Q: Why do Delivering independently AI proficiency participants rate the programme +7 pts above average — what does the D1 calibration data tell us?

The D1 calibration data does not include NPS broken down by AI proficiency band, so the +7 pt advantage cannot be directly explained by calibration scores alone.

The data brief confirms Delivering independently participants rated the programme above average NPS across the 3 cohorts analysed, but NPS by AI proficiency band is not reported — only headcounts per band are available (C19: n=10, C20: n=9, C21: n=13, C22: n=5). D1 calibration scores are reported at cohort level only (range 3.76–3.96), not segmented by proficiency band, so we cannot determine whether Delivering independently participants entered with higher or lower baseline confidence than peers. What the verbatim data does suggest is that the 'too basic' complaint appears almost exclusively from Operating at the frontier participants (e.g. McKinsey Developer nps=5, Fractional AI Developer nps=6, Bain Architect nps=6), not from Delivering independently participants — who contribute promoter verbatims citing applicable techniques and hands-on work. This implies Delivering independently may sit in a content sweet-spot: advanced enough to be novel, practical enough to be immediately usable. However, this is pattern inference from sparse verbatims, not a statistically supported finding. Key data points: · Delivering independently headcounts are small across cohorts: C19 n=10, C20 n=9, C21 n=13, C22 n=5 — all below the n=20 threshold for confident segment-level conclusions · D1 calibration scores are cohort-level averages only (3.76–3.96), not segmented by AI proficiency band — no Delivering independently baseline confidence delta is calculable · Detractor verbatims citing content being 'too basic' or 'already known' come from Operating at the frontier participants, not Delivering independently — suggesting a content-fit differential between the two bands Caveats: To answer this question with confidence, NPS by AI proficiency band and D1 calibration scores segmented by proficiency band would be needed. The cross-cohort correlations section explicitly flags AI proficiency correlation as 'insufficient data', confirming this gap.

2026-07-28 18:24 UTC

Q: What explains Deloitte's persistent above-programme NPS score across 2 cohorts, and what would change it?

Deloitte's above-programme NPS (+83 in C19, +44 in C20 vs. programme averages of +55 and +35) cannot be fully explained by the available data — sample sizes of n=6 and n=9 are too small for confident causal attribution.

Deloitte scored +83 in C19 (n=6) and +44 in C20 (n=9), both above their respective cohort NPS of +55 and +35 — a consistent pattern across the two cohorts where Deloitte appears in the org-level breakdowns. However, with combined Deloitte n=15 across both cohorts, no statistically confident explanation is possible. The verbatim evidence offers limited signal: C19 Deloitte promoters cite 'applicable techniques' and 'great food'; C20 Deloitte promoters cite 'real life scenarios' and hands-on Claude Code time, while one Deloitte detractor in C20 explicitly wanted more advanced content ('ontology and use of ontology with Claude to create agents'). This detractor pattern — a technically advanced participant finding content too basic — is the same risk that could erode Deloitte's NPS: if cohorts skew toward higher AI proficiency Deloitte attendees, the 'too basic' detractor signal becomes more likely to repeat. What would change it: sending more frontier-level Deloitte participants without an advanced track to match their expectations would convert promoters to passives or detractors. Key data points: · Deloitte C19 NPS: +83, n=6 (cohort average: +55) · Deloitte C20 NPS: +44, n=9 (cohort average: +35) · Combined Deloitte n=15 across both cohorts — below the 20-respondent threshold for confident segment conclusions · One Deloitte detractor in C20 (score: 6) cited content being too basic for advanced use cases · Programme-wide detractor theme across orgs: 'too basic for frontier-level practitioners' appears in C19, C21, and C22 verbatims Caveats: With n=6 and n=9, Deloitte's above-programme scores could reflect cohort composition luck (e.g., more 'Applying in practice' than 'Operating at the frontier' attendees) rather than a structural Deloitte effect. To answer this confidently, you would need Deloitte org-level AI proficiency breakdowns and at least 20+ Deloitte respondents across cohorts.

2026-07-28 18:24 UTC

Q: What is driving the high passive rate in C20 San Francisco July 14–15, 2026 — what do the verbatims say and what would convert them to promoters?

C20's 43.8% passive rate is driven by Transformation Leads (NPS +17, n=18) citing relevance gaps and content depth mismatches — not delivery quality.

