Participant feedback
10 responses from 23 participants (43%), collected 25-27 September. All ratings on a 1-5 scale.
Ratings distribution
Overall training rating
Trainer rating
Practical ways to work more effectively
Confidence shift, one line per participant
The three who did not move were already at 4/5 (two of them) or self-describe as needing more advanced content. Nobody went down.
What they said, grouped
Prompting was the takeaway
RoCoTaFo, structuring the ask, specificity, refining the first prompt rather than redoing it, challenging the output. Verification and the Pyramid Principle were the other two "most useful" answers.
Asks for what comes next
A follow-up or next-level session (2), hands-on prompt exercises, team-specific sessions (ChatGPT Excel add-in, Claude for Finance, connectors), an internal AI champion, and training on respectful internal use of AI output ("AI slop" and the reviewer's time cost).
Full day is too long
Three ask for 2 x half days or a half day; one wants it longer and split across a week. Same diagnosis, opposite prescription: the format, not the content, was the strain. Friday all-day is named twice.
Afternoon less relevant
Image and slide generation "least relevant for my role"; avatars section disliked (the Chinese shopping-channel clip felt "a little too real"); research techniques "a bit too deep". The morning was consistently rated the stronger half.
Tool access gap
"Excited about certain tools but know we can't use them." Claude is on request only; ChatGPT is the sanctioned tool. Both 3/3 raters cite this. One suggests speaking to Atul directly.
One pointed critique
Trainer "enthusiastic and clearly current", but the strategic framework "felt shoehorned into some other tired McKinsey models"; would ask him to "up the tempo" if re-engaged. Same respondent already knew the practical content.
So what
- Format: the loudest, most consistent signal. Offer BritBox 2 x half days (or morning-only core + optional afternoon) for the next cohort. It also answers the "difficult to juggle meetings" point.
- Scope the tool list to what they can use. Lead with ChatGPT (their sanctioned tool), summarise the rest in one slide. Both low raters would likely move to 4 on this alone.
- Trim or make optional: image/slide generation and the avatar section, for a finance/ops-heavy audience. Keep verification and Pyramid Principle, which landed well.
- Upsell surfaced by participants themselves: follow-up / advanced session, team-specific clinics, and a "respectful use of AI internally" module. Worth putting in front of the sponsor.
Score sheet
| # | Training | Trainer | Conf. before | Conf. after | Shift | Practical | Most useful |
|---|---|---|---|---|---|---|---|
| 1 | 3 | 3 | 2 | 3 | +1 | 3 | RoCoTaFo prompting |
| 2 | 5 | 5 | 1 | 3 | +2 | 4 | Prompting |
| 3 | 5 | 5 | 4 | 5 | +1 | 4 | A framework to expand existing AI use |
| 4 | 4 | 5 | 3 | 4 | +1 | 4 | Building prompts for the most helpful output |
| 5 | 4 | 4 | 3 | 4 | +1 | 4 | Prompting for best results |
| 6 | 4 | 4 | 3 | 4 | +1 | 4 | Verifying output |
| 7 | 4 | 5 | 2 | 4 | +2 | 4 | Tool pros and cons, structuring asks, frameworks |
| 8 | 3 | 3 | 3 | 3 | 0 | 2 | Focusing and refining prompts |
| 9 | 4 | 4 | 4 | 4 | 0 | 4 | Pyramid Principle |
| 10 | 4 | 4 | 4 | 4 | 0 | 4 | Prompts |
| Avg | 4.0 | 4.2 | 2.9 | 3.8 | +0.9 | 3.7 |
Verbatims
Will do differently: RoCoTaFo.
More / less: More time on tools available to us, feel excited about certain tools but know we can't use them. Felt like we went a bit too deep on the research techniques.
Will do differently: Think and add AI into my workflows.
More / less: The course just needs to be longer - maybe split over a week.
Support: Time to experiment.
Else: Fred is fab and we need more regular training on AI use. Thank you!
