01
Kittl Agentic AI
Current hands-on production experience: owned the Agentic AI project at Kittl, from architecture and technical direction through to production delivery.
AI ENGINEERING WORKFLOW SPRINT
Make AI-assisted speed survive the whole delivery workflow.
Coding agents can make implementation faster while the overall system remains just as expensive.
We take one real production workflow, baseline where the gain is being lost, implement the smallest useful engineering intervention, and test it on live changes.
PRODUCTION AI ENGINEERING
Production AI engineering — not prompt demos.
02
AI-assisted technical sign-off
Designed an AI-assisted engineering sign-off workflow adopted across approximately five product teams and used on 23 projects.
03
AI Change Verification — open source
We built and open-sourced an evidence-accountable verification workflow for AI-assisted code changes before human review — with deterministic evidence, unresolved-risk handling and explicit human judgment.
AI Change Verification on GitHubWHEN FASTER IMPLEMENTATION DOES NOT BECOME FASTER DELIVERY
The bottleneck often moves instead of disappearing.
01
Implementation gets cheaper, but delivery is not more predictable
Tasks move through coding faster while unclear intent, requirements, architecture decisions or handoffs still create delay downstream.
02
Individual AI usage does not become a team system by itself
Each engineer may have useful tools and personal workflows, while the team still lacks shared context, checks, conventions and measurement.
03
Senior judgment arrives late and expensively
Architecture, risk, edge cases and unresolved assumptions surface during review or rework instead of earlier in the workflow.
If AI-assisted development is not yet materially changing how the team implements software, this is probably not the right Sprint.
THE SPRINT
Fix the next delivery bottleneck — not “AI adoption” in general.
We select one production workflow, establish the current baseline, identify where AI-assisted speed is being lost, and implement the smallest useful intervention inside the existing engineering environment.
The mechanism follows the bottleneck. We do not force every team into the same AI process.
The goal is not maximum automation. The goal is a delivery workflow where AI does more useful work and human judgment is spent where it creates the most value.
- 01
Task and decision context
When implementation starts before expected outcome, constraints or unresolved decisions are sufficiently clear.
- 02
Shared AI / agent instructions
When individual workflows work well but team-level consistency is weak.
- 03
Architecture and approval checkpoints
When agents move quickly but system-level decisions are repeatedly revisited later.
- 04
Deterministic verification
When tests, linting, type checks, static analysis, reproduction steps or structural rules should happen before expensive human review.
- 05
Review readiness
When changes arrive technically plausible but reviewers still reconstruct intent, evidence and unresolved risks manually.
- 06
Human decision boundaries
When it is unclear which decisions may be delegated and which require explicit human ownership.
- 07
Before/after workflow measurement
Using one to three practical measures appropriate to the selected workflow.
Ten business days from baseline to live evidence.
- 01
Scope one production workflow
Define the workflow, team, primary repo, current AI usage and the workflow outcome the Sprint should improve.
- 02
Capture the baseline
Measure the current friction: rework, reviewer attention, repeated context reconstruction, handoffs, verification cost or another practical workflow signal.
- 03
Identify the real bottleneck
Distinguish the highest-value constraint from symptoms and select the smallest intervention worth testing.
- 04
Implement the intervention
Change the real workflow rather than producing a recommendation deck.
- 05
Test on live changes
Run the updated workflow on real engineering work and compare the observed evidence with the baseline.
- 06
Handoff and decision
Deliver the working intervention, result memo, unresolved items and a clear recommendation: roll out, iterate or stop.
FIT
For teams where AI is already changing implementation, but not the whole delivery system.
The Sprint is intentionally defined by workflow maturity and bottleneck severity, not company size.
GOOD FIT
- AI-assisted implementation is already materially used.
- There is one production workflow with visible friction.
- Engineering leadership can name the human cost of the current workflow.
- The team can test the intervention on real changes.
- A bounded workflow improvement is more useful than a broad transformation programme.
NOT A FIT
- The goal is generic AI training or tool rollout.
- There is no concrete workflow to test.
- The buyer wants an organization-wide transformation in ten days.
- The problem is entirely outside software delivery.
WORKING DIRECTLY WITH ENGINEERS
The engineers doing the work
AI & Software Engineering Lead
Yaroslava Suspitsina
Senior software engineer focused on production AI systems, agentic architecture and engineering workflows. She currently works at Kittl, where she owned the Agentic AI project from architecture and technical direction through to production delivery. She has also designed AI-assisted review, testing and quality workflows used across multiple product teams.
- Current Kittl Agentic AI project ownership and production delivery.
- AI-assisted technical sign-off used across approximately five product teams and 23 projects.
- Agentic architecture, testing, CI quality automation, Playwright, debugging and engineering enablement.
AI/ML Engineer
Valerii Safonov
AI/ML engineer with experience across machine learning, quantitative modelling, NLP/RAG, computer vision and production-oriented AI systems. MSc in Computer Science with a machine-learning focus, with applied AI engineering experience dating back to at least 2020.
- MSc in Computer Science with a machine-learning focus.
- Applied AI/ML work across quantitative modelling, NLP/RAG, computer vision and production engineering.
- Earlier applied technology-trend detection and evaluation work spanning data, architecture and engineering methods.
No handoff from sales to a junior delivery team. The people shown here work on the engagement.
FIXED-SCOPE SPRINT
One team. One production workflow. Ten business days.
€6,500 fixed
+ VAT where applicable
10 business days · 1 team · 1 production workflow · 1 primary repo · baseline + implementation + live trial + before/after evidence + handoff.
We scope the workflow before work starts. If the problem is too broad, too small or cannot be tested honestly within the boundary, we should not force it into the Sprint.
No ongoing retainer required. A broader rollout or another workflow is a separate decision after the Sprint.
Questions before the Sprint
Is this an AI adoption or training programme?
01No. It assumes AI-assisted development is already materially present and focuses on one real software-delivery workflow.
What kinds of bottlenecks can the Sprint address?
02Common examples include missing task context, architecture iteration, inconsistent agent practices, rework, verification cost, review readiness, handoffs and unclear human decision boundaries. The actual intervention follows the selected workflow.
Do we have to use a specific coding agent?
03No. The Sprint is tool-neutral and works around the engineering environment already in use.
Do we have to migrate platforms?
04No. A platform change is not a goal. Existing tools should remain unless a change is specifically justified by the workflow.
Is this only for a particular team size?
05No. Fit is determined by the workflow and its bottleneck rather than headcount.
Do you guarantee a percentage productivity gain?
06No fixed percentage is promised. The purpose of the bounded Sprint is to implement and evaluate one real intervention with observable before/after evidence.
What if our problem is only review and verification?
07The narrower AI Engineering Review & Verification Sprint is likely the better fit.
What if we are only a few people and our main problem is repeated agent supervision?
08The AI-Native Delivery System Sprint is designed specifically for that situation.
What happens after ten business days?
09The Sprint ends with handoff and a result decision. There is no automatic ongoing consulting commitment.
Contact
Show us where AI-assisted speed gets lost.
Briefly describe your team, the AI/coding tools already in use, and one production workflow where implementation has become faster but delivery still feels expensive or unpredictable.