San Francisco

Positive NPS of +31 — promoters outweigh detractors with moderate passive presence.

Moderate end-of-Day-1 build confidence — cohort is progressing; Day 2 has room to close remaining gaps (mean 3.9/5).

90 participants across 18 organisations — Transformation Lead majority with applying in practice the most common AI experience level.

What participants said — dominant theme from each NPS segment.

Praxent (NPS +77, n=13) and Fujitsu (NPS -100, n=4) bracket the cohort. Praxent had zero detractors and 10 promoters, and four of five promoter verbatims name the same drivers: hands-on building and direct client relevance — 'the hands on experience with actual building with Claude is incredibly insightful,' and 'guided instruction on how to really leverage Claude in real world contexts... highlighted many areas where we can provide more value at less cost to our clients.' Notably, Praxent's respondents span three AI-experience levels and all three personas, so the result is thematic rather than demographic. Fujitsu, at the other end, returned four detractors from four respondents, but only two left verbatims — both Transformation Leads, and both pointing the same way: 'Most of my colleagues are already familiar with these topics' and 'Didn't learn much.' The complaint is prior knowledge, not facilitation. That direction matters: where C29's detractor signal was content pitched too technical for non-technical participants, C30's is the inverse — experienced attendees finding the material too familiar. Read together, the programme signal is audience-level calibration in mixed-skill rooms rather than a single direction of travel. Both segments are small (n=13, n=4); Fujitsu especially is anecdotal, not representative.

Promoter open text (n=28) clusters on four themes, by frequency of mention: breadth and new concepts (10), hands-on building and exercise design (8), client relevance and applicability (5), and instructor quality and structure (4). Representative responses: 'Well structured and great material. The hands on experience with actual building with Claude is incredibly insightful and a great learning tool.' 'Guided instruction on how to really leverage Claude in real world contexts is extremely valuable.' And from a participant who came in without a technical background: 'This session was fast paced but in the best way possible! Even coming from a non-technical background, I felt that I was able to grasp the fundamentals of Claude's LLM and how it can be applied.' Within the Transformation Lead persona, mean build-confidence gain declines as proficiency rises — a pattern present in 8 of 12 cohorts with two or more readable proficiency bands. In this cohort: learning and exploring +0.89 (n=9), applying in practice +0.50 (n=20), delivering independently +0.14 (n=7). NPS does not track confidence gain consistently across the dataset; in 9 of 12 cohorts NPS also declines with proficiency, whereas in this cohort it rose (+11, +20, +71 across the same bands). This cohort's satisfaction gradient is therefore atypical and should not be generalised. Depth and pace responses distribute across different proficiency bands (n=90 Day-1 respondents). Depth rated "too basic" — 13 responses total: Delivering independently 6 of 13, Operating at the frontier 5 of 11, Applying in practice 2 of 52, Learning and exploring 0 of 14. Pace rated "too fast" — 15 responses total: Applying in practice 11 of 52, Learning and exploring 3 of 14, Delivering independently 1 of 13, Operating at the frontier 0 of 11. The two signals do not co-occur within bands. The two highest proficiency bands account for 11 of 13 "too basic" responses and 1 of 15 "too fast" responses. The "too fast" signal is concentrated in the "applying in practice" band, which returned 2 of 52 "too basic". Open-text responses within the Transformation Lead persona separate along the same axis. At practitioner level, responses describe an inability to reproduce the work independently rather than insufficient depth: "Understood workshop content but since so much scaffolding was already in Jupyter notebook I'm not sure I could independently do it end to end"; "should have been extended into a 4 day session... I thought the pace of hands on activities was too fast." One respondent identified language as a barrier during the hackathon. At director level and above, responses cite prior knowledge: "Didn't learn much"; "I feel this is for beginners level... that knowledge does not need 2 days of training. Was expecting more on how to build more multi agent solutions from business standpoint." These distributions do not support a uniform pace reduction: the depth and pace signals originate in separate segments and imply separate interventions. Band sizes range from 11 to 52; the depth signal rests on the two smallest bands (n=11, n=13). Seniority-level open-text segments (practitioner n=9, director+ n=13) fall below the 20-response threshold applied to two-way cuts and are reported qualitatively only, without segment-level scores.