vidtheque.

read-only demo

Videos 6bmM45jkMDY

You Can't Prompt the Room: The Last Skill AI Won't Replace - Balázs Horváth, VisualLabs

index_state ready data_status ok

AI Engineer· published 2026-06-29· 0:15:45· en· indexed 2026-08-11 05:38

Open on YouTube

Scene timeline

  1. Shot 0, 0:00 to 0:34, 1 of 1 keyframes kept
  2. Shot 1, 0:34 to 1:04, 1 of 1 keyframes kept
  3. Shot 2, 1:04 to 1:32, 1 of 1 keyframes kept
  4. Shot 3, 1:32 to 1:59, 1 of 1 keyframes kept
  5. Shot 4, 1:59 to 2:26, 0 of 1 keyframes kept
  6. Shot 5, 2:26 to 3:04, 1 of 1 keyframes kept
  7. Shot 6, 3:04 to 3:30, 1 of 1 keyframes kept
  8. Shot 7, 3:30 to 3:55, 1 of 1 keyframes kept
  9. Shot 8, 3:55 to 4:21, 1 of 1 keyframes kept
  10. Shot 9, 4:21 to 4:47, 1 of 1 keyframes kept
  11. Shot 10, 4:47 to 5:12, 1 of 1 keyframes kept
  12. Shot 11, 5:12 to 5:37, 1 of 1 keyframes kept
  13. Shot 12, 5:37 to 6:02, 1 of 1 keyframes kept
  14. Shot 13, 6:02 to 6:28, 1 of 1 keyframes kept
  15. Shot 14, 6:28 to 6:57, 1 of 1 keyframes kept
  16. Shot 15, 6:57 to 7:26, 1 of 1 keyframes kept
  17. Shot 16, 7:26 to 7:55, 1 of 1 keyframes kept
  18. Shot 17, 7:55 to 8:21, 1 of 1 keyframes kept
  19. Shot 18, 8:21 to 8:46, 1 of 1 keyframes kept
  20. Shot 19, 8:46 to 9:11, 1 of 1 keyframes kept
  21. Shot 20, 9:11 to 9:36, 0 of 1 keyframes kept
  22. Shot 21, 9:36 to 10:02, 1 of 1 keyframes kept
  23. Shot 22, 10:02 to 10:27, 0 of 1 keyframes kept
  24. Shot 23, 10:27 to 11:12, 1 of 1 keyframes kept
  25. Shot 24, 11:12 to 11:40, 1 of 1 keyframes kept
  26. Shot 25, 11:40 to 12:08, 1 of 1 keyframes kept
  27. Shot 26, 12:08 to 12:37, 1 of 1 keyframes kept
  28. Shot 27, 12:37 to 13:05, 1 of 1 keyframes kept
  29. Shot 28, 13:05 to 13:32, 1 of 1 keyframes kept
  30. Shot 29, 13:32 to 13:59, 1 of 1 keyframes kept
  31. Shot 30, 13:59 to 14:25, 1 of 1 keyframes kept
  32. Shot 31, 14:25 to 14:52, 1 of 1 keyframes kept
  33. Shot 32, 14:52 to 15:17, 1 of 1 keyframes kept
  34. Shot 33, 15:17 to 15:42, 0 of 1 keyframes kept
  35. Shot 34, 15:42 to 15:44, 1 of 1 keyframes kept

35 shot(s).

keyframes kept every frame deduplicated

What was stored

cues
124
whisperx 124
chunks
28
from 124 cues
keyframes
31
kept of 35 captured
frames with text
31
249 lines read
chapters
0
from the source metadata
keyframe bytes
2.0 MB
word timings on 124 cues

Provenance

Each pipeline stage, its state and the model that produced it
stage state model started took
fetch done — 2026-08-11 05:36 1m 11s
stt done — 2026-08-11 05:37 16s
chunk done — 2026-08-11 05:37 0s
text_embed done — 2026-08-11 05:37 0s
keyframe done — 2026-08-11 05:37 29s
ocr done — 2026-08-11 05:38 7s
frame_embed done — 2026-08-11 05:38 4s

