Can a Company Build Its Own RL Environment?
The slides and full script for this video
Will an RL environment save your company time and money? The business case first: what AI did for 5,172 support agents, what an untested chatbot cost an airline, and why practicing on your own past work before customers see it cuts errors and speeds up rollout. Then the technical side: the 3 parts, and how an old claim becomes a practice task with its outcome as the checker. Sources in this video: the Bank of England and FCA survey of AI in UK financial services (November 2024); Brynjolfsson, Li and Raymond, Generative AI at Work, Quarterly Journal of Economics (May 2025); Moffatt v. Air Cana...
What the video covers
- 0:00 Can a company build its own RL environment?
- 0:19 RL in one sentence: learning by trial and error
- 0:34 Most UK financial firms already use AI
- 1:09 Who I am: broker and technical advisor
- 1:30 The payoff: support agents solved 15% more per hour
- 1:58 Moffatt v. Air Canada: the cost of an untested chatbot
- 2:38 Practice on your own past cases first
- 3:13 Where the time and money is saved
- 3:41 The 3 parts of an RL environment
- 3:55 Example: an old insurance claim becomes a task
- 4:21 Where it does not fit
- 4:44 4 signs your company could build one
- 5:09 Cheaper than testing on your customers