Three C20 verbatims point to two distinct passive/detractor drivers: (1) relevance mismatch — a PwC Transformation Lead rated 6, saying 'not as directly relevant for my work in finance consulting,' and another PwC Transformation Lead rated 6 noting the content was 'very technical but a great way to learn what back end developers do' — suggesting non-technical personas felt the programme wasn't pitched at them; (2) depth ceiling — a Deloitte Architect rated 6 calling it 'very basic' and wishing for ontology/agent-building content. Passive verbatims (scores 7–8) use positive language — 'hands on and engaging,' 'very insightful,' 'well organized' — but the low scores suggest appreciation without genuine enthusiasm, consistent with a relevance gap rather than a delivery failure. The Transformation Lead persona drove the most risk: NPS +17 vs. Developer NPS +70 and Architect NPS +43, and PwC specifically scored NPS -11 (n=9), the only negative org score in C20. Confidence deltas were the lowest of all four cohorts (D1=3.76), indicating participants also entered with slightly lower baseline confidence, which may have amplified dissatisfaction when content skewed technical. Key data points: · Transformation Lead NPS +17 (n=18) vs. Developer +70 (n=10) — largest persona gap in C20 · PwC NPS -11 (n=9) — only negative org score in C20; two PwC detractor verbatims cite technical irrelevance · Passive verbatims (scores 7–8) uniformly positive on logistics/delivery but silent on business applicability — relevance gap, not execution failure · C20 had the lowest D1 confidence score across cohorts (3.76), suggesting a less experienced intake that may have found technical depth harder to absorb · Detractor verbatims split: 2 cite 'too basic/advanced depth missing' (Architect), 1 cites 'not relevant to finance consulting' (Transformation Lead) — two different fix paths needed Caveats: Only 3 passive and 3 detractor verbatims are available for C20 (n=48 total), so themes are directional, not statistically confirmed. A post-event pulse survey targeting the 21 passives specifically — asking about content relevance, depth, and role applicability — would give actionable segmentation data to confirm whether to split tracks or add persona-specific modules.

2026-07-28 18:25 UTC

Q: What is driving the high passive rate in C21 San Francisco July 16–17, 2026 — what do the verbatims say and what would convert them to promoters?

C21's 47% passive rate is driven by a content-depth mismatch: advanced participants found the material too basic, while non-technical roles found Jupyter notebooks irrelevant to their work.

Three C21 verbatims point to the same root cause from different angles. Fractional AI / Ode detractors (NPS -30, n=10) explicitly state 'it's a bit too basic for us' and 'pretty basic for most of my coworkers' — both from frontier-level practitioners (Operating at the frontier). A PwC passive (NPS 7) flags the opposite problem: 'very very heavy on Jupyter notebooks which is not relevant to my current role' — a Transformation Lead who is Delivering independently. A separate PwC detractor scores 5 and states 'not differentiated by level of technical skill', directly naming the structural issue. The Developer persona is the most distressed segment at NPS -6 (n=16), yet C21 has the highest engineering function share of any cohort (24 of 49). The concentration of Fractional AI / Ode attendees — a single org driving NPS -30 — is a compounding factor, but the depth complaint spans orgs. Converting passives to promoters requires either streaming by AI proficiency level or offering modular depth tracks, so frontier practitioners get advanced content while non-technical leads get role-relevant application framing. Key data points: · C21 passive rate: 46.9% (n=23 passives out of 49 respondents) · Fractional AI / Ode NPS: -30 (n=10) — single org heavily skewing detractor count; all complaints centre on content being too basic · Developer persona NPS: -6 (n=16) — worst segment in cohort; 24 of 49 respondents are in Engineering function · PwC passive verbatim: 'very very heavy on Jupyter notebooks which is not relevant to my current role' (Transformation Lead, Delivering independently) · PwC detractor verbatim (NPS 5): 'not differentiated by level of technical skill' — directly names the structural gap Caveats: Verbatim coverage is sparse — only 9 C21 quotes are available across all NPS bands from n=49 respondents, so themes are directionally consistent but not statistically validated. Knowing the AI proficiency breakdown of passives specifically (vs. the full cohort) would sharpen whether depth or relevance is the primary conversion lever.

2026-07-28 18:25 UTC

Q: What is driving the high passive rate in C22 London July 20–21, 2026 — what do the verbatims say and what would convert them to promoters?

C22 London's 53% passive rate is driven by a pace/depth mismatch: too little time for exercises and content pitched too basic for frontier-level attendees.