Frames, and what the machine read

  • 0:30 #0 done5 line(s)

    shot 0·sharpness 325.1

    1. AI.ENGINEER1.00
    2. You can't prompt1.00
    3. the room.1.00
    4. Cheap to build. Expensive to decide.0.99
    5. Balazs Horvath1.00

  • 1:01 #1 done6 line(s)

    shot 1·sharpness 180.7

    1. WHERE THIS STARTS0.97
    2. 171.00
    3. of 21 agents1.00
    4. abandoned1.00
    5. On data access. None of them reached the model.1.00
    6. Our hackathon0.98

  • 1:08 #2 done9 line(s)

    shot 2·sharpness 607.3

    1. THE BRIDGE1.00
    2. I've always prompted1.00
    3. the developers.1.00
    4. The room was always the hard part.0.99
    5. 13 years, ERP & CRM0.98
    6. US·UK·Hungary0.99
    7. VisualLabs, Microsoft partner1.00
    8. 031.00
    9. → navigate · F fuliscreen0.89

  • 1:48 #3 done9 line(s)

    shot 3·sharpness 609.4

    1. THE BRIDGE1.00
    2. I've always prompted1.00
    3. the developers.1.00
    4. The room was always the hard part.0.98
    5. 13 years, ERP & CRM0.98
    6. US·UK·Hungary0.99
    7. VisualLabs, Microsoft partner1.00
    8. 031.00
    9. → navigate · F fullscreen0.92

  • 2:02 #4 skipped

    shot 4·duplicate of #2

  • 3:00 #5 done8 line(s)

    shot 5·sharpness 312.2

    1. prompt1.00
    2. your code0.99
    3. prompt1.00
    4. your Al0.93
    5. prompt1.00
    6. the whole spec1.00
    7. You can't prompt1.00
    8. the room.1.00

  • 3:22 #6 done7 line(s)

    shot 6·sharpness 277.5

    1. WHY A MODEL CAN'T DO IT0.98
    2. Ask it what to build.1.00
    3. You get a faster horse.0.99
    4. what already exists1.00
    5. the average0.97
    6. a faster horse1.00
    7. after Henry Ford0.98

  • 3:33 #7 done7 line(s)

    shot 7·sharpness 280.0

    1. WHY A MODEL CAN'T DO IT0.98
    2. Ask it what to build.1.00
    3. You get a faster horse.1.00
    4. what already exists1.00
    5. the average0.96
    6. a faster horse1.00
    7. after Henry Ford1.00

  • 4:13 #8 done4 line(s)

    shot 8·sharpness 322.4

    1. THE CRAFT COMES BACK1.00
    2. The analyst's toolkit1.00
    3. is senior work now.1.00
    4. Story mapping is one example among several frameworks.0.99

  • 4:31 #9 done4 line(s)

    shot 9·sharpness 325.6

    1. THE CRAFT COMES BACK1.00
    2. The analyst's toolkit0.98
    3. is senior work now.1.00
    4. Story mapping is one example among several frameworks.0.99

  • 5:04 #10 done16 line(s)

    shot 10·sharpness 151.5

    1. THE STORY MAP1.00
    2. Contact1.00
    3. Triage1.00
    4. Resolve or route1.00
    5. Close1.00
    6. RELEASE 11.00
    7. Capture intent1.00
    8. Classify urgency1.00
    9. Draft a grounded answer1.00
    10. Log to system of record0.98
    11. LATER1.00
    12. Read sentiment1.00
    13. Route to a team1.00
    14. Suggest next action1.00
    15. Check satisfaction1.00
    16. 081.00

  • 5:32 #11 done16 line(s)

    shot 11·sharpness 154.8

    1. THE STORY MAP1.00
    2. Contact1.00
    3. Triage1.00
    4. Resolve or route1.00
    5. Close1.00
    6. RELEASE 11.00
    7. Capture intent1.00
    8. Classify urgency1.00
    9. Draft a grounded answer1.00
    10. Log to system of record0.98
    11. LATER1.00
    12. Read sentiment1.00
    13. Route to a team0.97
    14. Suggest next action1.00
    15. Check satisfaction1.00
    16. 081.00

