← Home
Partner Basecamp · Cohort 25

San Francisco

July 29–30, 2026
+44
Net Promoter Score
58
Day 1 Responses
91%
Matched Pairs
93%
Day 2 Response Rate
Programme Satisfaction
Positive NPS of +44 — promoters outweigh detractors with moderate passive presence.
+44NPS · n=54
50% Promoters44% Passives6% Detractors
95% CI: +28 → +60  ·  True NPS lies within this range with 95% confidence (n=54 respondents)
Confidence Arc
Moderate end-of-Day-1 build confidence — cohort is progressing; Day 2 has room to close remaining gaps (mean 3.9/5).
3.9end-of-D1 build / 5
4.0D2 Design / 5
4.2D2 Commercial / 5
4.3D2 Build / 5
Audience
58 participants across 15 organisations — Transformation Lead majority with applying in practice the most common AI experience level.
Top Organisations
PwC (13) DXC Technology (11) Deloitte (9) Ascendion (7) Bain (4)
Voice of Participant
What participants said — dominant theme from each NPS segment.
Promoters
Hands-on breadth and real-world framing resonated strongly — promoters found the content engaging, detailed, and practical, leaving with confident, sharper ways to apply Claude in their work.
Passives
Live coding walkthroughs and real-time demonstrations would deepen engagement beyond exercises and pre-recorded content.
Detractors
Setup failures and access issues prevented exercise completion; pre-training and laptop preparation are critical for non-technical attendees.
Segment Finding
PwC's NPS 0 (n=9) was driven almost entirely by corporate laptop / IT access blocking the hands-on exercises — all three detractors cite it directly, not the content. The same environment/setup friction recurs across the 44% passive segment. Highest-leverage fix: a pre-event laptop and access check. Secondary signal: the most experienced 'Applying in practice' attendees wanted more technical depth.
Survey & Data Quality
Survey & Data Quality
Day 1 responses58
Day 2 responses54 (93% of Day 1)
Matched pairs49 (91% of Day 2)
Orgs resolved58 of 58 respondents matched to named org
Day 2-only respondents5 (9.3%) — suppresses NPS by 3 points; no Day 1 data, excluded from persona & confidence analysis
Organisations
All Day 1 respondents · n=58
PwC13 (22%)
DXC Technology11 (19%)
Deloitte9 (16%)
Ascendion7 (12%)
Bain4 (7%)
Cognizant4 (7%)
Accenture2 (3%)
AlixPartners1 (2%)
IBM1 (2%)
KPMG1 (2%)
EPAM1 (2%)
AWS1 (2%)
Forgd.AI1 (2%)
Ascendion, Inc1 (2%)
NEC(Japan)1 (2%)
Function × Seniority
All Day 1 respondents · n=58
FunctionPractitionerSr PractitionerMgr / Sr MgrDirector+Partner / MDTotal
Engineering5432–14
Architecture–446115
Business Leadership––47314
Project / Engmt2265–15
Experience Profile
AI experience level · n=58 · % of row
By Persona
SegmentExploringApplyingDeliveringFrontiern
Transformation Lead31%48%7%14%29
Architect27%40%20%13%15
Developer14%64%14%7%14
By Seniority
SegmentExploringApplyingDeliveringFrontiern
Director, Senior Director, or Principal25%60%5%10%20
Manager or Senior Manager35%41%18%6%17
Senior practitioner (5–9 years)20%30%30%20%10
Practitioner (0-4 years in role)14%71%—14%7
Partner, Managing Director, or Executive25%50%—25%4
How would you describe your current experience with AI tools prior to Basecamp?
Learning and exploring
15 (26%)
Applying in practice
29 (50%)
Delivering independently
7 (12%)
Operating at the frontier
7 (12%)
Prior to today, how much had you worked with Claude or the Anthropic API?
Not at all
6 (10%)
A little
25 (43%)
Regularly
27 (47%)
Did the depth land for this audience?