Three passive verbatims from C22 London reveal two distinct friction points. First, time pressure: one Capgemini Architect (Learning & Exploring) scored 7 citing 'very technical content but little time to really run the exercises,' and an unmatched respondent also scored 7 requesting 'more presentations to present concepts and solutions rather than experiencing with it' — suggesting some attendees felt rushed through hands-on work without sufficient conceptual scaffolding. Second, depth insufficiency: the sole detractor, a Bain Architect operating at the frontier, scored 6 and explicitly requested an Anthropic engineer to cover advanced topics, calling content 'too basic.' The one passive who wrote only 'content' (IndiciumAI Developer, frontier proficiency) provides no actionable signal. Promoter verbatims praise interactivity and hands-on work, indicating the format itself is valued — the gap is calibration of pace and depth, not format. C22 also has a significant day-2 drop (55 Day 1 vs. 32 Day 2, only 24 matched), which means the verbatim base is thin and findings should be treated as directional only. Key data points: · C22 passive rate: 53.1% (n=32 respondents, but only 24 matched Day 1–Day 2) · 2 of 3 passive verbatims cite time/pace issues: 'little time to really run the exercises' and desire for more concept presentation before hands-on · 1 detractor (frontier-level Architect, Bain, score 6) calls content 'too basic' and requests Anthropic engineer for advanced topics · Frontier and 'Delivering independently' proficiency groups likely over-represented among passives: the IndiciumAI passive and Bain detractor are both frontier-level · Day 1 attendance was 55 but only 32 completed Day 2 survey — a 42% drop that inflates uncertainty in all C22 findings Caveats: C22 verbatim sample is very small (3 passive quotes, 1 detractor quote) — n < 20 for matched respondents (n=24), so no statistically confident conclusions can be drawn. To sharpen this analysis, mid-programme pulse data, full open-text responses from all 32 Day 2 respondents, and a breakdown of AI proficiency distribution among passives specifically would all help isolate whether depth or pace is the primary conversion lever.

2026-07-28 18:25 UTC

Q: What do promoters and passives say about the programme, and what specifically separates them?

Promoters cite learning, real-world relevance, and hands-on engagement; passives acknowledge quality but flag format mismatches — specifically too many Jupyter notebooks, insufficient concept framing, and content pitched below their level.

Across C19–C22, promoter verbatims consistently reference substantive learning ('learned a lot of applicable techniques', 'good course to get hands on time with Claude Code', 'very valid real life scenarios') and engagement quality ('great presenters, very interactive', 'explanatory and with a good level of hands on'). Passive verbatims are notably thinner in specificity — phrases like 'super fun and technical', 'very insightful content', 'well organized', and 'hands-on and engaging' register satisfaction but not transformation, suggesting the programme was good but not meaningfully differentiated in their experience. The clearest separating signal in passive verbatims is format and pacing friction: C21 saw a passive cite 'very heavy on Jupyter notebooks which is not relevant to my current role', and C22 passives flagged 'little time to really run the exercises' and a preference for 'more presentations to present concepts and solutions rather than experiencing with it'. This points to a structural gap — passives want better scaffolding before doing, not just doing. Promoters either had sufficient context to engage fully or found the hands-on format intrinsically rewarding; passives did not clear that threshold. Key data points: · 40.3% of respondents (n≈77 across C19–C22) are passives — highest in C22 London at 53.1% (n=32) · Passive verbatims across all cohorts use generic positive descriptors ('well organized', 'very insightful', 'hands-on and engaging') with no reference to specific skills, tools, or applicable takeaways · 3 of 12 passive verbatims sampled explicitly name format issues: Jupyter notebook overload (C21), insufficient time for exercises (C22), and desire for more concept framing before hands-on work (C22) · Promoter verbatims reference specific content or outcomes in 8 of 12 sampled cases: Evals, Claude Code, Agentix programming, real-life scenarios, and applicable techniques · Passives skew toward 'Applying in practice' AI proficiency in the verbatim sample — not beginners, not frontier — suggesting mid-level practitioners are least served by the current format Caveats: Verbatim sample is small (3 per segment per cohort) and not guaranteed representative — treat themes as directional hypotheses, not statistically validated conclusions. Structured follow-up with passives on format preference and content depth expectations would sharpen the conversion lever significantly.

2026-07-28 18:26 UTC