  • 5:47 #12 done16 line(s)

    shot 12·sharpness 152.4

    1. THE STORY MAP1.00
    2. Contact1.00
    3. Triage1.00
    4. Resolve or route1.00
    5. Close1.00
    6. RELEASE 10.99
    7. Capture intent1.00
    8. Classify urgency1.00
    9. Draft a grounded answer1.00
    10. Log to system of record0.98
    11. LATER1.00
    12. Read sentiment1.00
    13. Route to a team0.99
    14. Suggest next action1.00
    15. Check satisfaction1.00
    16. 081.00

  • 6:18 #13 done16 line(s)

    shot 13·sharpness 153.4

    1. THE STORY MAP0.98
    2. Contact1.00
    3. Triage1.00
    4. Resolve or route1.00
    5. Close1.00
    6. RELEASE 11.00
    7. Capture intent1.00
    8. Classify urgency1.00
    9. Draft a grounded answer1.00
    10. Log to system of record1.00
    11. LATER1.00
    12. Read sentiment1.00
    13. Route to a team1.00
    14. Suggest next action1.00
    15. Check satisfaction1.00
    16. 081.00

  • 6:31 #14 done7 line(s)

    shot 14·sharpness 319.3

    1. ONE STORY1.00
    2. As a support lead0.98
    3. I need open cases ranked by urgency1.00
    4. so that none of the escalations slip1.00
    5. persona whose problem1.00
    6. what the need1.00
    7. why the value0.97

  • 7:00 #15 done7 line(s)

    shot 15·sharpness 318.8

    1. ONE STORY1.00
    2. As a support lead1.00
    3. I need open cases ranked by urgency1.00
    4. so that none of the escalations slip1.00
    5. persona whose problem1.00
    6. what the need1.00
    7. why the value1.00

  • 7:30 #16 done7 line(s)

    shot 16·sharpness 318.9

    1. ONE STORY1.00
    2. As a support lead1.00
    3. I need open cases ranked by urgency1.00
    4. so that none of the escalations slip1.00
    5. persona whose problem0.98
    6. what the need1.00
    7. why the value0.98

  • 8:03 #17 done8 line(s)

    shot 17·sharpness 397.8

    1. BEFORE THE AGENTS START1.00
    2. Whose problem is this?1.00
    3. What does winning look like for them?0.99
    4. 31.00
    5. What would make them refuse to use1.00
    6. What decision does it change?1.00
    7. it?1.00
    8. 101.00

  • 8:31 #18 done6 line(s)

    shot 18·sharpness 399.7

    1. BEFORE THE AGENTS START0.98
    2. Whose problem is this?1.00
    3. What does winning look like for them?0.99
    4. What would make them refuse to use1.00
    5. What decision does it change?1.00
    6. it?1.00

  • 9:03 #19 done6 line(s)

    shot 19·sharpness 397.8

    1. BEFORE THE AGENTS START1.00
    2. Whose problem is this?1.00
    3. What does winning look like for them?1.00
    4. What would make them refuse to use0.99
    5. What decision does it change?1.00
    6. it?1.00

  • 9:21 #20 skipped

    shot 20·duplicate of #19

  • 9:42 #21 done9 line(s)

    shot 21·sharpness 332.6

    1. ONE PROBLEM, ALL THE WAY THROUGH0.99
    2. "Build us an agent that handles support."1.00
    3. Value1.00
    4. Process1.00
    5. Stories1.00
    6. frame the real need0.99
    7. how the work actually flows1.00
    8. precise, sequenced, shared1.00
    9. We call the shape VAD · Value » Architecture » Design0.97

  • 10:24 #22 skipped

    shot 22·duplicate of #21

  • 11:02 #23 done4 line(s)

    shot 23·sharpness 375.9

    1. THE QUESTION YOU'RE ALREADY ASKING1.00
    2. "Isn't this just1.00
    3. product management?"0.97
    4. Old skill. New economics.0.98