Technical depth perception · n=58 respondents · % of row
By Persona
SegmentToo basicAbout rightToo advancedn
Transformation Lead—66%34%29
Architect27%73%—15
Developer21%64%14%14
By AI Experience Level
SegmentToo basicAbout rightToo advancedn
Applying in practice7%76%17%29
Learning and exploring—53%47%15
Delivering independently57%43%—7
Operating at the frontier14%86%—7
Overall depth distribution
Too basic
7 (12%)
About right
39 (67%)
Too advanced
12 (21%)
Did the pace work across the room?
Session pace perception · n=58 respondents · % of row
By Persona
SegmentToo slowWell pacedToo fastn
Transformation Lead—93%7%29
Architect—100%—15
Developer—93%7%14
By AI Experience Level
SegmentToo slowWell pacedToo fastn
Applying in practice—97%3%29
Learning and exploring—87%13%15
Delivering independently—100%—7
Operating at the frontier—100%—7
Overall pace distribution
Well paced
55 (95%)
Too fast — not enough time to apply
3 (5%)
How confident were participants to build with Claude after Day 1?
End-of-Day-1 confidence · "How confident are you in your ability to build a client solution using Claude?" · n=58 · 1–5
Mean end-of-D1 build confidence by Persona
Architect4.13/5 · n=15
Developer4.00/5 · n=14
Transformation Lead3.38/5 · n=29
Programme mean: 3.9/5 · 18 cohorts
Mean end-of-D1 build confidence by AI Experience Level
Delivering independently4.43/5 · n=7
Operating at the frontier4.14/5 · n=7
Applying in practice3.76/5 · n=29
Learning and exploring3.13/5 · n=15
Overall end-of-D1 build confidence distribution
1
4 (7%)
2
4 (7%)
3
16 (28%)
4
14 (24%)
5
20 (34%)
How relevant was today's content to your current role?
Content relevance rating · n=58 respondents · 1 = not relevant, 5 = highly relevant
Mean relevance by Persona
Architect4.47/5 · n=15
Developer4.36/5 · n=14
Transformation Lead3.79/5 · n=29
Mean relevance by AI Experience Level
Delivering independently4.43/5 · n=7
Operating at the frontier4.43/5 · n=7
Applying in practice4.10/5 · n=29
Learning and exploring3.80/5 · n=15
Overall relevance distribution
1
1 (2%)
2
3 (5%)
3
9 (16%)
4
21 (36%)
5
24 (41%)
How likely are you to recommend attending this programme to a colleague?
n=54 Day 2 respondents · 93% of Day 1
Score distribution · 0 = not at all likely · 10 = extremely likely
Cohort012345678910
Cohort 2511111131116
■ Detractors 0–6■ Passives 7–8■ Promoters 9–10
NPS Breakdown
50%
44%
6%
NPS +44 (n=54)
Promoters (9–10)Passives (7–8)Detractors (0–6)
Programme mean: +42 · 18 cohorts
Promoters mean: 52%
Passives mean: 39%
Detractors mean: 9%
NPS by Segment
Organisations, personas, and experience levels · ≥4 respondents shown
By Organisation
Promoters (9–10)Passives (7–8)Detractors (0–6)
Ascendion
75%
25%
0%
NPS +75 (n=4)
Cognizant
75%
25%
0%
NPS +75 (n=4)
DXC Technology
70%
30%
0%
NPS +70 (n=10)
Deloitte
33%
67%
0%
NPS +33 (n=9)
Bain
25%
75%
0%
NPS +25 (n=4)
PwC
22%
56%
22%
NPS 0 (n=9)
By Persona
Promoters (9–10)Passives (7–8)Detractors (0–6)
Architect
54%
46%
0%
NPS +54 (n=13)
Developer
50%
50%
0%
NPS +50 (n=14)
Transformation Lead
50%
41%
9%
NPS +41 (n=22)
Programme means · 18 cohorts: Architect +51 · Developer +48 · Transformation Lead +36
By AI Experience Level
Promoters (9–10)Passives (7–8)Detractors (0–6)
Delivering independently
83%
17%
0%
NPS +83 (n=6)
Operating at the frontier
50%
50%
0%
NPS +50 (n=6)
Learning and exploring
50%
43%
7%
NPS +43 (n=14)
Applying in practice
44%
52%
4%
NPS +39 (n=23)
Programme means · 18 cohorts: Learning and exploring +45 · Applying in practice +44 · Delivering independently +41 · Operating at the frontier +40
What is the main reason for your score?
n=47 responses · organised by NPS segment
Promoters (score 9–10)· 25 responses
Hands-on breadth and real-world framing resonated strongly — promoters found the content engaging, detailed, and practical, leaving with confident, sharper ways to apply Claude in their work.
