITATION, 0.88)
- Teamwork Graph —améliore→ précision de l'IA (+44 %) avec −48 % de tokens (MESURE, 0.87)
- cadre de mesure à 4 dimensions —s_applique_à→ Speed, Efficiency, Quality, Satisfaction (CONCEPT, 0.86)
- Atlassian —surpasse→ clients externes (1,35× plus de PR assistées mergées) (MESURE, 0.85)

---
Canonical: https://www.thekb.eu/en/fiches/atlassian-ai-native-sdlc-paying-off-rovo-dev-2026-05-31/