Transcript

124 cues· 2,225 words· 12,001 chars

  1. 0:01 Hi, everyone.
  2. 0:03 I am Balazs Horvath, and today I will talk to you about what is the last thing that AI will take away from us as people in the software business.
  3. 0:13 So at a point where writing code is no longer the bottleneck, the real thing is figuring out what it is that you should be building.
  4. 0:26 And that comes down to people skills and being able to work the room because you can't prompt a room.
  5. 0:32 You can prompt your AI.
  6. 0:35 So at the beginning of the year, we held an internal hackathon where we had about 21 agents, agent ideas, and 17 of those were abandoned because they actually created no business value.
  7. 0:51 They either didn't have data access or just didn't make sense to build it.
  8. 0:59 And those four were the ones that actually had a very big impact on how we work today.
  9. 1:05 And it's a very good example of...
  10. 1:10 of just making sure that we are building what is worth building.
  11. 1:15 And throughout my career in the past 30 years, I've always been the bridge between business and IT and the developers.
  12. 1:27 I started writing, well, initially testing functional designs, specifications, and then I wrote them.
  13. 1:35 And as a functional consultant, I've worked with large ERP and CRM programs in the US and the UK, and then I founded Visual Labs.
  14. 1:44 And essentially,
  15. 1:46 I've trained my team on how to elicit those requirements in a way that we can turn them into good specifications for developers to build, for consultants to configure, and most recently for AI to build.
  16. 2:03 And what's not really changed over the years is how we interact with our customers, how we interact with systems, how we interact with AI is very much changing.
  17. 2:14 And that's the big thing now.
  18. 2:18 But if you can read the room, if you can elicit the right requirements, then you will be able to build more valuable software.
  19. 2:28 And that essentially the big shift over the past two, three years was that getting access to code and being able to build is no longer the bottleneck to the software development lifecycle.
  20. 2:40 Now, the real bottleneck is getting your people, your stakeholders, your decision makers into the room and being able to access them and elicit the requirement and being able to spend the time with them.
  21. 2:53 So that's the right, that's the real bottleneck.
  22. 2:55 figuring out what it is that should be built.
  23. 2:59 Because you can prompt your code, you can prompt your AI, you can prompt your whole specification, you can't prompt your room.
  24. 3:06 And
  25. 3:07 What a model can do is very similar to how Henry Ford's analogy of what he said about asking his users or his customers, if he'd asked them what it is that they needed, they would have said they needed more horses.
  26. 3:23 But in reality, he built a car and he made a very big success on them.
  27. 3:28 So if you're just using AI,
  28. 3:31 to make things, build things better, the chances are that you are replicating what already exists because AI by definition is coded to give you the most common answers.
  29. 3:46 So for us, the real job is to make sure that AI moves away from that average into what is better for us.
  30. 3:54 so we can just get to not a faster horse but actually produce a car that's a magnitude shift better than what we had.
  31. 4:04 So
  32. 4:05 It's really an interesting world where being able to write good code is no longer the most important skill to have.
  33. 4:16 Actually, the real skill now is becoming the analyst toolkit, which is things like story mapping, business model canvas, value canvas, and those good old things that we are so used to using as functional consultants, business analysts.
  34. 4:34 or in the world of design thinking.
  35. 4:39 So I'd like to zoom in on story mapping because that's the skill set that I found as the most valuable.
  36. 4:47 So once you have the story map with the backbones and understand at each step what your customers, your users are doing,
  37. 4:57 that would give them the ability to move forward in their processes.
  38. 5:04 So here's a support systems user story map, contacting, triaging, resolving, and then essentially closing the case.
  39. 5:14 With this, you can understand different stages of the process.
  40. 5:20 and then capture the user stories beneath them.
  41. 5:23 It is intended to stay at a fairly high level so you can get a big picture, and then you can decide what it is that you want to build and release.
  42. 5:33 One, like capturing intent, classifying urgency, drafting a grounded answer, and then logging it to a system of record.
  43. 5:41 That's essentially your MVP.
  44. 5:43 Those are the first things that you'd want to build.
  45. 5:45 And those are your first four user stories.
  46. 5:48 And beneath those, you've got the second set of user stories like reading a sentiment, writing to a team, suggesting next action, checking satisfaction, so on and so forth.
  47. 6:01 Those will be part of your backlog.
  48. 6:03 So what would allow you to...
  49. 6:07 to get really good agentic results is by honing in on these user stories and making sure that you use these user stories as a means to elicit discussions with your stakeholders, with your business, and then work out what that user story should really be about.
  50. 6:28 So the first user story, second user story would be as a support lead, I need to open cases right by urgency so that none of the escalations slip.

Open at this second