10“As someone with a more strategy/product owner background with very little (zero) developer or engineer experience, it was a great experience to understand more how the tools we use and implement work and to walk away with a better understanding of how the engineers I work with think through solving problems.”
Passives (score 7–8)· 19 responses
Live coding walkthroughs and real-time demonstrations would deepen engagement beyond exercises and pre-recorded content.
8“I would recommend it to technical people that are not actively working in AI. For me who does it was a bit more of a reminder and a way to piece things together. To someone else it could be the greatest program. Others who weren’t technical I see them struggling.”
Detractors (score 0–6)· 3 responses
Setup failures and access issues prevented exercise completion; pre-training and laptop preparation are critical for non-technical attendees.
5“On the PwC end, we need to be able to work out access before we come. I was never really able to run the modules. At some point, we have to ask: what is the point when we’re not able to run anything? I would highly recommend a pre-training module to ensure that someone can access before they come.”
End-of-Programme Confidence
n=54 respondents · 1 = not confident, 5 = very confident
How confident are you in your ability to design an evaluation for an AI solution?
12345Mean
2%4%19%43%33%4.02
Programme mean: 4.1/5 · 18 cohorts
Leaving today, how comfortable are you advising clients on AI when questions arise?
12345Mean
——11%56%33%4.22
Programme mean: 4.2/5 · 18 cohorts
How confident are you in your ability to lead a conversation about Anthropic and Claude with a client?
12345Mean
—2%7%46%44%4.33
Programme mean: 4.2/5 · 18 cohorts
Confidence Trajectory by Persona
D1→D2 delta · apply-AI confidence · matched pairs only · n=49
Positive delta = confidence grew · negative = dropped
Architect (n=13)
+0.46
Developer (n=14)
+0.21
Transformation Lead (n=22)
+0.59
What one takeaway will you share with a colleague or client?
n=39 responses
Participants prioritize model selection and prompt optimization over model switching as leverage points for better outcomes.
“The model is rarely the problem”
Participants recognize evaluation rigor and architectural complexity around inference as critical, underestimated components of implementation.
“the eval and the complexity of the architecture surrounding models”
Participants identify agent harness design and optimization as where significant cost and performance improvements materialize beyond model capabilities.
“a lot of work can be done in the harness to make better and cost effective agents”
Participants value applied, collaborative learning over theoretical content for embedding practical AI engineering skills.
“The hands on aspect was great”
Participants learn that actual deployment costs require empirical measurement across batching, multi-turn interactions, and parallel execution patterns.
“The cost of running is not strictly input plus output tokens, because of batching and multiple turns and parallel tool calls and such. The only way to estimate costs is to run it and measure.”
What will you build for yourself or apply at work in the next 30 days?
n=42 responses
Participants are moving from learning to implementation, prioritizing multi-agent architectures and orchestration workflows.
“Continue to improve agent orchestration and workflows”
Prompt engineering and prompt caching emerged as immediately actionable, with teams already investigating caching implementations.
“The biggest takeaway is prompt cacheing. Something I wasn't aware of and already I have my team looking into it.”
Participants recognize evaluation frameworks as critical, focusing on testing baselines and cost optimization levers.
“Testing/Baseline &incorporation of levers to make things run as inexpensively as possible”
Participants are identifying specific operational tools—PMO, project management, infrastructure discovery—as entry points for Claude integration.
“Mostly project management and infrastructure discovery and design related tooling.”
Participants see dual-track opportunities: external client solutions and internal team productivity through multi-agent customer motions.
