Level 1 of 5
Conversational prompting
Your team lives in two windows. Editor on the left, chatbot on the right. Copy, paste, fix the indent, repeat. The AI has never seen your repo, so every conversation starts from zero.
What to do with this
Bring AI into the editor and kill the alt-tab tax. Then give it repo context. The distance to agentic engineering is real, but every level is climbable.
Level 2 of 5
Inline suggestions
Autocomplete is muscle memory now. Tab, tab, tab. Your team is faster than last year, but the AI is finishing sentences, not taking tasks. This is exactly where 2025 industry data says most teams are stuck, and why the gains stay near 10%.
What to do with this
The jump to repo-aware and then agentic work is the single biggest leverage move available to your team. It is a workflow change, not a tooling purchase.
Level 3 of 5
Contextual pair programmer
The AI sees your whole repo and your developers trust it for stretches. Real leverage. But your developers are still the harness: prompt, review, run tests, repeat, every loop, all day.
What to do with this
Hand off whole tasks. Agent CLIs that run tests and open PRs move your team from driving every loop to reviewing outcomes. That is the Level 4 jump.
Level 4 of 5
Agentic engineering
Your team hands off scoped tasks and reviews diffs instead of typing them. You are ahead of the vast majority. The question now is consistency: is this two enthusiasts, or the team's default way of working, with numbers to prove it?
What to do with this
Make it the default and make it measurable. Champions, shared standards, and a metrics baseline turn individual speed into organizational throughput.
Level 5 of 5
Toward the Dark Factory
Specs in, reviewed software out. Your team orchestrates agents instead of writing code. You are operating where most organizations will be in years. The bottleneck now is review bandwidth and pulling the rest of the organization up.
What to do with this
Scale it beyond the pioneers. The playbook that got one team here is an asset. Rolling it out across teams is an organizational program, not a tooling one.
Most teams we assess land at Level 2 or 3. The measured gains start at Level 4. That distance is exactly what the Golden Team Program closes in 90 days.