Most useful: I have been using AI regularly at work, mostly to manage productivity but it's only now I realise how much more I can do with it. Today's training gave me a very good framework to expand my usage of AI.
More / less: More hands-on exercises of prompts.
Else: Highly enjoyable and extremely helpful training with practical ideas how to improve AI knowledge.
Will do differently: Think more about structure of what I am asking an LLM for and how to challenge it.
Will do differently: Prompt in a more clear, direct way.
More / less: I do feel the course could be condensed to a morning or two sections over a few days. It was a lot to take in over 8 hours and difficult to juggle meetings that needed to take place. 2 x half days would be more useful and mean that you do not get so tired or overloaded with information.
Support: A follow-up session.
Will do differently: Improved verification.
Support: PLEASE can we implement training for BritBox on respectful use of AI i.e. not distributing AI slop internally, how to respect the recipient of the internal documents you create using AI and so on. There is a clear but undiscussed net productivity loss if creating materials saves one person time but takes someone else much longer to review and interrogate. BritBox is behind the curve on this.
Will do differently: Using the frameworks - Blue Ocean, Pyramid etc.
More / less: I have a lot of survey data and tracking data and am interested in building a chatbot or agentic AI to query the data in natural language that I could then share with stakeholders to use safely. But do I have access to the right tools to be able to do this?
Support: A champion that does little sessions here or there, or a dedicated person that can help with the above idea.
Else: Not all the tools are available to us, especially Claude which seems to be on a request basis. Maybe one to speak to Atul directly. Also very long day and found myself waning - 2 x half days? But appreciate that it's not easy to arrange with so many of us.
Will do differently: Put more work into my initial prompts so that I get a good starting point, rather than so much re-doing the request.
More / less: I feel a bit mean saying this, but the last hour and a half was painful! I have literally no practical application for most of the stuff after the break and we can't use most of the apps referenced anyway. The main takeaway was about how to be better at prompts - this was useful but covered pretty comprehensively in the morning, so the course could be completed in half a day. Concentrate only on the tools we are able to use; just do a summary of the others.
Support: A shorter general session and more specific skills sessions for teams - the ChatGPT Excel add-in, Claude for Finance (currently limited), connectors (if available). Creating images and videos is unlikely to be relevant for finance.
Else: The morning was better than the afternoon - useful insights into building prompts but I don't think I learned anything new that I can apply to my work.
Most useful: I had not come across the Pyramid Principle before, which is a useful heuristic.
Will do differently: I was already aware of all of the practical advice on offer today so probably won't do anything differently, but it is good that everyone is getting a baseline level of training.
More / less: While sexier than learning how to write prompts, generating images and slides is the least relevant for my role at BBI. We won't be using GPT for any creative development anytime soon, especially slides, so that section could have simply been skipped.
Support: I need more advanced training than what was shared today; a follow-up or next level would be valuable.
Else: Fred is enthusiastic and clearly current across the landscape of AI tools. He's personable and friendly, but I question some of the business sense: the strategic framework he presented was not strategic, it felt shoehorned into some other tired McKinsey models. Admittedly, we weren't offering much in the way of participation - an all-day Friday training is pretty rough - so I give him credit for keeping his energy up. It's not an easy brief, considering the wide-ranging experience within the group, so he had to be patient. If there were a follow-up and he was being considered, I'd ask him to up the tempo. Thanks for organising, the wider exposure to this technology is good.
Most useful: I loved learning about prompts. That was very insightful!
Will do differently: Calibrate my AI using multiple source points, while ensuring I do not use any BBC/BritBox proprietary information or data that is not already open source or approved by the BBC.
More / less: I really enjoyed the training overall, but I didn't like the final section on avatars and creating AI versions of yourself. Starting with the video of AI avatars on the Chinese shopping channel felt a little too real for my personal comfort level, and I found the whole concept rather bizarre. To be clear, it was just the avatar section that I disliked. The rest of the training was 10/10!
Else: As I told my boss, this was worth the full day of training. I took tons away from it.