“Beyond client solutions, apply Claude for internal work for our teams.”
Partner Basecamp Survey Report · Cohort 25 · Generated by analyze.py v2.5
NPS Reason
n=47 responses · grouped by NPS segment
Promotersscore 9–10 · 25 responses
1It covers good breadth in a practical/ hands-on way10
2I learnt a lot on Day-210
3I learned so much about the possibility’s of Claud10
4Lot of hands-on activities10
5Very detailed and engaging. Lots of good materials10
6Class gave a lot of good information.10
7Insights for production-grade agentic systems10
8Good balance of timing, intensity, practice10
9.Hands on10
10The sessions were engaging and informative10
11Learning, short & interactive10
12Very insightful, methodical10
13Really worth it, rich content10
14The sessions were very interesting and help me better structure my approach when working with clients10
15I learned the power of Claude!10
16As someone with a more strategy/product owner background with very little (zero) developer or engineer experience, it was a great experience to understand more how the tools we use and implement work and to walk away with a better understanding of how the engineers I work with think through solving problems.10
17I’m an engineer so I was comfortable with the coding walkthroughs but it’s probably heavy for non-technical people9
18Great engagement and content9
19As someone who is not a programmer I think some more foundational knowledge like more proficiency with GitHub and VS should be required. Otherwise class was great.9
20Very informative and knowledgeable.9
21Hands on practice A lot of mazing people Could not be technically challenging for experience folks9
22Excellent framing & quite grounded in real world.9
23Good instructions and hands on exercises helped understanding9
24There is a bit of guessing what the correct actions are within several of the training sessions. A cheat sheet should be included with the day 2 activities. Half my evaluation did not work and yes, I asked Claude for suggestions to no avail.9
25To learn building blocks and exploring the black box so it’s easy to use. This was the first building block and great to know what matters to build an AI solution.9
Passivesscore 7–8 · 19 responses
1Great opportunity to become more conversant and impactful on how to align efforts around ai projects/investments8
2It’s engaging and great learning8
3Technical issues outside Anthropic’s control.8
4I found the content extremely informative. I think office hours for set up would have been nice so we could focus the full time on the excercises8
5I would recommend it to technical people that are not actively working in AI. For me who does it was a bit more of a reminder and a way to piece things together. To someone else it could be the greatest program. Others who weren’t technical I see them struggling.8
6architectural knowledge and hands on workshop8
7Because it was quite informative with useful resources. I learned something new!8
8I already was familiar with most of content8
9Better understanding on AI agents.8
10Good facilitating team who knew their stuff. Felt awkward going through Jupyter notebooks that don’t talk well with AI, and basically asking Claude to fill them for me. Oh, and great food!8
11Accountability skills for all AI developers7
12The sessions were good, but would be great if they were actually build alongside where instructor is going through things live while we follow along.7
13While a great session, I was more interested in knowing how we can use cluade better to generate better outcomes. More hands on and practical experiences to show it real time vs excercises7
14Chance for folks to get together on the topic and think creatively about the challenges we and others are facing.7
15content is relevant but the level is a bit too basic for Frontier AI delivery teams. sufgest breaking the attendance into different levels7
16Technical in nature and could use more time upfront for basic concepts7
17I liked the engagement part of it the hands on could be honestly async7
18Really useful once everything was set up correctly7
19I think the training was helpful and eye opening. For those less familiar with or using VS code for the first time, I think a quick hour training on VS code might be helpful to get familiar with it.7
Detractorsscore 0–6 · 3 responses
1I think from a tech pm perspective, i would need the pre course before attending this. The topics are interesting but I have a hard time following along. Also my company laptop blocked most things so I think you should include in pre read that people should bring their personal laptop6
2On the PwC end, we need to be able to work out access before we come. I was never really able to run the modules. At some point, we have to ask: what is the point when we’re not able to run anything? I would highly recommend a pre-training module to ensure that someone can access before they come.5
3Mismatch between skillset - I serve a functional role and have very very limited technical knowledge. Additionally, faced many technical problems so don’t end up completing any of the exercises.4
Most Valuable
What one takeaway will you share with a colleague or client? · n=39 responses
1Collaborate and work
2How to use Claud code
3Give a try and this will change your assumptions on AI
4The model is rarely the problem
5Great session for technical folks
6different ways to improve prompt/context engineering
7Evals and inference is important
8How to optimize our agents.
