Commission a study
Know exactly what AI can do
for your engineering team
We ran our methodology on 69,000 public Spring Framework tickets. The same analysis
on your own Jira history gives you numbers your stakeholders will actually believe.
69k
tickets already analysed
~1 week
typical turnaround
peer-reviewed
methodology, fully open
What you get
📊
AI readiness distribution
Every ticket in your Jira classified as Automate / Assist / Escalate using
outcome signals from your own resolution history — not generic benchmarks.
Broken down by project, team, ticket type, and trend over time.
🤖
Model benchmark on your tickets
We run the same four-model evaluation (GPT-4o, GPT-4o-mini, Claude Sonnet,
Claude Opus) on a sample of your resolved tickets with git commits.
You get pass rates, token overlap scores, and cost-per-ticket estimates
specific to your codebase and ticket writing style.
💰
Cost model
Based on your distribution and the model benchmark, we project the actual
cost of running AI on your ticket backlog — and the expected time savings.
Gives you a concrete ROI number for internal investment decisions.
📄
Report + stakeholder deck
A written report with all findings, methodology, and limitations — modelled
on our
published methods paper.
Plus a slide-ready summary your CTO or VP Engineering can present internally.
🌐
Interactive results page (optional)
A hosted or self-hosted version of the readiness chart and model comparison
page — same format as
our public results —
populated with your data. Share internally or publish externally.
How it works
1
Share access
Provide a read-only Jira API token and your project key(s), plus read access
to the corresponding git repository. We pull only resolved tickets and their
linked commits — no write access needed.
Your data never leaves your infrastructure if preferred
2
We run the pipeline
The same open-source scripts used for the Spring study — adapted to your
project key and git history. We compute outcome signals, classify every
ticket, build the git benchmark, and run the model evaluation.
Typical runtime: 1–2 days of compute.
3
Review draft findings
We share preliminary results and walk you through them in a 60-minute call.
You flag any surprises or context we're missing (e.g. certain ticket types
that should be excluded). We incorporate feedback.
4
Receive final deliverables
Final report, stakeholder deck, and (if included) the interactive results page.
All underlying data and scripts are yours — no lock-in.
Typical turnaround: 5–7 working days from access
🔒
Data privacy
We sign an NDA before any access is granted. The pipeline can run entirely
inside your infrastructure — we provide the scripts and you run them locally,
sending back only the anonymised results JSON. No ticket content, source code,
or credentials need to leave your systems. We're happy to discuss a data
processing agreement if required by your legal team.
Pricing
Starter
€1,900
one-time, up to 10k tickets
- AI readiness distribution
- Model benchmark (2 models)
- Written findings report
- One review call
- Raw data & scripts
Most popular
€3,900
one-time, up to 50k tickets
- AI readiness distribution
- Full 4-model benchmark
- Cost model & ROI projection
- Written report + stakeholder deck
- Interactive results page
- Two review calls
- Raw data & scripts
Larger archives or multiple projects? Get in touch for a custom quote.
Enterprise pricing available with on-premise delivery and extended support.
Common questions
Which Jira and git setups do you support?
Any Jira Cloud or Jira Server instance with REST API access. Git hosting on
GitHub, GitLab, Bitbucket, or self-hosted. The pipeline reads standard git
history — no special plugins needed.
Our ticket history isn't linked to git commits. Does that matter?
The tier classification (Automate / Assist / Escalate) works on any resolved
ticket history — no git link needed. The model benchmark requires some tickets
with matching commits, but even 20–30 matched tickets is enough for a meaningful
evaluation. We'll report exact coverage upfront.
How long does it take?
Typically 5–7 working days from when we have access. Pulling and indexing
50k tickets takes a few hours; the model evaluation runs overnight.
The review call and final report add 1–2 days.
Can we publish the results?
Yes — with your permission we're happy to add your study to our
research page (anonymised or attributed, your choice).
Publishing your results increases their credibility with candidates and the
wider engineering community.
What if the results are inconvenient?
We report what the data shows. If 80% of your tickets are Escalate, that's a
finding — it means your team is tackling genuinely hard problems, and AI tooling
should be scoped accordingly. We'll help you frame it constructively.
Get started