Meeting Report

Structured team training planned. Certification program activated. Valeria onboarded to Claude Code. Valeria learned Claude Code "yesterday" — first hands-on session. Team training session planned with Max Votek + Max Shevchenko for Mohammad, Kevin, and Valeria simultaneously. Anthropic certification program available; Arkhip Koloskin completed in 4 hours. All team members expected to certify. Andrey and Denis joining expands company-wide visibility.

Max Shevchenko published learnings to GitHub. Communication gap driving knowledge capture urgency. Max Votek noted Max Shevchenko "shared a pretty good result with the summary of learnings... several approaches which can be reused for the next customizing sessions." Need for project-specific GitHub with progress/issues/knowledge identified. Sessions to be recorded for quick-start knowledge. Training the AI on how to build (not just what) is itself a knowledge capture exercise.

Config workflow produces results but needs "how to build" training. QA and tech docs identified as next workflows. Build works but output is unreadable — needs teaching on proper worksheet population. Pre-sales estimator workflow actively being built (Kevin). Technical documentation workflow identified as critical client need. QA agent concept on roadmap. Video generation workflow explored. The gap is clear: AI can build, but must be trained to build in a human-readable, auditable way.

Valeria reviewed the build and found concrete gaps. UAT is actively happening, surfacing real issues. Worksheets empty (black box build). Fields not in expected objects. Calculations "seem accurate" but need full quarter comparison. Joint review session planned with Max Shevchenko + Valeria + Kevin. The fact that they're finding specific, actionable gaps means UAT is working — even though the build needs refinement. "Nothing is a showstopper."

First quantified staffing model: 3 roles → 2, developer eliminated, 73% hours reduction. Traditional: Solution Lead (76h) + Functional Consultant (144h) + PM (26h) = 246h. AI model: Architect (54h) + Developer (0h) + PM (11h) = 65h. Functional consultant role absorbed entirely. Developer role eliminated. This is the first concrete, numbers-backed proof of role compression. The slide confirms: "A small, skilled team can now deliver at a scale and pace that previously required a much larger organization."

Biggest mover: from vision-only to a concrete pricing model with margins quantified. Project price stays at ~$43K but internal cost drops from 246h to 65h at $175/hr = ~$11K internal cost vs ~$43K revenue. "Ideally we don't bill per hour... if it's going to be a fixed project... the margins will be a lot higher." Quick start vs custom decision tree discussed. Front-loading solution design (3 weeks) to eliminate mid-project surprises. Presenting to Salesforce on Monday.

No new PM tooling or process. Estimator is pre-sales, not PM. The staffing model comparison is useful for planning but isn't a PM system. Still cannot answer: profitability by project, planned vs actual hours, or forecasting accuracy. Taras structured next steps well but no tracking system discussed.

First concrete agent concepts: QA agent and pre-sales → delivery agent handoff. Mohammad discussed a QA agent to validate calculations. Also described pre-sales agent feeding into delivery agent via intake guide. Still conceptual — no agent built or tested. But these are the first specific agent roles articulated in a meeting.

Three packaged accelerators taking shape. Estimator near completion. Presenting to Salesforce Monday. (1) ROM Estimator: Kevin demoed working version with individual plan tabs, complexity matrix, integration section. Complete by end of week. (2) Health Check: AI assessment of existing SPIFF instances. Two AEs already requesting. Three-phase roadmap: train AI → assess → generate demand. (3) Automated Videos: Training/demo videos via Claude Code with personal avatars. Website concept with industry-specific success stories and AE booking flow. Professional slides already packaged.