9Evals
10Pick the right model for the task, optimize token usage
11There are levers to help seployments be successful. Find them.
12Cashing prompts and codes to improve efficiency
13a lot of work can be done in the harness to make better and cost effective agents
14Solutioning is so much more than developing a prompt to ensure the right level of value
15Let’s work on practices and governance, finally
16It’s never the model
17the eval and the complexity of the architecture surrounding models
18Eval adds so much value in your solutions
19Evals are really important
20Context engineering is important than pricier models
21Claude is awesome
22Model is not a problem Tokenomics in eval
23Evals to gain the confidence on AI Models
24I would say very valuable knowledge
25AI development is still a very technical and data heavy skillset. Non-developers who are vibe coding are just scratching the surface with demos, etc
26Let’s build the right solution for the use case with the right approach & tools.
27The hands on aspect was great
28Its a good session to go and familiarize with claude
29Let’s add contextual engineering
30How to use Claude with eclipse
31Evals, context engineering
32Claude's speed and ease of use.
33How to arrive at the correct or best model
34Keep learning and never give up if things don’t run
35Better planning for the care who will be onsite together. Quite a few of us thought we had everything "baked" before arriving only to learn we didn't.
36To PwC, don’t go if you can’t ensure Claude Code works first.
37Agents are getting more and more capable and are ready to add more skills
38The cost of running is not strictly input plus output tokens, because of batching and multiple turns and parallel tool calls and such. The only way to estimate costs is to run it and measure.
39It’s not always model.. analyze in depth before looking to change model
30-Day Intentions
What will you build for yourself or apply at work? · n=42 responses
1I want to implement multi agent
2My first agent
3The GAM metric auto run copilot
4Hands on!
5Continue to improve agent orchestration and workflows
6Use best practices when building ai agents
7Mostly project management and infrastructure discovery and design related tooling.
8Personal Assistant Agent
9Lots of applications
10Building agents to embed into our project delivery
11A web baseball game. Please give me tokens! Haha.
12Testing/Baseline &incorporation of levers to make things run as inexpensively as possible
13Model optimization
14Continuously working to be a better at prompt engineering
15agentic sales agents
16Numerous projects in mind personal and professional
17Focusing on prod-readiness of few pilots
18The biggest takeaway is prompt cacheing. Something I wasn’t aware of and already I have my team looking into it.
19agent economics solution
20Daily diary
21Evanls evals evals
22Eval
23Agentic solutions
24Data review solution & social entrepreneurship
25Agents which can transform legacy to modernized apps.
26App and logic reasoning tool
27Invoice monitoring agent dashboard
28Beyond client solutions, apply Claude for internal work for our teams.
29The hackathon project has great benefit in our internal org. I will continue to build.
30Integrated stock analysis and advisor systems
31I’ll utilize Claude every day and see what ways it can be helpful to my current - maybe a PMO tool
32Will keep studying about AI based development using Evals etc to get confident with score to be able to be explain it to customers
33Playwright testing for a client web app
34I will use Claude to perform tasks. Not necessarily agents.
35Customer sales motions. Multiple agent use cases
36We have automation scripts that I would like to streamline to see what Claude comes up with.
37I want to build a daily snapshot for the lead partner I work with that can give them daily or weekly updates on all their clients whether that be current news relevant to the client, mergers and acquisitions, news local to their community or others to give a personal and business snap shot to help our leaders spark conversation.
38I want to build security group and sandbox management application that checks salesforce qualifications
39I have an outline of a healthcare caregiver sentiment tool I will "play" with over the next 30 days.
40I will try to get my access sorted out with PwC.
41Use Claude for testing and analyzing data from my client’s AI voice and chat agent
42Couple of automation of daily task which can be handed over to my Claude coworker